Numerical Notation
This book is a crosscultural reference volume of all attested numerical notation
systems (graphic, nonphonetic systems for representing numbers), encompassing
more than 100 such systems used over the past 5,500 years. Using a typology that
defies progressive, unilinear evolutionary models of change, Stephen Chrisomalis
identifies five basic types of numerical notation systems, using a cultural phylogenetic framework to show relationships between systems and to create a general
theory of change in numerical systems. Numerical notation systems are primarily representational systems, not computational technologies. Cognitive factors
that help explain how numerical systems change relate to general principles, such
as conciseness and avoidance of ambiguity, which also apply to writing systems.
The transformation and replacement of numerical notation systems relate to specific social, economic, and technological changes, such as the development of the
printing press and the expansion of the global worldsystem.
Stephen Chrisomalis is an assistant professor of anthropology at Wayne State University in Detroit, Michigan. He completed his Ph.D. at McGill University in
Montreal, Quebec, where he studied under the late Bruce Trigger. Chrisomaliss
work has appeared in journals including Antiquity, Cambridge Archaeological Journal, and CrossCultural Research. He is the editor of the Stop: Toutes Directions
project and the author of the academic weblog Glossographia.
Numerical Notation
A Comparative History
Stephen Chrisomalis
Wayne State University
9780511679346
eBook (EBL)
ISBN13
9780521878180
Hardback
Contents
Acknowledgments
page vii
Introduction
Hieroglyphic Systems
34
Levantine Systems
68
Italic Systems
93
Alphabetic Systems
133
188
Mesopotamian Systems
228
259
Mesoamerican Systems
284
10
Miscellaneous Systems
309
11
360
Contents
vi
12
401
13
Conclusion
430
Glossary
435
Bibliography
439
Index
471
Acknowledgments
vii
viii
Acknowledgments
Further refinements and a new draft of the book were produced under a SSHRC
postdoctoral fellowship at the University of Toronto. While I was in Toronto,
Richard Lee, Trueman MacHenry, and David Olson were particularly helpful
to me and provided useful insights on the theories and concepts underlying my
work, forcing me to clarify my own positions in ways that I had not previously
done. Bob Bunker, John Gilks, Heather Hatch, Andy Pope, and Shana Worthen
read portions of the manuscript at this stage and provided very useful editorial
advice.
A work of this scope inevitably relies upon the individual and collective experience of regional specialists in the writing systems and mathematical practices of
various regions and periods, and of theorists working in cognitive and psychologically oriented anthropology and linguistics. I have benefited tremendously from
the specialized expertise of Priskin Gyula, Christopher Hallpike, Jim Hurford,
Joel Kalvesmaki, Eleanor Robson, Nicholas SimsWilliams, Matthew Stolper, and
Konrad Tuchscherer. A School of Advanced Research Advanced Seminar entitled
The Shape of Script was the key to moving my work into its final completed form,
and introduced me to many additional regional specialists whose advice has been
of assistance: John Baines, John Bodel, Beatrice Gruendler, Stephen Houston,
David Lurie, Kyle McCarter, John Monaghan, Richard Salomon, Kyle Steinke,
and Niek Veldhuis.
Scholars of numeration include historians, anthropologists, archaeologists, linguists, mathematicians, and psychologists, to name only a few, and it is all too
easy in such a disparate crowd of research traditions to lack a sense of disciplinary
cohesion and of ones scholarly influences. I therefore acknowledge my intellectual forebears in the comparative study of numerals, most notably Florian Cajori,
Genevieve Guitel, Karl Menninger, Antoine Pihan, and David Eugene Smith.
Although I disagree with his conclusions in many places, I thank Georges Ifrah,
whose gargantuan and important Histoire universelle des chiffres (1998) inspired me
to produce this volume.
Eric Crahan and Frank Smith at Cambridge University Press deserve great
credit for their skillful guidance of my work through the editorial process at all
stages. Russell Hahn guided the complex copyediting masterfully, and Leah Shapardanis prepared the index and read proofs. Many thanks to the fourteen anonymous reviewers who read and commented on one or more chapters on behalf of
the Press, and to the entirety of the production staff for their handling of dozens
of specialized typefaces.
To my family, all my love and thanks. Arthur Chrisomalis provided useful
firsthand insights into the childhood acquisition of lexical and graphic numeration, and rekindled his fathers wonderment at the magic of numbers. Finally, this
work is dedicated with love to my wife, Julia Pope, for her patience with me over
Acknowledgments
ix
the past decade, her keen editorial eye, her endless willingness to reread manuscript chapters, and her ongoing conviction that this work is worthwhile.
Despite the advice and assistance of the abovementioned, and any others I have
forgotten, I have doubtless made many errors of fact and interpretation, and I
eagerly anticipate the opportunity to broaden my knowledge of numerical notation systems in the future.
Notes on Style
Throughout the book I have used the conventions bc and ad to refer to chronological periods. Where no era indicator is associated, ad dates are assumed; I do
so only when the interpretation of a date is obvious.
chapter 1
Introduction
The Western world is a world of written numbers. One can hardly imagine an industrial civilization functioning without the digits 0 through 9 or a similar system.
Yet while these digits have pervasive social and cognitive effects, many unanswered
questions remain concerning how humans use numerals. Why do societies enumerate? How does the representation of numbers today differ from their representation
in the past? Why does the visual representation of number figure so prominently in
complex states? What cognitive and social functions are served by numerical notation systems? How do numeral systems spread from society to society, and how do
they change when they do so? And, despite their present ubiquity, why have the
vast majority of human societies not possessed them at all?
If you look up from this page and examine your surroundings, I am certain that
you will encounter at least one instance of numerical notation, probably more.
Moreover, unless you have a Roman numeral clock nearby, I am nearly certain
that all of the numerals you encounter are those of the HinduArabic or Western1
system. Numerals serve a wide variety of functions: denotation Call George,
1
The conventional term used in popular literature, Arabic numerals, and the term used
in most scholarly literature, HinduArabic numerals, can lead to considerable confusion because the scripts used to write the Hindi and Arabic languages use numerical
notation systems that differ from those of the West in the shape of the signs. I use the
term Western numerals to refer to this system because it developed in Western Europe
in the late Middle Ages, while fully acknowledging its Indian and Arabic ancestry.
Numerical Notation
8765000; computation 21.00 1.15 = 24.15; valuation 25 cents; ordination 1. Wash dishes, 2. Sweep floor, 3. Finish manuscript; and so on. Most of
the thousands of numerals we see each day barely register on our conscious minds;
regardless, we encounter far more written numbers in our lifetime than we do
sunsets, songs, or smiles. Until the past few centuries, the opposite was true for
most people.
These ten digits are so prevalent that it is easy to equate our numeralsigns
with the set of abstract numbers. In this view, 62 does not merely signify the
abstract concept sixtytwo it is the raw form of the number itself, the stuff of
pure mathematics (or perhaps pure numerology). That these signs are frequently
encountered and used in mathematical contexts contributes to the prevalence
of such attitudes. According to this view, our numeralsigns constitute abstract
number, and other systems (when recognized as such) are simply archaic deviations from the abstract entity comprised by these signs.
This view is erroneous, and rests on the confusion of a mental concept (signified) with its symbolic representation (signifier). Our numerical notation system has an extensive history, as do the more than one hundred systems that
have existed over the past five thousand years. Still, the worldwide prevalence of
Western numerical notation is undeniable. Most literate individuals worldwide,
as well as a sizable number of illiterates, understand them. Nor does any competing system have any reasonable chance of supplanting our system in the near
future. This has led many scholars to assert its supremacy solely on the evidence
of its nearuniversality (Zhang and Norman 1995; Dehaene 1997; Ifrah 1998).
Nevertheless, this situation does not imply that our system will dominate the
whole world forever. The study of numerical notation remains mired in a theoretical framework that has much more in common with late nineteenthcentury
unilinear evolutionism in anthropology than it does with early twentyfirstcentury critiques of unfettered scientific progress.
Despite this theoretical weakness, numerical notation as a topic of academic
study is a relatively common pursuit, with linguists, epigraphers, archaeologists,
anthropologists, historians, psychologists, and mathematicians all making significant contributions to the literature. These studies are mostly restricted to the
analysis of one or a few numerical notation systems, although a small number
of synthetic and comparative works dealing with numerical notation exist
(Cajori 1928; Menninger 1969; Guitel 1975; Ifrah 1998). However, such works
rarely consider more obscure numerical notation systems, such as those of subSaharan Africa, North America, and Central Asia. Similarly, social scientists such
as the anthropologist Thomas Crump (1990), the psychologist David Lancy (1983),
and the ethnomathematicians Marcia Ascher (1991) and Claudia Zaslavsky (1973)
have undertaken major comparative research on numeracy and mathematics in
Introduction
nonWestern societies. Yet numerical notation has not been a primary focus of
this body of research.
This study is a comparative analysis of all numerical notation systems known
to have existed throughout history approximately one hundred distinct systems,
most of which can be grouped into eight distinct subgroups. By presenting a
universal study of such systems and examining the historical connections and
contexts in which they are encountered, I will develop a framework that accounts
for cultural universals, identifies evolutionary regularities, and yet remains cognizant of idiosyncratic features, seeking to determine, rather than to assume, the
amount of intercultural variability among them. I will distinguish several major
types of numerical notation, evaluate their efficiency for performing specific functions, link their features to human cognitive capacities, and relate systems to their
sociopolitical contexts.
Definitions
A numerical notation system is a visual, relatively permanent, and primarily nonphonetic structured system for representing numbers. Signs such as 9 and 68, IX
and LXVIII, are part of numerical notation systems, but numeral words such as
nine and achtundsechzig are not. Though there are ties between numeral words
and numerical notation, a lexical numeral system, or the sequence of numeral
words in a language (whether written or spoken), has a languagespecific phonetic
component. Every language has a lexical numeral system of some sort, while numerical notation is an invented technology that may or may not be present in a
society.2 Some numerical notation systems contain a small phonetic component,
as in acrophonic systems whose signs are derived from the first letters of the appropriate numberwords in a language. However, since such systems are still comprehensible without having to understand a specific language, they are numerical
notation systems.
Numerical notation systems must be structured. Simple and relatively unstructured techniques, such as marking lines on a jailhouse cell to count ones days or
piling pebbles in a basket, are largely or entirely unstructured. They rely on onetoone correspondence, in which things are counted by associating them with an
equal number of marks or other identical objects. A numerical notation system,
by contrast, is a system of different discrete numeralsigns: single elementary symbols, or, in the terminology used in writing systems, graphemes, which are then
2
I will leave aside for the moment discussions of counterevidence questioning the assumption of the universality of lexical numeral systems (Hurford 1987: 6878; Gordon
2004; Everett 2005).
Numerical Notation
A few numeralsigns are more complex in that they graphically combine two or more
signs into one in order to represent multiplication, but they are treated as elementary
numeralsigns because their use is identical to that of all other simple signs in the systems in question.
Introduction
use of numbers (number theory being the most obvious example), large parts
of the discipline have only infrequent or peripheral encounters with numerical
notation. Numerical notation systems are not necessarily designed with mathematical purposes in mind. Even in contemporary industrial societies, where
mathematical ability is more extensive than in any other historical or modern
society, most numerical notation is nonmathematical.
Universal Comparison
The present study is, as far as possible, a universal one. I have not excluded any numerical notation system intentionally save where data are not plentiful enough to
undertake a reasonable analysis. Most comparative research in anthropology aims
to discover generalizations and patterns in human behavior, but using the universe
of cases is neither possible nor desirable in most crosscultural studies. In order to
use most analytical statistics on crosscultural data, each case must be independent
of the others, which requires that each case may not be historically derived or diffused from any other case. This issue, known as Galtons problem, is the thorniest
methodological issue in statistical crosscultural research (Naroll 1968: 258262).
The establishment of correlations between traits among historically independent
societies is enormously useful, and is the basis for most crosscultural research in
modern anthropology.
Yet to do so in a study such as this one, in which there are perhaps only seven
independently invented numerical notation systems, would be pointless. Firstly,
seven cases would be too small a sample to analyze statistically. Secondly, by
studying all cases, I am able to show that the total observable variability among
numerical notation systems is far greater than has previously been believed. This
variability cannot be understood by studying only a fraction of numerical notation systems. To paraphrase the old fable, if we study only the elephants trunk
or tail, we ignore most of the animal. Thirdly, I wish to explain structural variation among historically related systems, which frequently differ considerably from
their relations. This would be impossible using a sampling technique that omitted
related cases. Finally, were I to omit related cases, I could not analyze how systems
change over time or how new systems develop out of existing ones. By taking
events of change, rather than static systems, as the units of analysis in my comparisons, I am able to elucidate both synchronic and diachronic patterns among
numerical notation systems. It is worth noting that Galtons problem does not
apply to events of change of the sort I am analyzing, since every event is essentially
independent of every other, and can thus be analyzed statistically, where relevant.
I reject as false the dichotomy in anthropology between universalism (Tylor 1958
[1871], White 1949, 1959; Steward 1955; Harris 1968) and relativism (Lowie 1920,
Numerical Notation
Boas 1940, Sahlins 1976, Geertz 1984), both of which presume rather than evaluate
the degree of regularity that social phenomena display. While numerical notation
systems display remarkable regularities and even universals, historical contingencies also played a major role in shaping the cultural history of numerical notation.
Yet the only way to determine which features of numeration are crossculturally
regular and which are idiosyncratic is to undertake crosscultural comparison. The
best way to deal with the messiness of the world less universal than universalists
would like, less relative than relativists prefer is through a body of theory that
deals with constraints.
Most anthropological theory is predicated on the existence of very strong constraints on the forms possible within human societies. Some of these constraints
are so strong as to produce crosscultural universals (Brown 1991). Most cultural
relativists dismiss these universals as minimally true, but facile, irrelevant, and useless for understanding humanity (cf. Geertz 1965, 1984). The denial of comparativism on this basis is an overly negative position, given that those who criticize
comparativism most harshly are very often those who have not undertaken it. One
of the most crucial theoretical contributions of anthropology should be to indicate
the degree to which human societies are alike and the degree to which they differ.
While some aspects of human existence are truly universal, and others are almost
infinitely variable, most of the really intriguing domains of activity fall somewhere
in the middle.
In the early 1900s, Alexander Goldenweiser developed his principle of limited possibilities, which stated that for any social or cultural phenomenon,
there are a limited number of possible forms that can be expressed in human
societies (Goldenweiser 1913). Goldenweiser was particularly interested in the
limitations imposed by human psychology on the expression of cultural traits,
although, given the inchoate nature of psychological theory at the time, he was
unable to describe these mechanisms precisely. Bruce Trigger (1991) has rejuvenated the idea of constraints, proposing that anthropologists should use the
concept of constraint to describe the limitations on human sociocultural variation whether those constraints are biological, ecological, technological, informational, psychological, or historical in order to analyze statistical regularities
among cultures without implying determinism. We must be cautious, with both
the limited possibilities and the constraint approaches, not to restrict our
formulations and assume the restricting influence of various factors to be more
important than positive (enabling) effects. A very strong propensity in favor
of some trait is not the same thing as a very strong constraint against all other
possibilities. Constraints and inclinations can and do coexist, and the negative
limitations of one variable must be weighed against the positive inclinations of
another. Despite this caveat, I find a constraintbased approach to be the most
Introduction
promising theoretical perspective for explaining the regularities found in numerical notation systems, something to which I will return in Chapters 11 and 12.
In much of my analysis, I follow Joseph Greenberg (1978), whose analysis of significant regularities in lexical numeral systems presents a list of fiftyfour generalizations. Unlike much of his later work, Greenbergs study of numerals is universal
and cognitive in orientation rather than phylogenetic. It is synthetic, based on the
detailed empirical work of earlier scholars, such as the German linguist Theodor
Kluge, who spent years compiling sets of numeral terms in languages throughout
the world (Kluge 193742). While many of Greenbergs regularities are extremely
complex4 or have some exceptions, others reveal truly universal and nontrivial
features of every natural language; for instance, every numeral system contains
a complete set of integers between one and some upper limit each system is
finite5 and has no gaps (Greenberg 1978: 253255). Similarly, no natural language
expresses two as ten minus eight or twenty as onefifth of one hundred.
While every language has a set of lexical numerals, most premodern societies
functioned quite well without numerical notation. It is possible to conceive of
a world in which there are many regularities in lexical numerals, but in which
numerical notation systems are highly specific and unique responses to local
needs. We do not live in such a world. There is considerable uniformity among
the worlds numerical notation systems, and they display many synchronic and
diachronic regularities.
In fact, the number and variety of conceivable numerical notation systems is
far greater than what is attested historically. To take only a very limited example, a
numerical notation system can very easily be imagined that is just like the Western
system but instead of being a decimal system having a base of any natural6
number of 2 or higher. Yet most numerical notation systems have a base10 structure (and those that do not use multiples of 10). This does not preclude the existence of binary and hexadecimal numerical notation for specialized computing
purposes. Similarly, while there are only five basic principles of numerical notation
systems found historically (as described earlier), it is easy to imagine other types
that could have existed: a system where the size of a numeralsign is relevant to its
4
For instance: 37. If a numeral expression contains a complex constituent, then the numerical value of the complex constituent itself in isolation receives either simple lexical
expression or is expressed by the same function and in the same phonological shape,
except for possible automatic phonological alternations, stress shifts, or overt expressions
of coordination (Greenberg 1978: 279280).
This is not true of numerical notation systems, some of which (like our own) are truly
infinite.
Or even, as discussed in some aspects of number theory, having a fractional or negative
base!
Numerical Notation
value, or where all nonprime numbers are expressed multiplicatively using prime
number numeralsigns. Several modern writers, abandoning traditional principles
of numerical notation, have created new systems ex nihilo that rely on rather different principles than do the systems discussed in this study (Harris 1905; Pohl
1966; Dwornik 198081). Explaining regularities from a constraintbased perspective allows us to speculate about why certain numerical notation systems flourish
while others do not. Instead of denying the existence of exceptions, I use general
rules to explain why special cases are special, and why some imaginable systems are
unattested in the ethnographic and historical records.
Yet one might wish to contend that comparison of any sort, much less the
universal type of this study, is misleading because each culture, and hence each
numerical notation system, is a product of unique historical circumstances. If so,
comparing Egyptian hieroglyphic numerals to Shang oraclebone numerals and
Inka khipus might be misleading. At best, even if there is a core of features common to all numerical notation systems, I would be labeling oranges apples in order
to compare them to other apples. At worst, if these systems are entirely different
phenomena, I am trying to make apples out of abaci. Yet the relative ease of intercultural communication refutes the claim that all cultures are incommensurable.
The intercultural transmission of ideas relating to numerical notation systems is
frequent and poses a serious challenge to this degree of relativism. Prior to comparing phenomena among multiple societies, one cannot assume either that the phenomenon is crossculturally regular or that it is not. Having compared numerical
notation systems on a worldwide basis, I regard the systems as being sufficiently
similar to warrant their theoretical analysis as variations on a single theme.
I regard numerical notation as translatable crossculturally without significant
loss of information or change of meaning. The number 1138 is practically identical
in referent to MCXXXVIII or t\s\rrr\qqq\qqq\qq or any other representation. These
systems have very different structures, but, in Saussurean terms, the various signifiers refer to the same signified (Saussure 1959). Although the linguistic and symbolic signifiers for numbers may differ greatly (23, dreiundzwanzig, XXIII, viginti
tres, etc.), the correlation of both numeralphrases and lexical numerals with natural numbers is not culturally relative. Yet, while seemingly uncontroversial in the
exact sciences, the crosscultural universality of number concepts has been criticized recently by relativistic anthropologists and sociologists. In his recent work
on Quechua number and arithmetic, Urton (1997) asserts that Western concepts
such as odd/even are not appropriate to the Quechua arithmetical experience,
and that the Quechua use a fundamentally different ontology of numbers than the
Western one. Yet Quechua numbers can be understood in the same way as any others, and the Inka numerical notation used by Quechua speakers (Chapter 10) can be
compared to others without any particular difficulty. Relativist philosophers such
Introduction
as Restivo (1992) claim that 1 + 1 cannot equal 2 in any absolute manner, because
if one were to take a cup of popcorn and add a cup of milk to it, the result would
not be two of anything, but somewhat more than a cup of pulpy mush. Resisting
the temptation to describe such casuistry as pulpy mush, I simply point out that
addition is an arithmetical function that can only represent adding discrete objects
of a like nature. Such evidence does not convince me that the number concepts
of nonWestern societies are incommensurable with our own. On the contrary,
my own research suggests that these differences are relatively inconsequential in
comparison to the commonalities observed in all societies.
I acknowledge that, by treating all numerical notation systems purely as systems for representing number, I do not do justice to the complex symbolism that
complements many of them or to the scholarship on numerology (Hopper 1938,
Crump 1990). The arrival of the year 2000 was not simply another cause for celebration (or trepidation); rather, the nature of our numerical notation system and
the rolling over of the calendrical odometer on 2000/01/01 held great symbolic
and even mystical significance for much of the worlds population. My decision
to underemphasize numerology is based partly on space limitations, but also on
my theoretical interest in the comparable core of features underlying all lexical
numeral systems and numerical notation systems. These interesting differences do
not affect the validity of crosscultural comparisons, but merely highlight the need
to establish, rather than assume, the level of regularity in sociocultural phenomena. It may be true, as Geertz (1984: 276) famously asserted, that [i]f we wanted
home truths, we should have stayed at home, but if we want human truths, we
must compare.
10
Numerical Notation
Any natural number can be expressed as the sum of multiples of powers of some
base. In Western numerals, 4637 is 4 1000 + 6 100 + 3 10 + 7 1 or, to use
exponential notation, 4 103 + 6 102 + 3 101 + 7 100. Because the Western
numerals use the principle of placevalue, the value of any numeralsign in the
phrase is determined by its position position dictates the power of the base that
is to be multiplied by the sign in question. If the order changes, the value changes,
so that 6437, 3674, and so on mean different things than 4637. We could also write
the number out lexically as four thousand six hundred and thirty seven. Instead of
using placevalue, the powers (except for 1) are expressed explicitly thousand, hundred, ty. Because each multiplier corresponds to a word for a power, we could in
theory move each power and its multiplier to a different spot without introducing
ambiguity; German lexical numerals, among others, do exactly that viertausend
sechshundert sieben und dreizig four thousand six hundred seven and thirty.
Some numerical systems, however, do not use multiplication at all. To use the
Roman numerals, one simply adds up the values of all the signs: MMMMDCXXXVII 1000 + 1000 + 1000 + 1000 + 500 + 100 + 10 + 10 + 10 + 5 + 1 + 1.
Although there is no logical requirement that systems like the Roman numerals
list their powers in order, they almost universally do so. The Roman numeral
CCLXXVIII could be unambiguously read even if it were written as VIIICCXXL,
or even as XLIVIXCIC, if we omit the slight complexity of the occasional use of
subtraction. The fact that such disordered phrases are not valid tells us something
about systems that lack placevalue, however they too are structured as the sum
of multiples of powers. We can thus interpret the Roman numeral MMMMDCXXXVII as (1000 + 1000 + 1000 + 1000) + (500 + 100) + (10 + 10 + 10) + (5 + 1 + 1).
I will return in a moment to the issue of the signs for 500 and 5, and how they
affect our understanding of such systems.
The only major systematic attempt to date to classify numerical notation systems is Genevive Guitels Histoire compare des numrations crites (1975).7 Guitel
classifies approximately twentyfive systems (drawn from about a dozen societies)
according to whether they use addition alone to form numeralphrases (Type I,
like Roman numerals), addition and explicit multiplication (Type II, like English
lexical numerals), or implicit multiplication with placevalue (Type III, like Western numerals) just as I have done here. Each type is further subdivided according
to the systems base(s) and other features. Despite an admirable attempt, Guitels
analysis fails the most basic test of classification, which is that it must classify
similar systems together and separate dissimilar ones. It is problematic because its
primary division is made only on the basis of the degree to which multiplication
7
See also Zhang and Norman (1995). Ifrahs (1985, 1998) popular studies on the subject
follow Guitels typology.
Introduction
11
12
Numerical Notation
The product of the two signs determines the value of that power. The multiplicative principle is often used only to structure higher powers of the base.
Interexponential organization determines how the values of the signs for each
power of the base are combined to symbolize the value of each entire numeralphrase. It is subdivided into additive and positional subtypes. Additive systems
are those in which the sum of the intraexponential values in a numeralphrase
produces its total value. For instance, the Roman numeral CCLXXVIII consists of
two 100s (102), one 50 (5 101), two 10s (101), one 5 (5 100), and three 1s (100), for
a total of 278. Positional or placevalue systems, of which the Western system is the
best known, are those in which the value of a numeralphrase is determined not
only by its constituent numeralsigns but also by the place of each sign within the
phrase. The intraexponential values within a numeralphrase must all be multiplied by the appropriate powervalues before the sum of the phrase can be taken.
All numerical notation systems are structured both intra and interexponentially,
creating six theoretically possible pairings of principles. However, it is logically
impossible for a multiplicativepositional system to exist because multiplicative
systems represent the required positional value (10, 100, 1000, etc.) intraexponentially, leaving only five possibilities, as detailed in Table 1.1.
Cumulativeadditive systems, such as Roman numerals, have one sign for each
power of the base; the signs within each power are repeated and their values added,
and then the total value of the phrase is the sum of the signs. Cumulativepositional
systems likewise use repeated signs to indicate the value of each power, but this value
is then multiplied by the placevalues (in the Babylonian example used earlier, 60
and 1) before summing the phrase. In order to be entirely unambiguous, some sort
of placeholder or zero sign is required. Cipheredadditive systems have a unique sign
for each multiple of each power of the base (19, 1090, 100900, etc., in a base10
system like the Greek alphabetic system); the values of these signs are added to obtain
the value of the numeralphrase. Cipheredpositional systems like Western numerals
have unique unit signs from one up to but not including the base (e.g., 1, 2, 3 ... 9)
and a zero sign; the unitvalue is multiplied by the powervalue indicated by its position, and then the sum of these products gives the total value. Finally, multiplicativeadditive systems (like the traditional Chinese system shown earlier, but also for that
matter spoken English lexical numerals, e.g., three thousand six hundred and twenty
four) juxtapose a unitsign (or signs) and a powersign, which are multiplied together,
and then the sum of those products gives the total value of the phrase.
Most numerical notation systems use only one of these five combinations
throughout the entire system. However, some additive systems use one intraexponential principle (either cumulative or ciphered) for lower powers of the base,
and then use the multiplicative principle thereafter. These systems, which I call
hybrids, comprise about 30 percent of those I examine in this study. Systems that
Introduction
13
Positional
The value of each power
must be multiplied by a
value dependent on its
position before taking the
sum of the numeralphrase.
Cumulative
Many signs per
power of the base,
which are added
to obtain the total
value of that power.
Classical Roman
Babylonian cuneiform
(10 + 10 + 1 + 1 + 1) 60 +
(10 + 10 + 10 + 10 + 10 + 1 +
1 + 1 + 1) 1
Ciphered
Only one sign per
power of the base,
which alone represents the total value
of that power.
Greek alphabetic
Khmer
1434 = /auld
1434 =
(1 1000 + 4 100 + 3
10 + 4 1)
Multiplicative
Two components
per power, unitsign(s) and a powersign, multiplied
together, give that
powers total value.
Chinese (traditional)
LOGICALLY
EXCLUDED
1434 =
MCCCCXXXIIII
1434 =
1434 = b3
e4
(1 1000 + 4 100 + 3 10 + 4)
use two principles are not exceptions to my typology. They simply need to be
analyzed in two parts, with each part of the hybrid being assigned the appropriate principle. For instance, the version of the Greek alphabetic system shown in
Table 1.1 is cipheredadditive for powers below 1000 and multiplicativeadditive
for those above 1000.9 No numerical notation system employs more than two of
the five basic types, and no positional system uses more than one type.
Systems that have a subbase as well as a base require further typological clarification because they may use two intraexponential principles: one for units up
to the subbase, and another for multiples of the subbase up to the base. For instance, the cumulativepositional Babylonian system shown in Table 1.1 has a base
9
This feature is the one that leads Guitel to place this version of the Greek alphabetic
numerals in her Type II as opposed to Type I.
14
Numerical Notation
of 60 and a subbase of 10. In this case, we must know both how units from 1 to 9
are expressed and how tens from 10 to 50 are expressed in order to fully describe its
intraexponential structure. In this case, both the subbase and the base use the cumulative principle, so we might more properly describe this system as a (cumulativecumulative)positional system. However, no system uses a different intraexponential
principle for its subbase than for its base, so this elaboration is mostly unnecessary.
Again, none of this affects the interexponential structure of these systems.
Because it reflects both intra and interexponential principles, this typology
shifts the focus of analysis from systems to the structural principles that build
systems, and thus allows a more nuanced comparison of systems structures.
Moreover, it allows us to ask fruitful questions regarding the cognitive effects and
historical development of numerical notation.
Introduction
15
These authors go into far more detail on the various research programs undertaken to
study animal and human infant perception of numerosity than is warranted here.
16
Numerical Notation
was shown that the mathematical abilities of Clever Hans and other animal calculators were the result of subconscious cues passed from human trainers to these
purported prodigies (Fernald 1984). Yet, following in the footsteps of Koehlers
(1951) work on counting among birds, and the enormous literature studying primate numeracy (Matsuzawa 1985; Boysen and Berntson 1989, 1996), we now know
that many animal species are able to perceive quantity at least accurately enough
to perform tasks involving small quantities, mostly up to three to five units. It is
not yet known whether animal quantification is a homology inherited from an
ancestral species, a specific convergent adaptation in many species to the requirements of similar physical environments, or a general cognitive response of animals
of a certain level of neurological complexity. Many experiments involving many
different species have confirmed that something more than a Clever Hans phenomenon is being observed. The same is true in the case of human infants, who are
able to distinguish small numerical quantities (Gelman and Gallistel 1978, Wynn
1992). While our hominid ancestors did not need numerical notation, the ability
to distinguish between two gazelles and three gazelles would have been cognitively important and evolutionarily adaptive. As the survival of early hominids was
strongly predicated upon the ability to function in groups, the number concept
likely developed relatively early in human prehistory, although direct evidence is
limited. It is highly probable that by the time of the Upper Paleolithic (40,000
to 10,000 years ago), Homo sapiens sapiens possessed languages including two or
more numeral words and the ability to conceptually distinguish cardinal and
ordinal quantities (Marshack 1972; Wiese 2003, 2007).
Any human being (save those suffering from certain types of brain damage or
other serious mental deficiencies) has the capacity to learn how to use numerical notation. As a technology invented in particular historical contexts, however, its use is
limited to those who have encountered it. Anatomically and cognitively modern humans survived for millennia without any need for numerical notation, and the variability among numerical notation systems cannot be explained fully by the universal
human mathematical ability. Even so, this does not prevent us from considering the
possible effects of human cognitive capacities on the types of numerical notation
system that have been developed historically. It is very likely that the evolved capacity of some primates to distinguish five from six bananas is related to the human
visual capacity to distinguish five from six strokes on a tally or knots on a cord. Three
biological characteristics of humans pertain to the development of the concept of
number, which in turn is necessary for the development of numerical notation.
1. Perception of discrete external objects. The ability, common at least to all animals, to
distinguish foreground from background, to perceive the borders of external objects, is
necessary to the creation of the concept of oneness.
Introduction
17
2. Perception and cognition of concrete quantity. The ability to distinguish the quantity
of sets of objects is present in human infants and some animals, but is generally restricted to small quantities.
3. Possession of language. The ability to identify numbers using linguistic symbols, as
opposed to the prelinguistic quantitative abilities possessed by infants and animals,
permits the conceptualization of number through a series of lexical numerals, each
greater than the previous by one unit.
While numerical notation systems are useful because they enable the human
brain to conceptualize quantities efficiently, we must not assume that their structure and evolution can be derived entirely from the principles of cognitive psychology. Some neuropsychologists examine the development of numerical notation from a cognitive perspective (Dehaene 1997, Butterworth 1999). Dehaene
(1997: 115117) uses a stagebased unilinear scheme to describe the development of
numerical notation from its beginnings in onetoone correspondence, through
chunked groupings of notches and ciphered numerals, to the ultimate stage of
positional notation with a zero. However, I am very suspicious of such schemes
in the absence of significant historical documentation. The contention that the
history of technology can be understood as a sequence of eversuperior inventions
the better to fit the human mind and improve the usability of numbers is untested at best, ethnocentric at worst (Dehaene 1997: 117).
There are three sociocultural features that are likely prerequisites for the development of numerical notation. These are nonuniversal and derive from contingent historical circumstances, so it is possible to establish whether they are necessary conditions for the development of numerical notation using my universal
crosscultural methodology.
1. Presence of organizing principles that structure the number line. This refers to the
ability to structure the natural numbers in a manner most convenient to thought,
usually taking the form of a numerical base. No known numerical notation system has
ever been developed by speakers of any of the worlds many languages whose lexical
numerals have no base.
2. Presence of a nonstructured tallymarking system based on onetoone correspondence. Often claimed as the earliest stage of numerical notation through which
all societies must pass, tallying is a very intuitive way to represent number visually.
There is evidence for this form of representation as early as the Upper Paleolithic
(Marshack 1972).
3. Social need for longterm recording and communication of number. The social need
for a relatively permanent record of numbers is essential to the development of numerical notation. One of its main functions is to assist memory, so the social need to preserve
18
Numerical Notation
numbers beyond the ordinary limits of memory for whatever specific purpose is
probably necessary to its development. Related to this is the need to communicate
number outside a local community. While verbal numbers suffice for local communication, the ability of numerical notation to communicate numbers across barriers
of geography and language is an important feature that would make its development
likely in such circumstances.
Introduction
19
dichotomy. All associations of numerical structure with cognitive ability are untested,
and rely on the equally untested assumption that numerical notation develops from
concreteness to abstraction over time. By examining the diachronic patterns that
actually occurred in the evolution of numerical notation, I will show that these
patterns are multilinear, not unilinear. By comparing the structure of systems to
the functions for which they were used, I will examine the cognitive framework
within which different groups used numerical notation, keeping in mind that it
is only one part of a cluster of techniques that includes mental calculation, lexical
numerals, finger numbering, and computational artifacts.
Rather than assigning labels such as concrete and abstract to numerical notation systems, or identifying any other single factor on which the utility of a system should be judged, I focus on a constellation of features of numerical notation
systems that have cognitive consequences. This approach is similar to that adopted
by Nickerson (1988), who lists the relevant criteria as being ease of interpretation,
ease of writing, ease of learning, extensibility, compactness of notation, and ease
of computation. A set of nonhierarchical criteria for evaluating systems from a
cognitive perspective is a very valuable tool. Nickerson notes usefully:
If one accepts the idea that the Arabic system is in general the best way of representing numbers that has yet been developed, one need not believe that it is clearly
superior with respect to all the design goals that one might establish for an ideal
system. It may be, in fact, that simultaneous realization of all such goals is not possible. (Nickerson 1988: 198)
20
Numerical Notation
Introduction
21
22
Numerical Notation
language use. Numerical notation, by contrast, is largely nonphonetic and translinguistic, and may diffuse more readily than scripts. The Western numerals diffused
initially from India and passed through the Arab world before reaching Europe,
while the Roman alphabet is of Greek and Phoenician ancestry. This historical differentiation is not uncommon; the path of diffusion of numerical notation is often
radically different from that of the diffusion of scripts. Yet there may be a connection between the indigenous development of writing and numerical notation. In
several historically unrelated cases (Mesopotamia, Egypt, China, and Mesoamerica),
the independent invention of numerical notation immediately preceded or coincided with the development of a fullfledged script. Perhaps the social need for
numerical notation and a phonetic script tends to arise under similar circumstances
(i.e., during the formative phases of early civilizations). Alternatively, the idea of
numerical notation, once developed, might naturally suggest to its users that other
domains might also be represented visually. I return to this subject in Chapter 12.
Finally, the structures by which written lexical numerals and numerical notation
express number are quite different. The simple fact of being denoted visually is not
as important as the different principles used in the two symbol systems. Lexical
numerals (whether written or verbal) share a common structure that is very different
from that of numerical notation.11 For instance, while the cumulative principle is
commonly employed in numerical notation, it is nearly absent from lexical numeration. No known language expresses thirty as ten ten ten, even though cumulative
numeralphrases (like the Roman XXX = 30) are quite common. In lexical numeral systems that have a base, multiplicativeadditive structuring is overwhelmingly
prevalent, whereas numerical notation systems are only occasionally multiplicativeadditive. To take a familiar example, let us compare Western numerical notation
with English lexical numerals. The English lexical numerals eleven and twelve do not
follow the regular pattern for numbers between thirteen and nineteen, and words
like dozen and score add further complexity. Our numerical notation system is base10 and cipheredpositional, while English lexical numerals use a mixed base of 10
and 1000 (one million = 1000 1000; one billion = 1000 1000 1000), a situation
that becomes even more complex if we include British English, in which one billion
normally means one million millions (1012). Finally, while Western numerals are infinitely extendable one can add zeroes to the right of a number ad infinitum English
lexical numerals are only potentially infinite, since one needs to develop new words
to express higher and higher values. The highest number in many English dictionaries is decillion (1033 in American English, 1060 in British English).
11
The major exception to this disjunction is the classical Chinese numerals, which due to
the somewhat logographic nature of the Chinese script serve both as lexical numerals
and as numerical notation. I will return to this definitional issue in Chapter 8.
Introduction
23
The relationship between the origins of writing systems and numerical notation
systems is similarly complex. Visual number marks clearly precede phonetic writing by many millennia. A wealth of evidence from Upper Paleolithic portable artifacts (e.g., notched bones and stones) suggests that onetoone marking of numbers for calendrical or other mnemonic purposes has roots extending back at least
30,000 years (Absolon 1957, Marshack 1972, dErrico 1998, dErrico et al. 2003).
This may in turn have been related to the early use of the fingers and hands as a
visual, though nonpermanent, numerical system around the same time (Rouillon
2006).12 SchmandtBesserat (1992), on the basis of controversial interpretations
of Mesopotamian evidence from the protoliterate period, has been the strongest advocate for an evolutionary sequence from numeration to writing. Houston
(2004: 237) argues that most writing systems emerged as word signs bundled with
systems of numeration that probably had a different and farmoreancient origin,
and this may be correct. However, numerical notation (as opposed to nonbasestructured tallies) does not greatly precede, if at all, the earliest writing. As I shall
show, in all the independent cases of the development of numerical notation, written numerical systems with bases emerge alongside other conventionalized signs,
not as a unilinear predecessor to them.
For these reasons, to analyze numerical notation systems as adjunct components of scripts does not do them justice. Nevertheless, throughout this study I
will sometimes refer to numeralsigns and numeralphrases as being written.
When I do so, it is mere conventionality, and this usage does not indicate any
specific relationship between numerals and scripts.
It is absolutely clear that manual counting and numerical notation are connected in later
societies known through written and oral evidence (see Chapter 12).
24
Numerical Notation
Introduction
25
26
Numerical Notation
Few inventors of numerical notation systems have ever provided detailed information about the contexts of their inventions. Thus, I must build a circumstantial
case for the origins of most systems. In order to demonstrate cultural affiliations
between numerical notation systems, I use both internal (structural and graphic)
resemblances between systems and external (contextual and historical) considerations. The main criteria I use are as follows:
1. Use of the two systems at the same point in time. This criterion is nearly
unavoidable; some chronological overlap in the periods during which two systems
are used is needed to sustain a hypothesis of cultural transmission. An extinct system might conceivably be revived and modified by a later society (for instance, on
the basis of old inscriptions), but this is hardly a sufficient basis for a hypothesis of
cultural transmission. Alternately, a system that is not attested to have survived may
in fact have done so; this is the basis of the controversial theory that the Mycenean
Linear B numerals (Chapter 2) gave rise to the Etruscan numerals (Chapter 4).
Such hypotheses cannot be dismissed immediately, if other factors suggest that
they could be true, but they require much more evidence.
2. Similarity in structural features. Because there are only three intraexponential principles (cumulative, ciphered, multiplicative), two interexponential principles (additive, positional), three common bases (10, 20, 60), and two subbases
(5, 10), no one aspect that is similar in two systems is sufficient to prove a connection. However, when two systems are alike in all or most of these respects, cultural
contact becomes a much more likely explanation for the resemblance. Many of
the cultural phylogenies of systems that I discuss share a common structure; for
instance, all the Italic systems (Chapter 4) are cumulativeadditive with a base of
10 and a subbase of 5. This does not mean that all identically structured systems
must be placed in that phylogeny the Ryukyu shochuma numerals (Chapter
10) and modern Berber numerals (Chapter 10) do not fit because they were used
much later and have different numeralsigns. The use of structural features as evidence of contact suffers from the weakness that, if many systems in a phylogeny
are identical or similar, it is often impossible to choose between several equally
likely candidate ancestors.
3. Similarity of forms and values of numeralsigns. Because many graphic
symbols are very complex, they are unlikely to have developed independently.
If the forms of numeralsigns used in two systems are identical or very similar,
and if those signs represent the same numerical values in the two systems, it is
likely that cultural contact resulted in the invention of the later system based on
the earlier one. The more signs that are shared between two systems, the more
likely it is that there is a historical connection between them. However, when two
systems use similar signs for different numerical values, this is not good evidence
of such a connection. For instance, and J represent 10 and 20 in the Kharoh
Introduction
27
numerals (Chapter 3) but mean 7 and 9 in the Brhm numerals (Chapter 6).
In this instance, even though the two systems were used in the same region at
the same time (the Indian subcontinent in the fourth century bc) and have two
similar numeralsigns, the dissimilarity of their values reduces the likelihood of
a historical connection. Caution must be exercised when invoking this criterion
for very simple symbols vertical and horizontal lines, dots, circles, crosses, and
the like because such designs are crossculturally common. This is especially
true in the case of the use of lines and dots with the value of one, since these
signs may have been part of tallying systems before being used in numerical
notation systems. Cases where signs are similar but not identical must also be
treated with caution. There is no general paleographic principle for identifying
relations among graphically similar signs; hence, such efforts usually proceed on
an intuitive basis.
4. Known cultural contact between the regions where the two systems are
used. In general, where one cultural trait is transmitted from one region to
another, multiple traits are likely to have been transmitted. Thus, where there
is a known pattern of shared nonnumerical features in two societies, or where
there is substantial evidence of interregional trade, migration, or colonization,
such evidence supports a postulated ancestordescendant relationship between
two numerical notation systems. Determining whether known cultural contact is sufficient to postulate the diffusion of a numerical notation system is
always a tricky matter and involves an evaluation of various lines of evidence.
For instance, one of the difficulties in postulating that the Brhm numerals
(Chapter 6) are descended from the Egyptian demotic ones is that, despite
structural and graphic resemblances between the two systems, Egypt is well
down on the list of areas with which ancient India had contact. In no case do
I postulate a connection between two systems solely on the basis that they
were used at approximately the same time and in a single region. There must
always be some structural or graphic resemblance between postulated ancestor
and descendant systems.
This problem is made more complex by stimulus diffusion, a complex blend of
inventive and diffusionary processes in which awareness of an invention is transmitted, but, because of some obstacle to transmission or acceptance, the actual
invention does not take hold in the adopting society (Kroeber 1948: 368370).
However, because the general principle is seen as useful, some members of the
adopting society, stimulated by the original idea, invent their own version of the
innovation. The most widely cited example of stimulus diffusion is the development of the Cherokee syllabic writing system by Sequoyah in the nineteenth century, based on his rudimentary knowledge of the Western alphabet. While several
numerical notation systems resulted from stimulus diffusion (e.g., the abortive
28
Numerical Notation
Cherokee numerals, never used in the syllabary), no principles exist to help identify stimulus diffusion. It is tempting to postulate stimulus diffusion even when
the basic fact of incomplete transmission cannot be established. However, I use
stimulus diffusion as an explanation only when it can be established that the form
of cultural contact that occurred between two regions fits Kroebers model.
5. Use of ancestor and descendant systems in similar contexts. If two systems
serve similar purposes, on similar media, or among similar social groups in their
respective societies, this can serve as further confirmatory evidence that the two
systems are related historically. This factor, while useful, is never sufficient on its
own to demonstrate such a connection, but it may provide further support. For
instance, the spread of the Greek alphabetic numerals into Armenia and Georgia
(Chapter 5), though poorly documented, is confirmed not only by the striking
similarities in the systems but also by the systems use in Bibles and other liturgical texts. The similarities among some of the cuneiform systems of Mesopotamia
(Chapter 7) rest on their common use of a wedgeshaped stylus on clay media as
much as on specific resemblances in the numeralsigns or their organization.
6. Geographic proximity of the regions where two systems were used. All
other factors being equal, a system is more likely to have been modeled on one
that is used by neighboring groups than on one used more distantly. This is a
particularly dangerous criterion to invoke, especially where there is less cultural
contact between neighboring regions than with more distant regions. Many times,
two very different and unrelated systems are used in proximity to one another, and
other times, closely related systems are used at considerable distances from one
another. Geographical proximity is such a weak measure that I will use it only as a
last resort, and never as the sole factor for hypothesizing transmission.
Establishing links between ancestor and descendant systems, within the limits
of the available data, allows me to describe phylogenies of related systems. These
are, however, analytical descriptions, which allow the explanation of evolutionary
patterns of change in numerical notation systems. These explanations are analogical, because they describe independent recurrences of cause and effect. However,
they are also explaining homological processes resulting from cultural contact and
the transmission of knowledge among many societies. This is a paradox only if we
accept the notion that these two concepts stand in opposition to one another. A
phylogenetic perspective is both homological and analogical, seeking to describe
particular historical contexts, but also to derive general processes by which numerical notation systems are related to one another. Diffusion may be, as Harris
contends, a nonprinciple, but it is not a nonprocess. Comparing ancestor and descendant systems, and understanding the nature of the process of borrowing and
adoption of cultural features, is absolutely essential to an evolutionary perspective
on cultural change.
Introduction
29
Such perspectives accept without proof that the Western numerals are the most
efficient ever developed, and are not only the best in existence but also perfect the
best that could ever be conceived. Their adoption by the vast majority of human societies today is perceived as a natural and inevitable consequence of this
superiority, only minimally mediated by social factors. Other, more cumbersome
systems are to be evaluated in relation to the Western system, and in particular to
their utility for arithmetical calculations and higher mathematics. Since so many
modern technologies require mathematics, Western numerical notation is a partial cause of these evolutionary developments. The corollary of this proposition,
often left unstated, is that those societies that did not develop or adopt Western
numerals failed to compete politically with the West in part because of this.
30
Numerical Notation
I consider the decimal, cipheredpositional system of numerical notation developed in India in the sixth century ad and transmitted by Arab scholars to Western
Europe to be a very remarkable invention. Its brevity, unambiguity, and ease of
learning make it conducive to the practice of written arithmetic and mathematics. How well numerical notation systems represent number strongly affects the
development of new systems, their acceptance after being transmitted, their modification over time, and their eventual abandonment. This pattern of longterm
sociocultural change can meaningfully be called evolutionary.
The primary difficulty with the assumption of the evolutionary progress of numerical notation is not the notion of evolution itself. The problem is that the efficiency of numerical notation systems cannot be evaluated in the abstract, but only
by considering the purposes for which they were developed and used. It is often
assumed that the function of numerical notation is to perform written computations. For instance, Ifrah, whose work is the most popular and influential study of
the history of numerical notation, writes:
To see why placevalue systems are superior to all others, we can begin by considering
the Greek alphabetic numeration. It has very short notations for the commonly used
numbers: no more than four signs are needed for any number below 10,000. But that
is not the main criterion for judging a written numeration. What matters most is the
ease with which it lends itself to arithmetical operations. (Ifrah 1985: 431)
Introduction
31
one corner (probably to record the number of fivedollar bills received at some
event). None of these numeralphrases was actually ever used to compute.13 Numbers denote far more often than they reckon, even in our highly numerical society.
This was doubly true in preindustrial contexts.
In defining a numerical notation system as a system for representing numbers, I
am explicitly making a functional statement. At minimum, whatever else numerical notation may mean in a particular society, it must always express number as
one of its functions. I am not saying that a numerical notation system must be
fully integrated with other sociocultural phenomena, that it must be perfectly
adapted to serve social needs, or that the purpose for which it is used must be that
for which it was developed. The representational function of numerical notation
is general enough that it can be stimulated by a variety of social or political needs.
While trade is the most obvious one making transactions possible over long distances, enabling monetary calculations, or recording results to facilitate accurate
bookkeeping it is not the only one. For instance, the main impetus behind the
origin of the Mesoamerican numerical notation systems was probably astronomical and calendrical, while the Shang numerals were first used in the context of
Chinese divination. We should not expect to find a single specific domain of activity correlated with the origin of numerical notation, and we should be skeptical of
universal or unilinear schemes.
If we wish to compare the efficiency of various numerical notation systems, we
must compare systems that served a common purpose in terms of how well they
served that purpose. Because all systems represent number visually, some general
criteria can be used. A system that represents numbers using few numbersigns is
more efficient than one that requires many signs. One could then argue that the
Roman numerals are not as efficient for representing number as Western numerals
are because 1492 is much shorter than MCCCCLXXXXII (or MCDXCII). While
the situation is slightly more complex MMI is shorter than 2001, for instance the
Roman numerals are more concise for only a small fraction of all natural numbers.
Two other criteria that are relatively easily definable are a systems signcount (how
many signs it uses in total) and extendability (the highest number expressible).14
I return to these criteria in Chapter 11 and show how they can be used to ask
13
14
One might protest that the numeral on the bill is used in doing arithmetical computations such as providing change for purchases. To refute this, one need only go into a bank
and ask for $100 in fivedollar bills, and see whether the teller looks at the number on
each bill, or whether in fact he or she merely counts out twenty bills while doing mental
arithmetic. The numeral on the bill denotes its value, but is not used in calculation.
See Nickerson (1988: 189197) for a different, but related, list of criteria used in comparing numerical notation systems.
32
Numerical Notation
fruitful questions that help explain synchronic and diachronic patterns among
attested systems.
By contrast, efficiency for computation is a Westerncentered and historically inaccurate benchmark for comparing numerical notation systems. Zhang
and Normans (1995) paper on the visual representation of numbers through numerical notation is a major step forward in our understanding of how numerical
notation systems work. They analyze how specific systems visually represent (or
fail to represent) numerical information, describe three general means by which
numerical notation systems are structured (shape, quantity, and position), and
then examine how these features are combined in numerical notation systems. Yet
their analysis falls apart because they compare and evaluate different numerical
notation systems based on their ability to aid in multiplication. Even if Western
numerals are the best system for doing arithmetic (which would best be resolved
through the use of the systems rather than abstract theorization), most other systems were never designed or used for such a purpose. The situation is analogous
to denigrating screwdrivers for being inefficient hammers. The fact that one can
use a screwdriver handle to drive in nails does not justify that comparison, just as
the fact that one might use Egyptian hieroglyphic numerals to multiply does not
justify comparing them to systems such as the Western numerals. To add insult
to injury, even though Zhang and Norman recognize that calculation technologies such as the abacus are frequently better than numerical notation for doing
arithmetic, they suggest that part of Western numerals superiority is that they are
used for both calculation and representation, while other societies employed two
separate systems (Zhang and Norman 1995: 293). They thus blame the carpenter
for using both a hammer and a screwdriver where just the screwdriver would do.
Such arguments are little more than elaborate rationalizations for a historical fact
(the nearuniversality of Western numerals) that eludes simple explanation.
The only way to compare numerical notation systems fairly is to use functional
criteria that apply to all systems (namely, those related to simple representation).
Yet, even where there are definite answers to these efficiencyrelated questions,
this does not mean that individuals testing out a new system will immediately
perceive its advantages and disadvantages. A familiar but in some respects inefficient system, so long as it is not entirely unworkable, may be retained, despite
its obvious inferiority. There may be a steep learning curve preventing the easy
adoption of the alternative system, or there may be cultural or political reasons
for retaining ones present system. Moreover, numerical notation, as a system for
communicating information to others, requires not only that specific individuals
adopt it, as would be the case with a more efficient plough or a better mousetrap,
but also that an entire social group learn it before its usefulness will be evident.
A system with many users is functional for that reason alone, because it can be
Introduction
33
used to communicate with more people than one with few users. Thus, it is inappropriate to evaluate numerical notation systems only in terms of their structural
features. Rather, these features must be considered in the broader social context in
which systems develop and are used. I examine these social factors and show how,
far from negating structural factors, structural and social explanations of regularities combine to produce a more complete understanding of numerical notation
than has previously been possible.
I turn in the following chapters to the body of data itself. I endeavor to highlight the ways in which the general theoretical principles just discussed relate to
the data. I have organized these data according to cultural phylogenies of related
systems, presenting the earliest systems first, leading forward to systems developed more recently. The first five phylogenies are probably related to one another
historically, so I treat them together, but no other principle has been used in the
ordering of chapters. The eight major phylogenies, each of which merits a full
chapter, are as follows:
Chapter 2: Hieroglyphic systems historically descended from the Egyptian hieroglyphic
numerals;
Chapter 3: Levantine systems used in the Levant, descended from the Aramaic and
Phoenician numerals;
Chapter 4: Italic systems used in the circumMediterranean region, descended from
the Etruscan numerals;
Chapter 5: Alphabetic systems whose signs are mainly phonetic scriptsigns, descended
from the Greek alphabetic numerals;
Chapter 6: South Asian systems originating on the Indian subcontinent and descended
from the Brhm numerals;
Chapter 7: Mesopotamian systems used in Mesopotamia, descended from the protocuneiform numerals;
Chapter 8: East Asian systems descended from the Shang numerals;
Chapter 9: Mesoamerican systems descended from the Mesoamerican baranddot
numerals.
chapter 2
Hieroglyphic Systems
The hieratic and demotic systems are too complex to be included on this chart; consult
their individual entries for their numeralsigns.
34
Hieroglyphic Systems
35
10
100
1000
Egyptian hieroglyphic
q
\=
s
0
u\
Cretan hieroglyphic
Minoan Linear A
Mycenean Linear B
Cypriote syllabic
Hittite hieroglyphic
Egyptian Hieroglyphic
The hieroglyphic script is the bestknown ancient Egyptian script. It was used
between about 3250 bc and 400 ad, making it the longest surviving of all scripts
(Loprieno 1995). However, its use was restricted geographically to the Nile Valley
and nearby areas under Egyptian control. While the hieroglyphic script may well
have arisen because of stimulus diffusion and trade with Mesopotamia, the scripts
in these two areas emerged essentially simultaneously and show no substantial
resemblances. Hieroglyphic inscriptions are written from top to bottom, left to
right, or right to left, with the last of these three options being the most common (Ritner 1996: 80). The script is mixed in principle, with both phonograms
(consisting of one, two, or three consonants) and logograms indicating words
nonphonetically (Ritner 1996: 74). The later hieratic and demotic scripts used to
write the ancient Egyptian language, as well as the Meroitic hieroglyphic script,
are directly derived from the hieroglyphic, while the early scripts of the Levant and
the Aegean are probably its less direct descendants.
Numbers other than one are very rarely expressed through lexical numerals in
Egyptian, making it difficult to determine their structure, although evidence from
some Old Kingdom Pyramid Texts and later Coptic writings establishes that they
had a purely decimal structure with words for each power of 10 up to 1,000,000
(Loprieno 1995: 71). Most hieroglyphic inscriptions express numbers using numeralsigns rather than words, however, with separate signs corresponding to each power.
These signs are shown in Table 2.2 (cf. Gardiner 1927: 191; Allen 2000: 97).
Numerical Notation
36
10
100
1000
10,000
100,000
1,000,000
RL
q
q
r
r
v\
Lex.
fn
LR
68,257 =
The system is purely decimal and cumulativeadditive, with each sign repeated
up to nine times as necessary, and ordered from highest to lowest rank. The direction in which a numeral is read is always the same as the direction of writing, but
varies depending on the inscription in question. The set of signs in the top row of
Table 2.2 are those used when the direction of writing is from left to right; when
righttoleft writing is used, the signs are mirrored (i.e., q\r\\\\). Occasionally,
when days of the month are being expressed, the signs for 1 and 10 were placed on
their side: or ^ instead of r or q (Gardiner 1927: 191). Numeralsigns could
be used either cardinally or ordinally, with ordinals from second through ninth
adding the ending nw (masculine) or nwt (feminine) to the numeralphrase, and
those from tenth upward adding m (masculine) or m t (feminine). To aid in
reading long numeralphrases, five or more identical signs were usually grouped in
sets of three or four rather than placed on a single line. Thus, 5 is written as a row
of three signs above a row of two signs, 6 as a row of three above a row of three, 7
as a row of four above a row of three, 8 as a row of four above a row of four, and 9
either as a row of five above a row of four or as three rows of three.2
The sign for 1 is a simple vertical stroke. Gunn (1916: 280) suggests that in early
wellexecuted inscriptions, the sides of the vertical bar are curved inward slightly,
thus making a biconcave bar, and postulates that it may represent a small object
of bone or wood used in some kind of tally or aid to reckoning, but I tend to
think that it is simply an abstract stroke. The sign for 10 has been described as a
heel bone (Kavett and Kavett 1975: 390), a tie made by bending a leaf (McLeish
1991: 42), or even, anachronistically, as a croquet wicket (Boyer 1959: 127). In fact,
it corresponds to the phonetic value m (masc. m w, fem. m t) hook, handle,
2
However, other groupings were sometimes used when it was more convenient for the
scribe to do so.
Hieroglyphic Systems
37
and is a rebus for the Egyptian lexical numeral for 10 (m ) (Sethe 1916: 2). The
higher power signs also have specific representational qualities and can also represent phonetic values in Egyptian apart from their use as numerical symbols. The
sign for 100 ( t) is probably a coiled length of rope; that for 1000 ( ) is a lotus
plant; the sign for 10,000 ( b ) is an extended finger; and that for 100,000 ( fn)
is a tadpole. These numeralsigns, as well as the overall structure of the system,
remained remarkably stable throughout its history. In some older instances in
which the sign for 1000 occurs, rather than grouping the signs in clusters of three
to five separated signs (as in the numeralphrase mentioned earlier), multiple lotus plants were depicted as emerging from a single bush (e.g., 3000 = ). The sign
for one million ( ) could also mean multitude or a countless quantity, just as
the Greek word myriad can mean a group of ten thousand or, more generally, a
large quantity (Loprieno 1995). After the Early Dynastic period, this nonspecific
lexical sense predominated over the specific numerical value. In most other respects, Predynastic hieroglyphic numerals would have been completely intelligible
to Late Period scribes.
The earliest known Egyptian hieroglyphic numerals are those from Tomb
Uj at Abydos, which dates to around 32503200 bc (late Naqada II or early
Naqada III period), and also contains the earliest examples of Egyptian writing (Dreyer 1998). Numeralsigns occur on many drilled bone and ivory tags
found in this royal tomb, which were probably once attached to containers of
goods. Other tags have other signs that resemble later Egyptian hieroglyphs,
but only a very few contain both numerals and hieroglyphs (Baines 2004:
154157). Some tags have six to twelve vertical or horizontal strokes, others the
sign for 100, and one has both a sign for 100 and a sign for 1 (Dreyer 1998:
113118). This system has three distinctive features as compared to the mature
hieroglyphic system: it uses both horizontal and vertical strokes for units;
there is no attested numeralsign for 10; and there are tags with more than
nine unitstrokes. Dreyer (1998: 140) explains the first two of these differences
simultaneously by noting that on some Old Kingdom linenlists, horizontal
strokes stand for 10. The Tomb Uj tags are very similar to others found at
Naqada and Abydos that date from the Naqada III and Early Dynastic periods, which contain the sign for 10 and use only vertical strokes for 1 (Dreyer
1998: 139). The very early date of the tags suggests that the system was developed independently of Mesopotamian influence, although the Uj tags are
essentially contemporaneous with the Uruk IV tablets. The margin of error
and discrepancies in the different radiocarbon dates from Tomb Uj are large
enough that no definite conclusion regarding priority can be reached (Baines
2004: 154). Even though the Uj tags are apparently administrative or commercial, the context of the discovery (a royal tomb) suggests instead that the
38
Numerical Notation
Figure 2.1. A scene from The Narmer macehead, a late Predynastic ceremonial artifact
bearing early but recognizable Egyptian hieroglyphic numerals. At the far right, a quantity
of 120,000 prisoners is indicated, while the lower register indicates quantities of 1,422,000
goats and 400,000 cattle. Source: Quibell 1900: Plate XXVI B.
signs were part of a nascent visual high culture of interest to elites seeking to
legitimate their authority (Baines 2004: 170171).
While we have no evidence for numeralsigns higher than 100 from the Tomb
Uj tags, by the Early Dynastic period the system was fully developed. Figure 2.1
depicts the Narmer macehead found at Hierakonpolis, which may describe the
unification of Upper and Lower Egypt by around 3100 bc, and which demonstrates
that even the very highest signs were being used at that time (Arnett 1982: 42).
The macehead indicates an exaggerated tally of 400,000 oxen, 1,422,000 goats,
and 120,000 humans (Quibell 1900: 89, Pl. XXVI). This is traditionally interpreted as an exaggerated and propagandistic tally of booty and prisoners acquired
through Narmers military victories. Millet (1990), however, provides an alternate
interpretation of the macehead inscription as a yearidentifier and suggests that
the numbers are purely artificial, meant only to signify the taking of a census.
Another early example of hieroglyphic numerals is found on the Second Dynasty
statue of Khasekhem indicating the slaughter of 47,209 of the pharaohs enemies
(Guitel 1958: 692). These large numerical values figure prominently in early hieroglyphic inscriptions, further supporting the idea that early Egyptian monumental
writing was primarily oriented toward display purposes relating to the ideological
justification of the kings authority (cf. Baines 2004). Other numerical inscriptions from the Early Dynastic include tags for commodities from Naqada like the
Hieroglyphic Systems
39
earlier ones found at Abydos, but using ordinarily structured hieroglyphic numerals
(Imhausen 2007: 14).
In the Old Kingdom (25752134 bc), variants on the basic hieroglyphs were not
uncommon. Clagett (1989: 5657) discusses a variant of the hieroglyphic numerals
used on the Palermo Stone (a Fifth Dynasty / 2400 bc pharaonic annal), in which
certain notations of the aroura measure of land are represented with unitstrokes
in a quasipositional manner. However, this system is not used regularly throughout the Palermo Stone and is not found in any other inscriptions; hence its value
for understanding the hieroglyphic numerals is somewhat limited. Another unusual Old Kingdom system has been proposed by PosenerKriger (1977) to have
been used in papyrus documents from Gebelein indicating area measures of fabric on socalled linenlists (mentioned earlier). In this system, a single cubit sign
meant one square cubit, horizontal strokes ten, and long vertical strokes meant
one hundred each of which were followed by short vertical strokes indicating
how many of the requisite units were denoted. Both of these variant systems were
employed solely for enumerating particular types of goods, and were never used
more generally for denoting numbers.
In the Ptolemaic era (33230 bc), the hieroglyphic numerals became more
complex. The sign for 1,000,000 was reintroduced into the numerical sequence,
though it is unclear whether its numerical meaning was truly understood. In a few
inscriptions from this period, a ring sign is found in the sequence between
v and w. While Sethe (1916) believed that the ring sign was a meaningless addition, Gunn (1916: 280) protested that perhaps, in order to lengthen the series of
numerals without assigning the god w a subordinate place, was assigned the
value of 1,000,000, while w either shifted upward in value to ten million or else
retained its lexical meaning of an uncountable number. Curiously, on the stela
of Ptolemy Philadelphos (r. 282246 bc) at Pithom, the sign used for 100,000 is
not v but rather , with the ring sign placed underneath the ordinary tadpole sign
(Sethe 1916: 9).
Another curious change in the late hieroglyphic numerals is the occasional
use of cryptographic ciphered numeralsigns for many numbers, as shown in
Table 2.3. These signs replaced the standard cumulative sets of signs with single
signs whose association with the number was homophonic, pictorial, religious, or
related to the corresponding hieratic numeralsign. They were used as early as 950 bc
on a wooden votive cubit rod of Sheshonk I, but are found on no artifacts between that point and the Ptolemaic era (Priskin 2003). The most common of these
signs is that for 5, a fivepointed star, which often combines with unitstrokes in
the same way as V = 5 in Roman numerals (Sethe 1916: 25). However, unlike the
Roman numerals and related systems, no signs were developed for 50, 500, or
other halfpowers. The origin of this sign is almost certainly pictorial, from the five
Numerical Notation
40
Q
U
Z
A
B
points of the star. Other common signs were a human head for 7, from the Egyptian understanding of the head as having seven orifices, and a scythe for 9, from
the resemblance between that sign and the hieratic numeralsign for 9 (Sethe 1916: 25).
In addition to the signs for the units 1 through 9, there were cryptographic hieroglyphs for 60 and 80, both of which were derived from resemblances to hieratic
numerals (Fairman 1963). These new signs never led to a fully cipheredadditive
set of hieroglyphic numeralsigns, and were often included in otherwise perfectly
ordinary cumulative numeralphrases.
Hieroglyphic numerals are largely written on monumental inscriptions, but not
exclusively so. Texts including hieroglyphic numerals include seals, funerary stelae
and tomb inscriptions, annals, lists relating to conquest and plundered goods, and
certain administrative texts. An oftenoverlooked source of hieroglyphic numerals
is the wide variety of stone balanceweights bearing inscriptions indicative of their
weight (Petrie 1926, Petruso 1981). Numerals indicated dates, weights and measures, and a wide variety of quantities of goods, animals, and people. In all of these
texts, the numerals are formed in the ordinary fashion just described.
The hieroglyphic numerals were rarely if ever used for mathematics and calculation. The vast majority of Egyptian literary texts, and all Egyptian mathematical
texts, are written in the hieratic or later demotic scripts (cf. Gillings 1978: 704705).
Nor are there hieroglyphic numerals marked on potsherds, tallies, or other such
media that would suggest their use as an intermediate step in performing calculations. Some hieroglyphic numerals are used in an inscription from the tomb of
Methen (Fourth Dynasty, twentysixth century bc), which indicates the calculation
of the area of a rectangle, but this inscription indicates only that the calculation was
done; the numerals were not actually used in the calculation process (Peet 1923: 9).
Yet, because Egyptologists regularly transliterate documents in the hieratic
script into regularized hieroglyphs, historians of mathematics have sometimes
Hieroglyphic Systems
41
inferred wrongly that the Egyptians calculated using hieroglyphic numerals. The
hieroglyphic system is cumulativeadditive, while the hieratic system is cipheredadditive, but since this difference in structure was underemphasized by Egyptologists of earlier generations (e.g., Gardiner 1927: 191; Peet 1931: 411), historians of mathematics frequently presume that the hieroglyphic numerals were the
only ones available to Egyptian scribes (cf. McLeish 1991: 42; Guedj 1996: 3435;
Palter 1996: 228229; Dehaene 1997: 97). It is to be hoped that the presence of
new Egyptological literature may remedy this deficiency (Ritter 2002, Imhausen
2007). It is necessary to treat the hieroglyphic and hieratic systems separately, not
despite their very strong historical connection, but because of that connection,
inasmuch as the two systems were different in structure and used in entirely different functional contexts.
A very few hieroglyphic inscriptions express large numbers (particularly multiples of 100,000) through multiplicative formations instead of purely additive
ones. In one Ptolemaicera text, the number 27,000,000 is expressed by placing
a single sign for 100,000 above the ordinary additive hieroglyphic phrase for 270
(Brugsch 1968 [1883]: III, 604).3 The only other way to write 27,000,000 would
have been to use 270 signs for 100,000 or 27 signs for 1,000,000, neither of which
is an attractive option. Such phrases would be unappealing from the perspective
of Egyptian aesthetic canons, in addition to the clear economy of symbols enjoyed
through multiplicative phrases. In a second instance (from the time of Amenhotep
III, around 1400 bc), 100,000 is expressed multiplicatively using the tadpolesign
v placed above a vertical stroke 100,000 1 = 100,000 (Sethe 1916: 9; Loprieno,
personal communication), thus, curiously, requiring more signs than the standard
numeralphrase. Finally, multiplicative hieroglyphs are found on a number of votive cubits from the New Kingdom and later, cubitlong polygonal stone objects
inscribed with metrological and religious information, with clear multiplicative
phrases for multiples of 100,000 and possibly also for 100 and 1000 (Ritter 2002:
308309). Yet there is no evidence that this multiplicativeadditive structure was
widespread. The cipheredadditive hieratic numerals use multiplicative forms far
more frequently, and earlier, than do the cumulativeadditive hieroglyphs (Sethe
1916: 810; Mller 1936: I, 59). Yet because Egyptian grammars mention the hieratic examples (e.g. Gardiner 1927: 191; Allen 2000: 97) but transcribe the numerals
as hieroglyphic numerals, it is easy to conclude that multiplicative expressions are
common in the hieroglyphic numerals, when in fact almost all such expressions
come from hieratic texts.
3
While Brugsch interprets this figure as 100,270, the figure being represented is the
amount (in arouras) of land in Egypt, for which 27,000,000 is the only reasonable interpretation (cf. Kraus 2004: 225).
42
Numerical Notation
The question of when this borrowing took place remains open; the first hieratic
documents to use this structure date to the Middle Kingdom (2040 to 1652 bc),
while the first hieroglyphic example (mentioned earlier) dates to about 1400 bc.
Because the hieratic multiplicative numeralphrases are more common and earlier
than the hieroglyphic, I think it likely that the ancestral system (hieroglyphic)
borrowed the feature from its descendant (hieratic). Because hieroglyphic numerals were used only for monumental purposes at that time, numbers higher than
100,000 would have been expressed only infrequently. It is entirely possible, given
the small number of hieroglyphic inscriptions using multiplication, that it was an
exceptional response to the occasional requirement for expressing high numbers
in hieroglyphic numerals. There is no evidence supporting Guitels assertion that
this occasional use of multiplication, which is paralleled in certain Aztec texts
(Chapter 9), represents even an abortive step toward a fully positional notation
(Guitel 1958: 692695; 1975: 44). Rather, it represents an alternative means of
increasing the conciseness of some (but not all) numeralphrases and extending a
systems capacity to write numbers while retaining its basic structure.
The Egyptian hieroglyphic script possessed two distinct systems for representing
fractional values, both of which normally expressed only unitfractions those in
which the numerator is 1. The first such system, the standard system for expressing
fractional quantities, simply required the scribe to place the sign r, which could
also mean part or mouth, above any hieroglyphic numeralphrase to indicate
the corresponding unitfraction (Loprieno 1986: 1307). If the mouth sign was too
small to place over the entire phrase, it was simply placed over the signs for the
highest power of 10. This system also used special symbols for some of the most
commonly used fractions: 1/2 (gs), 1/4 (r4), 2/3 (rwj), and 3/4 (hmtrw) (Sethe
1916: Table II; Allen 2000: 101). The last two of these are not unitfractions, and
are thus exceptions to the general rule. This system was not used in the Predynastic
era, but is found in abundance during the Old Kingdom and thereafter. While,
in theory, this system could express any fraction, most have denominators smaller
than 20.
The second system was used primarily for measurements of volume of grain,
fruit, and liquids by indicating fractions of the heqat ( t), a measure probably
equal to 292.24 cubic inches, or roughly 4.8 liters (Chace et al. 1929: 31). This notation is sometimes known as Horuseye fractions because the six hieroglyphic
symbols for fractional values can be combined to form the glyph of the w t or
eye of Horus (#), a symbol of health, fertility, and abundance. The sum of these
signs is only 63/64; symbolically, the remaining 1/64 would be supplied magically by the god Thoth when he healed the Eye of Horus, thus producing unity
(Gardiner 1927: 197). Ritter (2002) makes a persuasive case, however, that these
signs were originally nonpictographic, hieratic submultiples of metrological units
Hieroglyphic Systems
43
1/4
2/3
3/4
Horuseye fractions
1/2
1/4
1/8
1/16
1/32
1/64
& *
rather than pure numerals, and that they were not originally associated with the
Eye of Horus at all, and thus he renames them capacity system submultiples.
Despite the early caution of Peet (1923), the Horuseye interpretation has misleadingly become standard among historians of mathematics. The hieroglyphic
forms of these signs are shown in Table 2.4, presuming a leftright direction of
writing (Sethe 1916: Table II; Gardiner 1927: 197).
This systems binary structure was probably most useful for dividing and multiplying by two, a standard operation needed when manipulating volumes of goods.
The system probably originated in an earlier hieratic series of fractional signs, of
which the earliest example is from the Abusir Papyri of the Fifth Dynasty, and
only later did the signs assimilate to resemble the parts of the Horuseye symbol
(Reineke 1992: 204). Other than one possible sign for 1/2 from the Fifth Dynasty,
the earliest hieroglyphic Horuseye fractions are from the Nineteenth Dynasty or
later (Priskin 2002: 76; Ritter 2002: 304).
The Egyptian hieroglyphic numerical notation system has several direct descendants, the most direct of which is the cipheredadditive Egyptian hieratic
system (to be discussed later), which developed as early as the First Dynasty as
a scribal shorthand for the hieroglyphs (Peet 1923: 11). Egyptian scribes would
have learned both the hieroglyphic and hieratic numerals during their education,
and used both systems in the appropriate contexts the hieroglyphs on stone
monuments, and the hieratic numerals written in ink on papyrus and ostraca
(inscribed potsherds). It is also very likely that the civilizations of the Aegean
used the Egyptian hieroglyphic numerals as the model for their own indigenous
numerals the Cretan hieroglyphic system, the Linear A and B numerals, and the
HittiteLuwian hieroglyphic numerals. There was considerable commercial and
political interaction between Egypt and the Aegean in the second millennium
bc, when the Aegean numerical notation systems began to emerge (Cline 1994).
Despite the dissimilarity in the numeralsigns of the two systems, they are identically structured, and thus a hypothesis of diffusion is likely correct. A less direct
descendant of the Egyptian hieroglyphs is the PhoenicianAramaic system, which
44
Numerical Notation
was developed around 750 bc, blending the numeralsigns and script tradition of
the Egyptian hieroglyphs with the structure of the AssyroBabylonian cuneiform
numerals (Chapter 7). This development marks the formation of the Levantine
systems (Chapter 3), reflecting the intermediary position of the Levantine civilizations between the larger polities of the eastern Mediterranean.
By the GrecoRoman period, the use of the hieroglyphic script and numerals
had declined greatly, and both writing and numerals had increased in the number
of signs used and the complexity thereof, to the point where it was considered to
be a purely symbolic or cryptographic script by outsiders (Ritner 1996: 81). By the
third century ad, Egypt was becoming increasingly Christian in its religion, and
its language was being written in the Greek and Coptic scripts. The latest dated
hieroglyphic inscription is on the temple of Isis at Philae, and dates from August
24, ad 394 (Griffith 1937: I, 126127). By the fifth century, knowledge of how to
read and write hieroglyphs had disappeared. The hieroglyphic numerals, as well
as their immediate descendants, were replaced by the Coptic alphabetic numerals
(Chapter 5).
Egyptian Hieratic
The hieratic script was developed around 2600 bc by Egyptian scribes as a sort
of cursive shorthand for the earlier hieroglyphic script (Loprieno 1995) and, like
its forerunner, used a mixture of logographic and phonographic components.
However, unlike the hieroglyphs, hieratic writing was designed for cursive writing
on papyrus and on ostraca, making it suitable for administrative and literary
purposes. Furthermore, while the hieroglyphs could be written in a variety of
directions, hieratic texts are always linear and written from right to left. While the
form of the hieroglyphs was very regular and formalized, hieratic writing varied
greatly by period, location, and the idiosyncrasies of the scribes handwriting. The
Old Kingdom divergence of Egyptian scripts into monumental (hieroglyphic)
and cursive (hieratic and demotic) variants continued throughout the remainder
of ancient Egyptian history.
A base10 cipheredadditive numerical notation system accompanied the hieratic script. The hieratic numeralsigns, like the script itself, changed considerably over the systems extensive history. The paleographic development of hieratic
numerals is traced in the charts provided by Mller (1936). In Tables 2.5, 2.6, and
2.7, I present three distinct sets of numerals, the first and earliest from the Kahun
papyrus (Twelfth Dynasty / 20001800 bc), the second from Pap. Louvre 3226
(fifteenth century bc), and the third from the Harris papyri (twelfth century bc)
(Mller 1936, vol 1: 5963, vol. 2: 5559). These three texts contain mostly complete sets of numeralsigns at least as high as 1000, and are thus very useful for
Hieroglyphic Systems
45
a
b
10s
j k
100s
s t
1000s
1 2
10,000s
: ::
100,000s >
4367 = gou4
1s
c
l
u
3
:::
d
m
v
4
=
e
n
w
5
f g h
o p q
x y z
6
8
i
r
0
9
A B C D E F G H I
10s
J K L M N O P Q R
100s
S T U V W X Y Z [
1000s
\ ]
657 = GNX
1s
a b
10s
j k
100s
s t
1000s
 }
10,000s
8 9
100,000s ,
56,207 = gt4;
1s
c
l
u
1
0
d
m
v
2
:
e
n
w
3
;
f
o
x
4
<
g
p
y
5
=
h
q
z
6
>
i
r
{
7
?
46
Numerical Notation
comparative purposes. The Harris papyri numerals, from Table 2.7, include all of
the numbers up to 100,000; this is the only text to do so.
Looking only at the signs for 5, 6, 7, and 9, the three series appear remarkably
distinct. At the same time, however, most of the hieratic numeralsigns show remarkable continuity. Many of the hieratic signs used in the Old Kingdom would
have been perfectly comprehensible to a scribe in the Late Period or even the Ptolemaic era. Many of the numeralsigns are very similar to others from the same period; for instance, it is very difficult to distinguish 400 from 600 or 3000 from 5000 in
Table 2.7. When used to express days of the month, hieratic numerals, like hieroglyphic numerals, were often rotated ninety degrees counterclockwise to reflect
their function. Given the nature of the Egyptian calendar, these forms exist only for
numerals less than 30. To write fractional values, a small dot was placed above the
numeralphrase for an integer to indicate the appropriate unit fraction (1/x).
The hieratic system is primarily cipheredadditive, and its signs each represent a
multiple of a power of 10. Many of the hieratic numeralsigns bear a clear relationship to their cumulativeadditive hieroglyphic forerunners, seen particularly in
the signs for 1 through 4, 10, 10,000 through 40,000, and 100,000. Other hieratic
numerals show no clear correspondence with their hieroglyphic ancestors except
in very early periods. The cipheredadditive hieratic system thus shows traces of its
cumulativeadditive ancestry. For this reason, I include the hieratic system in this
chapter even though it is structurally different from its hieroglyphic ancestor.
In some hieratic texts, irregular numerical systems were used in conjunction
with grain measures (Allen 2000: 102). One early Middle Kingdom system notated
sacks of grain using regular numerals, and heqats (1/10 sack) using one to nine
dots in a cumulative fashion. Later in the Middle Kingdom, a notation developed
whereby ordinary numerals placed before a heqat unit indicated multiples of 100,
those after the numeral multiples of 10, and then one to nine dots for the units.
These systems were not used outside of this metrological context. As discussed
earlier, hieratic fractions were frequently written in unitfraction form or through
capacity system submultiples (Ritter 2002).
For writing many numbers above 10,000, multiplicative notation was used in
the hieratic numerals; for instance, the sign for 60,000 is written by placing the
sign for 6 below the sign for 10,000. This principle is not used for 10,000 through
30,000, but was used occasionally for 40,000, and normally for 50,000 through
90,000 and for values above 100,000. While the multiplicative principle is seemingly used for certain values of the hundreds and the thousands, paleographic
analysis of the numeralsigns shows that the sign for 300 represents the abbreviation of the first two of three cumulative 100signs and the extension of the third
rather than the juxtaposition of 3 and 100. Imhausen (2006: 2526) discusses an
Hieroglyphic Systems
47
300
Hieroglyphic
Old Kingdom
Kahun papyrus
(Dyn. 12)
P. Louvre
3226 (Dyn. 18)
P. Harris
(Dyn. 20)
qqq
qqq
qqq
qqq
qqq
ostracon from the New Kingdom workers village of Deir el Medina, a scribal
exercise in which 600,000, 700,000, 800,000, and five, six, and seven million are
expressed multiplicatively using the hieratic signs for 100,000 and one million,
respectively.
The regular use of multiplicativeadditive structuring allowed very high numbers
above 100,000 to be expressed easily in hieratic numerals by placing the appropriate
multiplier below the tadpole sign. The earliest multiplicative hieratic numerals are
from the twentieth century bc Kahun papyrus fragments, which are arithmetical
problems and accounting documents, suggesting that this technique originated in
the context of mathematical or arithmetical practice (Griffith 1898: 16). The development of hieratic numerals was thus a highly creative process involving both the shift
to ciphered notation and the use of multiplication where it was deemed useful.
The strong similarities between the hieratic numerals and the earlier hieroglyphic numerals, coupled with the indisputable historical connections between
the two scripts, prove the historical relationship of the two systems. Whereas the
hieroglyphic numerals are found in Predynastic inscriptions, hieratic numerals
first appear in the First Dynasty (Peet 1923: 11). Their use became widespread from
the Old Kingdom onward, with the two systems (hieroglyphic and hieratic) being used for parallel purposes. The earliest hieratic numerals were little more than
cumulativeadditive cursive forms of the appropriate hieroglyphic numerals. Over
time, the numerals became increasingly removed from their hieroglyphic ancestors as multiple strokes were condensed into single strokes, probably for greater
ease of writing. By the Fifth Dynasty, the numerals written in the Abusir papyri
(archives of royal funerary cults) had acquired a strongly cursive character that had
moved away from the original cumulative signs (Goedicke 1988: xvixvii). Table 2.8
48
Numerical Notation
compares how the numbers 6, 9, and 300 were written in Old Kingdom hieratic
to the numeralsigns from the three sets of numerals presented earlier.
While ciphered signs were the ordinary ones, the systems origins were not completely forgotten; cumulative hieratic numeralsigns were occasionally employed
even into the New Kingdom. No single individual invented ciphered notation
in Egypt; rather, its development was a process of abbreviating and combining
cumulative signs by scribes over many centuries until, by the Late period, very few
hieratic signs bore any resemblance to their hieroglyphic counterparts. It is even
possible that the scribes making these changes were not really aware of the importance of the new structural principle they were using. Hence the origin of ciphered
notation may, in some sense, have been accidental. Strikingly, in some hieratic
documents from the Ptolemaic era, there is a reversion in the numeralsigns away
from the ciphered signs used in older hieratic texts and back to the common use of
the cumulative principle. In several texts (Leinwand, P. Bremner, IsisN., Leiden
J. 32, and P. Rhind),4 hieratic units were expressed with repeated vertical strokes,
tens with horseshoeshaped curves, and hundreds with coils, in an exact imitation of the hieroglyphic numeralphrases of the same value (Mller 1936: vol. III,
5960). While some of these documents retained the ciphered signs for some values, there is a trend over time toward the use of cumulative numeralsigns in these
late hieratic documents. Some scribes may have forgotten the ciphered signs; more
likely, however, the reversion to cumulativeadditive numerals was a deliberate
archaism, resulting from the desire to emulate hieroglyphs more exactly.
Egyptian scribes would have learned both hieroglyphic and hieratic writing
and numerals during their education, and used whichever was appropriate according to the context. Accordingly, while the functions of the hieratic numerals
are quite distinct from those of the hieroglyphic numerals, the users of the two
systems would have been the same individuals. For the hieratic numerals, two
functions stand out above all others: administration and mathematics. Almost all
extant Egyptian legal, commercial, educational, and literary texts from 2600 to
600 bc are written in hieratic, and numerals abound on such documents. While
hieroglyphic numeralphrases were very lengthy, requiring an enormous number
of symbols to express many small values, hieratic numerals were highly concise,
facilitating their use in accounting, commerce, and law, as well as for expressing
dates and cardinal quantities. Because they would have been learned and used by
only a small and welleducated segment of the populace (i.e., the scribes), their
main disadvantage the large number of signs one needed to learn in order to use
the system would not have been a serious problem.
4
P. Rhind does not refer here to the famous Rhind Mathematical Papyrus, but to a different text dating to 9 bc and having nothing to do with mathematics.
Hieroglyphic Systems
49
A limited but interesting set of hieratic texts directly concern mathematical subjects. The hieratic numerals were first used in Egypt for arithmetic and
mathematics in the late Middle Kingdom and the early Second Intermediate
Period (Twelfth and Thirteenth Dynasties). The wellknown Reisner, Berlin,
Kahun, and Moscow mathematical papyri all date from the nineteenth century bc
(Gillings 1978: 704705). Later, around 1650 bc, during the period of Hyksos domination, the Egyptian Mathematical Leather Roll and the famed Rhind Mathematical Papyrus were written using hieratic numerals, though the latter may be a copy
of an earlier document. Egyptian mathematics was never perceived as a separate
field of activity, but was thoroughly enmeshed and embodied within daily scribal
practice, so to search for pure mathematics divorced from administrative activities is futile and ethnocentric (Imhausen 2003: 386). While the most thoroughly
mathematical texts from ancient Egypt date from roughly 1900 to 1650 bc, these
texts are but a minuscule fraction of the total number of hieratic texts containing numerals, and cannot represent the full scope of mathematical practice over
four millennia of Egyptian history. A full discussion of the mathematics of ancient
Egypt is well beyond the scope of this work (cf. Peet 1923; Neugebauer 1957; van der
Waerden 1963; Gillings 1972, 1978; Rossi 2004; Imhausen 2003, 2006, 2007).
Tracing the diffusion of the hieratic numerals is quite difficult. As the system was
primarily used for administrative purposes, it spread wherever Egyptian domination
extended for instance, into Canaan in the Nineteenth and Twentieth Dynasties
(Millard 1995: 189190). The early Israelites used a minor variant of hieratic numerals
(to be described later) starting in the tenth century bc. In addition, the hieratic numerals gave rise to two distinct descendant systems. By the eighth century bc, the
hieratic of Upper Egypt (abnormal hieratic) was no longer mutually legible with
that of Lower Egypt, which is now known as demotic, and which eventually
replaced its ancestor. In addition, the Meroitic cursive script, found on ostraca in the
Sudan starting in the third century bc, contains numeralsigns to which Griffith
(1916: 23) assigns ancestry from the hieratic numerals. While the hieratic numerals
have relatively few direct descendants, through their demotic descendant they are
ancestral to a great number of systems.
In the Twentysixth Dynasty (664 to 525 bc), the demotic script and numerals,
which had only begun to diverge from hieratic a century or so earlier, were accorded royal preference for most purposes. After that point, demotic began to replace
hieratic for more and more functions throughout Egypt. By the early Christian
era, when hieratic was encountered by the Greeks, it was used only in religious
texts by which means it got its name, hieratikos sacred. The name that we
now give to this script and numerical notation system is, ironically, taken from a
purpose for which it was rarely used throughout over two millennia of its history.
By around 200 ad, even these religious functions had ceased.
50
Numerical Notation
Hebrew Hieratic
In the ninth century bc, the Egyptian scribal tradition, including the use of the
hieratic script and numerals, was adopted by the ancient Israelites, who incorporated a great deal of Egyptian learning into their own thought (Millard 1995,
Rollston 2006). Prior to this point, there is no evidence that the Israelites used
any numerical notation whatsoever, although some may have become familiar
with Egyptian notations while in Egypt. From around the ninth5 through the
sixth centuries, Hebrew scribes combined their own (PaleoHebrew) script with a
variant form of hieratic numerals, which I will call Hebrew hieratic for simplicity, although they differ only paleographically from ordinary Egyptian hieratic
numerals (Kletter 1998: 142).
The earliest Hebrew inscriptions containing hieratic numerals are the Samaria
ostraca from the late ninth or early eighth century bc. The most notable and
complete example of Hebrew hieratic numerals is a large (30 22 cm) ostracon,
KBar6, excavated from Tell elQudeirat (Kadeshbarnea) in 1979, depicted in Figure 2.2, with the numerals transcribed in Table 2.9. The ostracon contains a very
complete series of hieratic numerals; only the signs for the units 1 through 9 and
60 were missing, blurred, or unreadable (Lemaire and Vernus 1980; Cohen 1981:
105107). It was probably a scribal exercise or practice text in writing numerals and
measures (DobbsAllsopp et al. 2005: 251260). The number 10,000 is expressed
by combining the Egyptian hieratic sign for 10 with the lexical numeral thousand
in PaleoHebrew script, and there are also Hebrew units of measurement (heqat,
shekel, gerah) on the ostracon, so it cannot have been written by an Egyptian
scribe. The numeralsigns are paleographically very similar and structurally identical to the late hieratic ones, indicating that these numerals were directly borrowed
under conditions of economic domination by and cultural contact with Egypt.
Ostraca found at Arad and Lachish and mostly dating to the late seventh and
early sixth centuries record weights and measures in letters and accounts, and
served an administrative function (Levine 2004: 433).6 While Gandz (1933: 61)
argued that the numerals from the Samaria ostraca had an Aramaic origin, the numeralsigns are in fact hieratic in origin (Lemaire 1977: 281). The Samaria ostraca
signs for 5, 10, and 20 are , , and , respectively, which closely resemble the
late hieratic e, j, and k, but not Aramaic aaa\aa, A, and B.
Sheas claim (1978) that this system is attested on a Late Bronze Age jar handle from
the thirteenth century bc has not been addressed by Semitic epigraphers, but both the
paleographic interpretation and the date are questionable.
Some of the Arad ostraca record quantities in a West Semitic (AramaicPhoenician)
notation rather than in hieratic numerals (see Chapter 3).
Hieroglyphic Systems
51
Figure 2.2. The KBar6 ostracon from Kadeshbarnea, most likely the work of a student practicing writing numerals in Hebrew hieratic numerals. Source: Cohen 1981: 106.
Reprinted with the permission of the American Schools of Oriental Research.
The Hebrew hieratic numerals were also used on Judaean inscribed limestone
weights, a metrological system that reflected the growing commercial importance
of Judah within the context of Near Eastern commerce between the late eighth
and early sixth centuries bc. Kletters (1998) study of 434 of these Judaean weights
represents the most thorough examination of these artifacts to date. Although it
was once argued that these notations were cumulativeadditive and indigenous in
origin (Allrik 1954, Yadin 1961), apart from the paleographic variability one would
expect when transferring a cursive notation onto stone, they are clearly hieratic,
Table 2.9. Hebrew hieratic numerals from KBar6 ostracon from Kadeshbarnea
1
10s
100s
1000s
52
Numerical Notation
cipheredadditive numerals like those on the ostraca (Aharoni 1966). The numerals on these weights range from 1 to 50 to indicate multiples of the shekel (approx.
11.3 grams), but curiously, other than the 1 and 2shekel weights, the numerals are
in a 5:4 ratio to the expected masses of the weights. This puzzle has been the source
of considerable recent debate among metrologists and remains incompletely resolved, but in any case the numerical interpretations of the signs are paleographically secure (Ronen 1996, Kletter 1998).
The ostraca and shekel weights are of a relatively early date in the history of
Hebrew writing, and must be understood in the context of growing administrative needs in the Iron Age states of the Levant. This period marks the first in which
literacy was relatively widespread in Judah (Kletter 1998: 144). Rollston (2006)
rightly sees the presence of hieratic numerals and artifacts such as KBar6 as positive (if not conclusive) evidence for the introduction of formal schooling in late
Iron Age Israel. After Judah lost its independence in 586 bc, the system appears to
have become obsolescent. The Hebrew hieratic numerals are not directly ancestral
to the cumulativeadditive Levantine systems that emerged in the eighth century
bc, among Phoenicians and Aramaeans. These systems were used contemporaneously with the hieratic numerals and, for instance, occur in the same contexts on
ostraca from Arad. There is no evidence of an indigenous Hebrew numerical notation system until about 125 bc, when the use of the familiar alphabetic numerals
(Chapter 5) began.
Meroitic
The kingdom of Mero, which flourished from roughly 300 bc to 350 ad, made
use of two distinct scripts. The first, Meroitic hieroglyphs, were based on Egyptian
hieroglyphs and were used on some stone monuments. The attested Meroitic hieroglyphic inscriptions contain no numerals. The other script, the Meroitic cursive, was
written from right to left on ostraca as well as on stone, and was accompanied by a
set of numerals. Almost all of our information on the Meroitic numerals rests on the
work of F. Ll. Griffith, the original decipherer of the Meroitic scripts. Unfortunately,
because the Meroitic language has no known relatives, we are largely unable to read
Meroitic inscriptions, even though the values for the signs of the cursive script are
more or less fully deciphered. Griffith (1916: 22) first presented the interpretation
of the Meroitic numeralsigns shown in Table 2.10. On structural and paleographic
grounds, the values for the units, 10, and all of the hundreds are unquestionable, and
the remainder of the numeralsigns are fairly certain.
This system is cipheredadditive and decimal, and written from right to left
(like the Meroitic cursive script). The number 2348 as shown in Table 2.10 appears
on the stela of Akinidad, which dates to the late first century bc (Griffith 1916: 22).
Hieroglyphic Systems
53
a
b
10s
j k
100s
s t
1000s A
B
2348 = hmuB
1s
c d e f g
l m n
p
u v w x
C
As with the hieratic numerals, the signs for the units, low hundreds, and possibly
1000 through 3000 are somewhat cumulative. There is only one case (again, from
the Akinidad stela) where a number greater than 10,000 is expressed; interestingly, where hieratic uses a single sign for 10,000 (8), Meroitic appears to use a
multiplicative formation (10 1000). However, this evidence is far too limited to
conclude that the Meroites regularly used multiplicativeadditive structuring to
express higher powers. In addition, cumulative sets of one to nine dots apparently
indicated tenths from 1/10 to 9/10, while a dot in a semicircle (0) represented
1/20 (Griffith 1916, 1925). Griffith (1916: 2223) believed this system to be purely
metrological, representing tenths and twentieths of some larger unit of measure
rather than abstract numbers.
By the time of the development of the Meroitic scripts, the hieratic script and
numerical notation system had largely been replaced by demotic throughout
Egypt. Nevertheless, on paleographic grounds (citing especially the signs for 6,
10, and 20, but also the cumulative unitsigns), Griffith (1916: 23) argued that
the Meroitic numerals resemble the late hieratic numerals (eighth to third centuries bc) more closely than the demotic forms, even though the characters of the
Meroitic cursive script are almost certainly derived from a demotic rather than a
hieratic prototype (Millet 1996: 85). More paleographic analysis is desirable to
settle this question.
The Meroitic numerals were used for administrative purposes such as tax records
and mensuration, as well as in funerary and monumental contexts indicating yeardates and quantities of individuals. Griffith suggests that something akin to the
Egyptian heqat or artaba measures, used to indicate volumes of produce such
as corn or dates, was probably indicated on some ostraca (1916: 23). There is no
evidence that the Meroitic numerals were ever used for arithmetic or mathematics. Even on ostraca upon which multiple numerals have been written, Griffith
was unable, except in one instance, to establish any arithmetical correspondence
54
Numerical Notation
between the numerals that would indicate that a tally or sum had been taken
(1916: 24).
The Meroitic numerals were used until the fourth century ad, but did not
outlast the kingdom of Mero. Millet (1996: 84) suggests that the script may have
continued in use until the introduction of Coptic Christianity in the sixth century, but there is no textual evidence to establish whether the Meroitic numerical
notation system existed during this late period. The Coptic and Ethiopic numerals, both of which are derived from the Greek alphabetic numerals, were used
widely in the region from the sixth century onward.
Egyptian Demotic
The demotic script developed in the late eighth century bc (Twentyfifth Dynasty) and began to replace the hieratic script about a century later. It was a cursive script consisting largely of consonantal characters, derived from the business
hand hieratic used in the Nile Delta (Ritner 1996: 82). During the Late period
and the Ptolemaic era, demotic writing was used very widely for administrative
and literary purposes, and more sporadically throughout the Roman period. A
set of cipheredadditive, base10 numerals accompanied this script throughout
its history. As with the hieratic numerals, there is a great deal of variation in the
demotic numeralsigns; the ones presented in Table 2.11 (after Sethe 1916: Table I)
are typical of those found in papyri of the Late and Ptolemaic periods. Griffith
(1909: 415417) provides an interesting paleographic comparison of the demotic
numeralsigns found on a selection of papyri dating from the Twentysixth
dynasty to the Roman period.
The demotic numerals are a base10, cipheredadditive system, written from right
to left. They are less reliant on the cumulative principle than their hieratic ancestor
(compare hieratic c and demotic C for 3). Some of the signs for the thousands may
be vaguely multiplicative, as there is a general resemblance between the signs for the
hundreds and the corresponding signs for the thousands, but it is more likely that
they are simply further reductions of the nonmultiplicative hieratic signs. Sethe
(1916: Table I) suggests that additive phrases incorporating two lower signs (3000 +
2000, 4000 + 3000) were used for the missing 5000 and 7000 signs. Above 10,000,
the demotic numerals, like the hieratic ones, are multiplicative (though such large
expressions are fairly rare); for instance, Parker has found multiplicative expressions
for 90,000 (7, = 9 10,000) and 100,000 (8, = 10 10,000) in his study of demotic
mathematical papyri (Parker 1972: 86). As in the hieratic numerals, a small dot
placed above a numeralsign indicated the corresponding (1/x) unit fraction.
The demotic numerals are directly derived from the hieratic forms used in the
eighth century bc; as the hieratic numerals were used as late as 200 ad, the two
Hieroglyphic Systems
55
A B C
10s
J K L
100s
S T U
1000s @
# $
6268 = HOT&
1s
D E
M N
V W
%
F G H I
O p Q R
X Y Z !
&
* (
systems were used side by side in Egypt for nearly a millennium. This long coexistence can be explained in part by regional variations, with Upper Egypt retaining
the abnormal hieratic numerals and Lower Egypt using demotic. Unlike the corresponding writing systems, the hieratic and demotic numerals would have been
largely mutually intelligible until the Ptolemaic period at least, which may have
facilitated communication between different parts of Egypt. The demotic script
and numerals were accorded royal preference in the Twentysixth Dynasty, and
thus they were used for most royal functions thereafter, while the hieratic system
was retained primarily for calligraphic religious texts (Ritner 1996: 8182).
Unlike the hieratic script and numerals, which were rarely written on stone
except at the very end of their history, demotic inscriptions are found on stone as
well as ceramics and papyrus. Like their predecessor, demotic numerals were used
for a wide variety of commercial, legal, and other administrative functions, as well
as for indicating dates. A number of demotic mathematical papyri have survived
from the Ptolemaic period, confirming the suitability of the system for arithmetical and mathematical purposes (Parker 1972, Gillings 1978). However, as with the
hieratic numerals, most demotic texts that contain numerals serve no mathematical function. Much of our paleographical knowledge of the demotic numerals
comes from administrative texts, such as dowry records and educational papyri
(Griffith 1909). An extensive set of demotic numerals is found in P. Tsenhor, the
private archive of a sixthcentury woman (Pestman 1994).
The importance of the demotic numerical notation system lies not in any structural feature or unusual function, but rather in its historical role as the immediate
ancestor of several other numerical notation systems. The demotic numerals are
almost certainly ancestral to the Greek alphabetic numerals (Chapter 5). These
numerals, which are structurally identical to the demotic numerals, first appear
in the sixth century bc in Ionia and Caria, at which time Greek trade with Egypt
was beginning in earnest, and when the Ionian trading city of Naukratis in the
56
Numerical Notation
Nile Delta was the major center for trade between Egypt and Greece (Chrisomalis
2003). Furthermore, the alphabetic numerals became common in the late fourth
century bc, at which time Egypt came under Ptolemaic control. Remarkably, the
similarities between the demotic and Greek alphabetic numerals have been substantially ignored over the past century, with most scholars inclined to treat the
latter system as a case of independent invention (but cf. Boyer 1944: 159). Secondly, there are strong similarities between the demotic numerals and the Brhm
numerals (Chapter 6), which began to be used in India around 300 bc. In this
case, the historical connection between the two regions is not as clear, but the
structural similarities between the two systems suggest some connection. While
trade between Egypt and India became common only in the Roman period, there
are strong indications of overseas trade dating from the Ptolemaic period and perhaps even somewhat earlier. Again, few historians of mathematics have proposed
this connection, although it has held some popularity among Indologists for over
a century (Bhler 1896, Salomon 1998).
By the Roman period (30 bcad 364), the demotic numerals were used increasingly rarely, as the general decline of Egyptian cultural institutions continued apace.
However, even though Roman imperialism was the immediate circumstance surrounding the decline of the demotic numerals, they were not replaced with Roman
numerals, but rather with the Coptic numerals, which were themselves descended
from the demotic through the Greek alphabetic numerals. As Christianity began to
take hold in Egypt, and the Coptic script and numerals became more widespread,
demotic suffered a fatal decline. The last text with demotic numerals is a graffito
on the temple of Isis at Philae (the same temple that contains the last evidence for
hieroglyphs), which dates to December 2, ad 452 (Griffith 1937: I, 102103). The last
known demotic text of any sort (dated in Greek) was written nine days later.
Linear A (Minoan)
The Linear A script was the standard script used in the Minoan civilization of
Crete between 1800 and 1450 bc (Bennett 1996: 132). It is perhaps the most famous
of all undeciphered scripts, having foiled decades of effort to interpret it. Only the
numerals and a few other ideograms for commodities can be deciphered. Linear A
is written from left to right and is almost certainly a mixture of syllabograms and
logograms. Its wellattested numeralsigns are shown in Table 2.12 (Sarton 1936b:
378; Ventris and Chadwick 1973: 36).
The Linear A numerical notation system is decimal and cumulativeadditive,
and is written from left to right with the powers in descending order. Where appropriate, signs are grouped in two rows of up to five signs each rather than placing them in an uninterrupted row. The variant dot symbol for 10 is found only in
Hieroglyphic Systems
57
10
100
7659 =
1000
qqqqq
qqqq
58
Numerical Notation
numerals. A similar system of small circles and large circles inscribed on cylindrical
stone weights from the palace at Knossos may have indicated one and ten units of
some metrological value (Evans 1906). While either of these systems could be related
to Linear A, at present the hypothesis of borrowing from Egypt best explains the
structure of Linear A numerals, with the numeralsigns developed indigenously.
Numeralsigns are the only known means of representing numbers in Linear
A; although it remains possible that lexical numerals were written using syllabic
signs, the closely related (and deciphered) Linear B script does not do so, suggesting that this is unlikely. The vast majority of Linear A documents are clay tablets
having an accounting or bookkeeping function, and thus we have many examples
of the use of numerals. Vertical strokes that probably represented numbers have
been found in other contexts for example, on Minoan balance weights; these
marks, however, do not show any clear relation to the Linear A signs found on the
clay tablets and are probably simply unstructured unitmarks or tallies (Petruso
1978). What are likely Linear A numerals occur on a number of pieces of pottery
from Bronze Age Cyprus (Grace 1940). Stieglitz proposes that a numerical graffito
found at Hagia Triada and containing the sequence of numbers (1, 1 1/2, 2 1/4,
3 3/8), in which each number is 1.5 times the previous one, represents a series
of musical notes or tunings for a stringed instrument (Stieglitz 1978). I think it
equally likely that the series served an economic function such as calculating interest. Since we do not have significant literary or monumental texts in Linear A, we
do not know if the numerals were ever used in other contexts.
While the Cretan hieroglyphic numerals were formerly thought to be ancestral to
Linear A, it now appears that Linear A predates the Cretan hieroglyphs, perhaps by
as much as a century. The exact historical relationship between the two numerical
notation systems is unclear, but I believe it most likely that the Cretan hieroglyphic
numerals were a local variant of the Linear A system. The Linear B Mycenean script
used on Crete and the Greek mainland definitely derived from Linear A. Its numerals
(to be discussed later) are nearly identical to those of Linear A. The precise relation
between the peoples using the Linear A and B scripts is still unclear, as is the question
of the cause of the collapse of the Minoan civilization in the fifteenth century bc.
Presumably, during this period, the Greekspeaking Myceneans adapted Linear A for
their own language, resulting in Linear B. The two scripts coexisted in Crete from
about 1550 to 1450 bc, after which time Linear B replaced Linear A completely.
Cretan Hieroglyphic
The Cretan Hieroglyphic or Pictographic script was first identified by Sir Arthur
Evans (1909) based on his work at Knossos. While it was once considered ancestral to the other Aegean scripts, it probably developed about the same time as, or
Hieroglyphic Systems
59
10
100
1000
\\=
8357 = \000\\\====
===
slightly later than, the Linear A script. Its use is generally thought to have lasted
from 1750 to 1600 bc (Bennett 1996: 132). It is found on around 300 attested
sealstones and clay documents (Olivier et al. 1996). While the script is still undeciphered, it is probably of a mixed syllabic and logographic structure, like other
Aegean scripts. Among the few Cretan hieroglyphic signs that can be interpreted
securely are the numerals, which are shown in Table 2.13 (Evans 1909: 258; Sarton
1936b: 378; Ventris and Chadwick 1973: 3031).
The system is cumulativeadditive and decimal, and most often written from
left to right, although righttoleft numeralphrases are also attested. Groups of
multiple repeated signs were sometimes organized using two rows, one above the
other, each with no more than five signs, but this rule was not strictly applied, and
in other cases the organization of signs was more haphazard. Figure 2.3 depicts a
Cretan clay rectangular bar on which numerals are written on three of the four
long sides plus the base (Evans 1909: 177). While the number 483 is written at the
bottom of side (b) according to this principle, for instance, many of the other
numerals are oriented irregularly, grouping signs in clusters of six or more. Evans
(1909: 257) assigns the uncommon sign P the value 1/4 because it is repeated
not more than three times at the end of a few numeralphrases, while Dow and
Chadwick (1971: 12) suggest a quite different fractional system with signs for 1/2,
1/4, and 1/8.
Since the Cretan hieroglyphs are largely undeciphered, it is difficult to speculate on the history of their numerals. As with other Aegean scripts, an Egyptian
origin for the system has been proposed (Sarton 1936: 378), though this cannot be
demonstrated conclusively. There is limited similarity between the numeralsigns
for the Cretan hieroglyphs and any other system, except that the use of the dot for
10 is common to some early Linear A inscriptions. Dow and Chadwick (1971: 14)
suggest that the differences between the Cretan hieroglyphs and Linear A are attributable to the fact that the former were designed for chiseled inscriptions while
the latter were intended to be written with ink. The Cretan hieroglyphic numerals
are probably a local variation of the Linear A numerals or, less plausibly, a direct
borrowing from the Egyptian hieroglyphic numerals. The contexts in which the
60
Figure 2.3. Cretan hieroglyphic inscriptions on a clay bar containing numerals. For instance, at the bottom of face b, the
numeral 483 is represented with four diagonal strokes, eight dots, and three curved strokes. Source: Evans 1909: 177.
Hieroglyphic Systems
61
numerals are found are similar to those for Linear A. The Cretan hieroglyphic
inscriptions include information on commodities such as wheat, oil, and olives
and thus are probably records of transactions, inventories of goods, and similar
administrative documents (Ventris and Chadwick 1973: 31). By around 1600 bc,
Cretan hieroglyphs had been entirely replaced by Linear A.
Linear B (Mycenean)
The Linear B script was used on Crete and the Greek mainland in the middle to
late second millennium bc to write an archaic Greek dialect on clay administrative tablets. It is written from left to right, and consists of a syllabary with a large
repertory of logograms and taxograms (classifiers), including a numerical notation
system. The Linear B numerals are shown in Table 2.14.
The Linear B signs are mostly identical with the Linear A signs, except that the
sign for 10 is always a horizontal stroke (never a dot), and there is a sign for 10,000
that is not found in the earlier system. The 10,000 sign is probably a multiplicative
combination of the signs for 10 and 1000. The structure of the system is cumulativeadditive and decimal, with the highest powers on the left, written in descending order and with five or more identical signs divided into two rows.
Unlike the Linear A numerals, Linear B lacks a separate system for expressing fractions; instead, specific logograms express divisions of metrological units
and then combine with numeralsigns as appropriate (just as one might say 10 cm
instead of 0.1 m). Ventris and Chadwick (1973: 5455) note that some of the
Mycenean logograms for metrological units resemble the Minoan signs for fractions, and may have originally indicated specific ratios of two types of units, which
further shows the indebtedness of Linear B to its Minoan forerunner.
The Linear B system definitely originated through direct contact with the
Minoan civilization and the Linear A numerals. The earliest Linear B inscriptions
date from the sixteenth century bc, so the two scripts coexisted on Crete for about
a century. Their numerical notation systems are so similar that some authors do
not distinguish between the two (Ventris and Chadwick 1973: 53; Struik 1982).
The distinction between the two is not nearly as great as between the two scripts,
which record different languages. Throughout the history of the Linear B numerical notation, there is no observable change in the form of the numeralsigns or in
the structure of the system.
Linear B numerals are found almost solely on clay tablets serving accounting and financial purposes (Olivier 1986: 384386). Numerals are used both for
counting discrete objects (men, chariots, etc.) and for measures of dry and liquid
volume and weight. Almost all Linear B documents relate to administrative and
bookkeeping functions, suggesting a very limited level of literacy and numeracy
Numerical Notation
62
10
100
1000
10,000
68,357 = \\\
\\\\\\\\\\\\\\\\ \\\\
throughout Mycenean society. Even so, the consistency of the numerals throughout several centuries and across a substantial geographic area suggests that some
sort of scribal education system was in place to transmit knowledge of both the
Linear B script and its numerals. We do not know if Linear B numerals were written on papyrus or other materials, though such uses are certainly possible.
We also do not know whether the Myceneans used their numerals for arithmetical purposes. Andersons (1958) theory on the means by which such calculations
could be undertaken suffers from the defect that it involves aligning and manipulating numbers as one would in Western arithmetic, although there is no evidence
that such a procedure was ever undertaken. Dow (1958: 32) and Anderson (1958:
368) both point to a clay tablet found at Pylos (designated Eq03) in which tallying
in groups of five units is used to reach 137. Other tablets from Pylos discussed by
Ventris and Chadwick (1973: 118119) show that the Myceneans could successfully
compute complex ratios in order to determine the contributions of goods required
from towns of different sizes. Rather than proving that the Myceneans used numerical notation for arithmetic, however, these examples indicate that tallying by
units and in groups of five, rather than the purely decimalstructured numerical
notation, was the method used for computation. None of this denies that clay tablets recorded the results of rather complex computations done mentally, through
tallying, or perhaps by some other method.
There is no relationship between the Mycenean numerals and either of the later
Greek numerical notation systems (the acrophonic and alphabetic systems). It is
conceivable, however, that there is some relationship between the Mycenean and
Etruscan numerals (Chapter 4). Both Haarmann (1996) and Keyser (1988) have
raised this claim, which will be discussed in detail when considering the origins of
the Etruscan system. Mycenean settlements have been found in Sicily and southern Italy, providing one possible locus for cultural contact. However, this theory
is controversial, not least because of the time elapsed between the latest known
Linear B documents (twelfth century bc) and the first Etruscan ones (seventh century bc). A more likely descendant of Linear B numerical notation is the Hittite
hieroglyphic system, which was invented around 1400 bc and used by Hittite and
Hieroglyphic Systems
63
Luwian speakers in Anatolia. The Hittite signs for 1 and 10 are identical to the
Linear B ones, and at the time when the Hittite numerals were developed, there
were Mycenean settlements in western Anatolia (such as at Miletus) and on
Cyprus that were engaged in trade throughout the eastern Mediterranean. The
contemporaneity of the two systems makes this scenario plausible, if not proven.
The perplexing and apparently violent end of the Mycenean civilization in the
twelfth century bc, and the repeated razing of major sites such as Mycenae and Pylos, marks the end of the Linear B inscriptions and the start of the Dark Age of
Greek civilization. No writing or numerical notation of any kind is attested from
the Aegean region between 1100 bc and the introduction of the Greek alphabet a
few centuries later.
Hittite Hieroglyphic
The Hittites lived in central Asia Minor from about the end of the third millennium bc. The Hittite and closely related Luwian languages are the first IndoEuropean languages for which we have solid textual evidence. By the middle of the second millennium bc, two distinct scripts were in use in the Hittite Empire. Firstly,
a cuneiform script (borrowed from Mesopotamia) was used to write the Hittite
language. Its numerals are closely related to the AssyroBabylonian cuneiform system, and so will be treated in Chapter 7. Additionally, an indigenous hieroglyphic
script was used to represent the Luwian language on monumental inscriptions,
on a few lead tablets, and probably also on wooden tablets that have not survived
(Melchert 1996: 120). This script was used from about 1500 to 1200 bc, during the
apogee of the classical Hittite Empire, and then is found only sporadically until the
rise of the NeoHittite kingdoms between around 1000 and 700 bc, during which
time it was again common (Hawkins 1986: 368). This script is known as Hieroglyphic Hittite or Hieroglyphic Luwian, and has a mixed syllabic and logographic
structure. Among the purely ideographic signs, the Hittites used a set of written
numerals as shown in Table 2.15 (cf. Laroche 1960: 380400).
The system is purely cumulativeadditive and uses a base of 10. Numeralphrases
were written from left to right, right to left, or top to bottom, depending on
the overall direction of the inscription. As in the Egyptian and Aegean systems,
Hittite numeralsigns were sometimes but not always grouped in clusters of three
to five unitsigns. Laroche (1960: 395) indicates that 9 was variously written using
three rows of three strokes, a row of five above a row of four, or simply with nine
strokes in sequence on a single line.
The Hittite hieroglyphic numerals were most likely based on one of the Aegean
numerical notation systems. Both the Linear A and Linear B scripts were in use
around 1500 bc, when the first Hittite hieroglyphic inscriptions are found, but
Numerical Notation
64
Table 2.15. Hittite numerals
1
10
100
1000
^
(
)
3635 = )))((((((*qqqqq
Linear A was almost extinct by that time. Like the hieroglyphs, the three Aegean scripts
use a combination of syllabograms and logograms. The Linear A, Linear B, and Hittite hieroglyphic numerical notation systems are all decimal and cumulativeadditive,
and use a horizontal stroke for the units and a vertical stroke for the tens. There
was a significant degree of intercultural contact between the Aegean and Asia
Minor during this period. The Myceneans had settlements in western Anatolia
and traded throughout the eastern Mediterranean, and were possibly the Ahhijawa (Achaeans) mentioned in the Hittite archive from Bogazkoy. Because the
Luwian language was spoken primarily in western Asia Minor and only later was
used in the Hittite Empire, the transmission of the numerals from the Aegean to
western and then central Anatolia is plausible (Hawkins 1986: 374). An alternate
hypothesis is that the Hittite system was based directly on the Egyptian hieroglyphic numerals, since the Hittites were in contact with Egypt at that time.
Due to the paucity of extant examples, little can be said about the function
and use of the system. The numerals are found on a variety of stone inscriptions and lead tablets. Most notable among these are the Kululu lead strips (mid
to late eighth century bc), which record village census data using an abundance
of numerical signs (Hawkins 2000: 503505). The Hittite numerical notation is
used far more frequently than lexical numerals, which is also true of the Egyptian
hieroglyphs and Aegean scripts. There is no discernable change in the structure
or signforms of the system throughout its history, even though there is little evidence for its use between 1200 and 1000 bc, following the invasion of Phrygians
and others who ended the classical Hittite kingdom. During these two centuries,
the hieroglyphs were likely used only on perishable materials, such as wooden
tablets (Hawkins 1986: 374).
A few inscriptions on clay jars found at the Urartian site of Altintepe (in eastern Asia Minor) use a syllabary closely related to the Hittite hieroglyphs to write
single words in the Urartian language, starting in the early eighth century bc
(Laroche 1971, Klein 1974). Many of these inscriptions contain numeralsigns
for small numbers using either pitted dots or vertical strokes to represent units
(i.e., 5 = 554 or 11111), but never to express numbers larger than eight, making
this system an unstructured tally system having no base. Klein (1974: 93) accurately states that this usage should thus be viewed as an isolated and shortlived
Hieroglyphic Systems
65
phenomenon, possibly not outlasting the career of a single (foreign?) scribe. The
numerals that accompany the Cypriote syllabary, which was invented around
800 bc, are also potentially derived from the Hittite hieroglyphic numerals. The
proximity of the NeoHittite kingdoms to Cyprus, the extensive trade relations
between the regions, and the identical structure of the two systems all suggest
that such a derivation is likely. However, there are too few Cypriote syllabic inscriptions containing numerals to establish an accurate chronology or even to
secure values for certain numeralsigns. Less plausible descendants of the Hittite
hieroglyphic system are the earliest Levantine systems, Phoenician and Aramaic
(Chapter 3). However, these systems developed around 750 bc, at the very end of
the Hittite systems history, and are structurally distinct from it, since they have a
sign for 20 and are multiplicativeadditive above 100.
The subjugation of the NeoHittite kingdoms under the Assyrian empire ended
the use of Hittite hieroglyphic numerals around 700 bc, and the system was replaced for all functions by the AssyroBabylonian common numerals. Later numerical notation systems developed for related peoples of Asia Minor, such as
the Lycians, were based on a Greek model and display no obvious relation to the
Hittite hieroglyphs.
Cypriote Syllabary
As its name suggests, the Cypriote syllabary was a syllabic script used only on the
island of Cyprus. It was used between about 800 and 200 bc for writing the Greek
language, and thus coexisted with the much more prominent and longlasting
Greek alphabetic script (Bennett 1996: 130). Cypriote is always written from right
to left. None of the synthetic works concerning numerical notation have dealt
with the (admittedly small) evidence for a distinct Cypriote numerical notation
system. However, Masson (1983: 80), whose discussion of the Cypriote syllabary is
the most detailed currently available, presents about a dozen inscriptions in which
the system shown in Table 2.16 was used.
This rudimentary system was decimal and cumulativeadditive and, like the
syllabary itself, was written from right to left. The numbers expressed using the
system are very small; unless certain undeciphered signs are in fact numeralsigns (as
discussed later), the largest number expressed in any Cypriote inscription is 22. This
system parallels the Aegean Linear systems from which the Cypriote numerals
are probably derived. This is strongly suggested by the use of the CyproMinoan
script, which was very probably borrowed from Linear A, on Cyprus as early as
1500 bc. However, eastern Cyprus was under Phoenician domination well into the
period of the use of the syllabary, and the Phoenician numerical notation system
is also written from right to left, and uses vertical strokes for units and horizontal
Numerical Notation
66
10
strokes for tens. Furthermore, Masson (1983: 80) notes the use of two unusual
symbols: , found in but a single inscription but possibly indicating 100 on the
model of the Aegean systems, and , also in only a single document, but possibly
signifying 20. It is notable that the Phoenician system used @\and D at various
times as the sign for 20. Because Cypriote inscriptions do not contain dates, it is
often difficult to place them in chronological context, but it is possible that the
Cypriote system is either ancestral to or descended from the Phoenician system.
A final complexity is that the Hittite hieroglyphic numerals, which were still in
use in the NeoHittite kingdoms in 800 bc, also use a vertical stroke for 1 and
a horizontal one for 10. Trade between Cyprus and Asia Minor was substantial,
and it would have been an extremely short sea voyage between the two regions.
None of this material categorically excludes the possibility that the aberrant signs
found by Masson are nonnumerical and that the Phoenician, Hittite, and Cypriote numerals are unconnected except by their temporal and geographic proximity
on the island of Cyprus. The corpus of inscriptions containing numerical signs is
simply too limited, and the numbers expressed too small, to resolve the issue of
their origin.
Summary
Despite the enormous amount of work being done in the archaeology of the eastern Mediterranean, the genetic relations among the systems of this phylogeny have
not been analyzed adequately in the past. The connections between the Egyptian
hieroglyphic, hieratic, and demotic systems are well established, but more data
are needed to establish the specific links between the Egyptian and Aegean systems. Nevertheless, on the basis of a shared set of features that distinguish it from
other, superficially similar phylogenies such as the Levantine (Chapter 3) and Italic
(Chapter 4), the inclusion of all the hieroglyphic systems in a single group is warranted. First, all the hieroglyphic systems have a base of 10, but they do not use a
subbase of 5 or any additional structuring signs. Second, they mostly have a cumulativeadditive structure, although the hieratic, demotic, and Meroitic systems
are cipheredadditive reductions of the original structure. Third, large numbers of
cumulative signs in a numeralphrase are grouped in sets of three to five. Fourth,
their direction of writing can be quite variable (leftright, rightleft, topbottom,
Hieroglyphic Systems
67
chapter 3
Levantine Systems
69
Pahlavi
Manichaean
Sogdian
Middle
Persian
Old Syriac
Hatran
Kharoh
Nabataean
Palmyrene
Phoenician
Aramaic
a
a
a
a
a
a
a
10
a b d
C D
G
A
A
A
H A
N K
> A
> A
R
g
BA
C
B @ D
J
J
J
J
@
20
F\
A
I
L
100
500
1000
10,000
Numerical Notation
70
Table 3.2. Aramaic numerals
1
10
20
100
1000
A C
F
G
2894 = \0\aaa\A\CCCC\F\aa\aaa\aaa\G\aa
crosscultural comparison have not previously been addressed, and much work
remains to be done.
Aramaic
The Aramaeans, who originally inhabited a large portion of modernday Syria,
are first recognizable in the archaeological and written records around the end of
the second millennium bc. During the ninth and eighth centuries bc, Aramaeans
ruled a number of small states in the Levant, until these came under the domination of the Assyrian empire. Around this time, they developed a consonantal script
on the model of the preexisting Phoenician consonantary. By the eighth century
bc, Aramaic inscriptions began to include numerical signs, shown in Table 3.2.
The system is purely cumulativeadditive for numbers up to 99, written (like
the script itself ) from right to left, using signs for 20, 10, and 1. The unitsigns are
grouped in threes, since up to nine such signs could be required. Occasionally,
when an ungrouped unitstroke was present in a numeralphrase, it was written
at a slight angle (so that 7 would be 0\aaa\aaa). Because there was a sign for 100, no
more than four 20signs and one 10sign would ever be required, obviating the
need for such groupings for higher values. The 10sign appears to have originally
been a simple horizontal stroke, with a tail added cursively. The 20sign is almost
certainly a ligatured combination of two 10signs, as shown by the occasional use
of a variant form B. There is a gradual trend over time toward the use of a special
sign for 5 (H), which Lidzbarski (1898: 199) notes appearing on an Assyrian clay
tablet as early as 680 bc. However, the majority of Aramaic numeralphrases do
not use a symbol for 5.
Above 100, the Aramaic numerical notation system is multiplicativeadditive
rather than cumulativeadditive, and it is thus a hybrid system. To form 800, for
instance, eight unitsigns (appropriately grouped) were placed in front of the sign
for 100 in order to indicate that the values should be multiplied. The same principle was followed for the thousands. There were apparently two signs for 1000;
the first, G, is actually no more than the final two letters of the Aramaic lexical
Levantine Systems
71
numeral LP thousand (Gandz 1933: 6970), while the second, , is the same as
the corresponding Phoenician numeralsign (Lidzbarski 1898: 201202).
While there is no distinct sign for 10,000 in the Aramaic system used in the
Levant (though see the following discussion for Egyptian variants), rarely numbers greater than 9,999 were written using 10 and 20signs in conjunction with
the sign for 1000. Fractions are apparently found in a handful of inscriptions in
which ungrouped unitstrokes aaaa and aaaaa mean 1/4 and 1/5, and one inscription
contains a special sign for 2/3 () (Lidzbarski 1898: 202), but they were normally
written out lexically.
The first Aramaic inscription with numerical notation is an eighthcentury bc
ostracon from Tell Qasile, in which 30 is expressed as three horizontal strokes (*)
rather than the normal form (Lemaire 1977: 280). However, it may be a Hittite
hieroglyphic numeralphrase (Chapter 2), since that system was still in use in the
eighth century bc in the NeoHittite kingdoms to the north. The earliest uncontestable examples are from the Assyrian bronze lionweights found at Nimrud by
Layard in the nineteenth century. These eighthcentury inscribed weights have
texts in Aramaic and Akkadian; on the largest (BM 91220; CIS II/1, 1), dating to
the reign of Shalmaneser V (726722 bc), the number 15 indicates the objects
weight of fifteen minas in three different ways on its three lines of text: in Aramaic
lexical numerals, as fifteen ungrouped single strokes, and according to the structure detailed above (aa\aaa\A) (Fales 1995: 35). This threefold repetition using different
methods of representation suggests that the system was unfamiliar, either because
of its novelty or because it was intended for speakers of several languages.
Structural similarities between the Aramaic system and the AssyroBabylonian
common system (Chapter 7), with which it shares a decimal base and the use of
multiplicativeadditive structuring for the hundreds and thousands, suggest a historical connection (Gandz 1933: 69; Ifrah 1985: 356). The conquest of the Aramaeans in 732 bc by the Assyrian empire establishes a clear historical context in which
this transmission could have taken place. The lionweights from Nimrud may
well have been taken from the Levant as war booty shortly after this time
(Fales 1995: 54). Yet the Aramaic system is also similar to the Egyptian hieroglyphic
system. Aramaic speakers would certainly have had considerable contact with
Egypt in the eighth century bc, and by the sixth century bc the Aramaic script was
being used by settlers in Egypt at Elephantine and Saqqara. There are a number
of similarities in the forms for signs. Like the Egyptian hieroglyphs but unlike the
AssyroBabylonian system, Aramaic uses vertical unitstrokes grouped in threes to
express the units. A relationship between Aramaic A and hieroglyphic r (both
signifying 10) has also been postulated (Schroder 1869: 186), although it is more
likely that the hooked Aramaic sign is simply a cursive alteration of a horizontal
stroke. Regardless, both signs are very different from the cuneiform AssyroBabylonian
Numerical Notation
72
Aramaic
424 =
\aaaa\C\F\aaaa
4
Hieroglyphic
424 =
qqqqrr
4
AssyroBabylonian
424 =
20 100 4
20
400
4i\b4
4 100 20 4
system. The Aramaic use of unit fractions along the Egyptian hieroglyphic model,
including the exception of having a special sign for 2/3, further suggests Egyptian
borrowing. West Semitic accounts, like those in the Egyptian hieroglyphic and
Aegean scripts, are written with the item being enumerated placed before the
numeral, in contrast to Mesopotamian texts, which follow a quantity + item
order (Levine 2004: 435). Finally, Egyptian hieroglyphic numeralphrases are primarily written from right to left, as in Aramaic, whereas the AssyroBabylonian
system runs in a leftright direction, although of course the Aramaic script is also
written right to left, so this cannot be taken as positive evidence in its own right.
These differences are compared in Table 3.3.
To muddy the waters even further, two other Hieroglyphic numerical notation
systems were used in the eastern Mediterranean around 750 bc and could potentially have been known to the early users of Aramaic numerals. The NeoHittite
kingdoms, although on the wane by that time, were still present in southeastern
Anatolia, immediately abutting the Aramaeans. Moreover, the Cypriote numerals
were invented just before that time, and there was enormous trade between Cyprus
and the Levantine coast. But while the Cypriote and Hittite systems are cumulativeadditive and decimal and use vertical strokes for 1 and horizontal strokes for 10,
they lack the other characteristics that might identify them as potential ancestor
systems.
The Aramaic numerals were likely developed under a dual cultural influence
from Egypt and Mesopotamia. The systems basic structure is very similar to
the AssyroBabylonian common system, but many paleographic and contextual
similarities are far more similar to the Egyptian hieroglyphs. Geographically and
historically, the Aramaeans and other Levantine peoples were peripheral to both
civilizations in the mid first millennium bc, at the time of the systems invention.
Levantine Systems
73
Although cultural phylogenies for scripts and numerical notation systems are usually arranged in accordance with a biological taxonomic scheme, cultural phenomena may have multiple origins, each making a contribution to the descendant,
much as biological parents contribute to a childs genetic makeup. If this explanation is correct, we must ask why the Egyptian hieroglyphic numerals, rather than
the hieratic, would be chosen as a model for the Aramaic numerals. As I discussed
in Chapter 2, the hieratic numerals were widely used in the Kingdom of Judah
in the first half of the first millennium bc. Ostraca from Arad in the Negev dating to the late seventh and early sixth centuries attest both hieratic and Aramaic
numerals (Levine 2004: 433). Like Millard (1995: 190191), I find the failure of the
Aramaeans to adopt the hieratic numerals to be rather curious.
The existence of a distinct sign for 20 in Aramaic, and the recombination of
features of two quite different systems, demonstrates that the Aramaeans were
numerically inventive. In most of the Semitic languages, the word for twenty
is etymologically the plural of ten for example, Hebrew eser ten versus esrim
twenty (Menninger 1969: 14). This may explain why the graphic etymology of
the Aramaic numeralsign for 20 is two ligatured 10signs. This development of
a special sign for 20 outside the regular decimal base of the numerical notation
system is a unique development of the Levantine numerical notation systems; neither the AssyroBabylonian system nor the systems of the Hieroglyphic phylogeny
have this feature.
Like the script to which it was attached, the Aramaic numerical notation was
used in the Levant, Egypt, Mesopotamia, and farther afield throughout the second
half of the first millennium bc. Segal (1983) gives ample evidence for the use of
the system among Aramaic texts from Saqqara in Lower Egypt throughout the
fifth and fourth centuries bc, and Aramaic papyri found at the fifthcentury bc
military colony at Elephantine demonstrate the use of the system in numerous
administrative documents. While the system as used in the Levant had no special
sign for 10,000, the Aramaic papyri found at Saqqara and Elephantine do use such
a sign (), which obeys the multiplicative principle in the same way as detailed
earlier for 100 and 1000 (Segal 1983: 131; Ifrah 1985: 335). An alternate sign for 100
() was also used in Egyptian Aramaic, but it resembles none of the signs used
in the Levant and is not similar to any of the Egyptian demotic or hieratic signs
used at that time.
The Aramaic script was widely used throughout the Achaemenid Empire from
the sixth to fourth centuries bc on clay administrative and legal tablets, stone
monuments, and leather and papyrus. While the scripts used in official royal proclamations and dedications were Old Persian, Babylonian, and Elamite, Aramaic
was the lingua franca of the empire and served most administrative functions. As
such, it was used as widely as Lower Egypt, Asia Minor, and the Transcaucasus and
74
Numerical Notation
even as far east as the Indus River. Throughout its history, the Aramaic numerical
notation was used extensively on monumental inscriptions, ostraca, and administrative papyri. In literary and religious texts, however, numbers were more often
written using lexical numerals only. Aramaic numerals were used to record the
results of calculations used in commerce and administration, but none of the
extant inscriptions demonstrate the use of written arithmetic.
The end of the Achaemenid Empire did not spell the end of Aramaic influence over the Middle East; however, it did result in the fragmentation of what
previously had been a unified script and numerical notation into several regional
variants. During this period, Greek alphabetic numerals were often used administratively, although Aramaic numerals continued to be used in a variety of contexts. By the second century bce, political and ethnic divisions in the Near East
had led to the emergence of variant numerical notation systems. The Hellenized
Palmyrene, Nabataean, Hatran, and Edessan Syrian populations of the Levant
each possessed its own variant numerical notations based on Aramaic. In these
variants, the use of a distinct sign for 5 was far more prominent than in Aramaic
numerals. The Kharoh numerical notation used in parts of modern Afghanistan
and Pakistan, and the Middle Persian and Pahlavi systems used in Iran, are also
variant forms of Aramaic.
Phoenician
The Phoenicians, who inhabited various cities (Tyre and Sidon foremostly) along
the Levantine coast in the first millennium bc, were perhaps the greatest mercantile people of the ancient Mediterranean. The Phoenician consonantal script
was a descendant of the earlier Canaanite consonantary that diverged from its
ancestor late in the second millennium bce. However, none of the earliest Phoenician inscriptions contain numerical notation. While the Aramaic writing system
developed from the earlier Phoenician, the Aramaic numerals appear to be slightly
prior to the Phoenician, and there is no reason to assume that script and numerical notation must be borrowed jointly. The Phoenician numerical notation system
is similar in structure to the Aramaic, with distinct signs for 1, 10, 20, 100, and
1000. These signs (including some paleographic variants) are shown in Table 3.4
(cf. Schroder 1869; Lidzbarski 1898; Gandz 1933; van den Branden 1969: 4243).
Like Aramaic, this system is purely decimal with the exception of the 20sign,
cumulativeadditive below 100 and multiplicativeadditive thereafter. Unitsigns are
simple vertical strokes, although a leftslanting stroke is often used for ungrouped
single strokes, and are grouped in threes, as in the Egyptian hieroglyphic and
Aramaic systems. Like the Phoenician script itself, numeralphrases are nearly
always read from right to left, although van den Branden (1969: 43) notes at least
Levantine Systems
75
10
20
A B @
D
697 = 0\aaa\aaa\ADDDD\\aaa\aaa
100
1000
76
Numerical Notation
is virtually simultaneous with that of the Aramaic system. However, one unitstroke
is scant evidence for this. From the seventh century onward, however, Phoenician
texts containing numerals are relatively common, including inscriptions on stone,
ink writings on clay, administrative documents on papyrus, and, at a somewhat
later period, inscriptions on coins. Phoenician numerals were often used to enumerate regnal years and for record keeping of commodities. However, between
the seventh and first centuries bc, Phoenicia was politically dominated in turn
by the Assyrian, NeoBabylonian, Achaemenid, and Alexandrine Greek empires.
The Phoenician numerical notation thus predominated in the Levant only during its early history. However, in the Phoenician colonies in North Africa and
Spain (including, most importantly, Carthage), the Phoenicians continued to use
the system detailed here, without significant regional variation, until Roman and
Greek conquests in the second century bc effectively ended its use. Coins from
Akko, Tyre, and Sidon used Greek alphabetic numerals as early as 265 bc, though
at Arvad and Marathus, Phoenician numerals were used on coins until about 110 bc
(Millard 1995: 193).
The fruitful transmission of the Phoenician consonantal script throughout the
Aegean and the Middle East has led some to speculate as to the transmission of
its numerical notation system. Millard argues that Phoenician may have been the
model for the Greek acrophonic (base10, subbase 5, cumulativeadditive) numerical notation system (Millard 1995: 192). However, the acrophonic systems subbase
of 5, coupled with the more obvious derivation of acrophonic numerals from the
very similar Etruscan system, makes such an origin unlikely. Schroder (1869: 187f )
suggests that the Lycian numerals (Chapter 4) are a variant of Phoenician, but
that system much more closely resembles the Greek acrophonic numerals than the
Phoenician. It is entirely possible that one or more of the later Levantine systems
have a Phoenician as opposed to Aramaic origin, but there is no good way to
demonstrate this in most circumstances. The lack of a symbol for 5 in Phoenician
numerals suggests that most of these later systems were Aramaicbased. Phoenician numeration, then, unlike the Phoenician writing system, is essentially a side
branch of the broader Levantine family.
Palmyrene
Palmyra was an important mercantile city located in modern Syria around
200 km northeast of Damascus, and whose inhabitants, Aramaicspeaking Semites,
managed to retain considerable control over their own affairs despite Greek and
Roman influence in the area. Palmyrene inscriptions are found dating from the
first century bc to the mid third century ad, continuing the tradition of the earlier
Aramaic script. Palmyrene numerical notation retained much of the structure of
Levantine Systems
77
a
178 =
10
20
H
A
J
aaa\\H\\\A\\J\J\J\\A\a
the older Aramaic system, while introducing new numeralsigns. Despite their
relative obscurity, the Palmyrene numerals were first analyzed over a quartermillennium ago by Swinton (175354). The Palmyrene system had distinct signs
for 1, 5, 10, and 20, as shown in Table 3.5.
These four symbols express any number less than 100. While in earlier Aramaic
scripts the sign for 5 appeared only sporadically, it was a fundamental part of the
Palmyrene system. Because of this, only four unitsigns were required at most, so
there was no need to group sets of unitsigns into threes. Like its Aramaic ancestor, Palmyrene numerical notation is base10 and cumulativeadditive below 100.
For numbers greater than 100, Palmyrene, like Aramaic, is multiplicativeadditive,
with the complexity that the sign for 100 is identical to that for 10. The possibility of confusion is avoided by the requirement of having one or more unitsigns
before the 100sign, whereas no such signs could precede a 10sign.
While this feature resembles the use of the positional principle, such phrases are
multiplicative, not positional. To represent 100, the sign A had to be combined
with unitsigns; alone, it always meant 10, not 100. Cantineau (1935: 36) contends
that the original Palmyrene sign for 100 was a horizontal stroke placed above a
dot, but that it was later reduced until it was identical to the 10sign. If so, the
identity of the two signs may be largely coincidental. In monumental inscriptions
on stone, Palmyrene numerals are among the clearest and most unambiguous of
all the Levantine systems. Figure 3.1 is a memorial inscription dating to the year
492 (ad 181); the phrase is clearly visible on the last three lines of the inscription
(Arnold 1905: Plate IV).
Palmyrene numerical notation was restricted geographically and temporally to
the city of Palmyra during the period from about 100 bc to 275 ad. During that
time, however, it was used widely on inscriptions and records of commercial transactions, though not normally in literary contexts. We do not know the extent to
which it may have been used in a broader range of genres due to the poor survival
of evidence.
The importance of Palmyra as a commercial center rested on its strategic location on the Roman frontier and its trade ties with peoples outside the empire.
Despite considerable Hellenisation and Latinisation, Palmyra retained its script
and numerical notation through the third century ad, though Greek alphabetic
78
Numerical Notation
Figure 3.1. A Palmyrene memorial inscription dating to ad 181; the yeardate (492) is
clearly visible on the final three lines of the inscription, including multiplicative use of the
sign for 100. Source: Arnold 1905: Plate IV.
and Roman numerals came to be used more frequently for administrative and
mercantile purposes. In 273 ad, following the shortlived independent rule of
Queen Zenobia over the province (266272 ad), Palmyra was destroyed by the
Roman emperor Aurelian, abruptly ending its importance as a commercial center.
Thus, political factors, rather than criteria of function and efficiency, led to the
complete replacement of the Palmyrene numerical notation system by those of
Greek and Roman colonizers. It has sometimes been argued that Palmyrene is
ancestral to the Syriac numerical notation, though I will show below that this is
only one of many possible scenarios of transmission.
Nabataean
The Nabataeans were a South Semitic people of Arabian ancestry who inhabited
the area between Syria and Arabia in the southeastern Levant in the late first millennium bc and into the Christian era. Though not Aramaeans, they came under
considerable Aramaean influence and adapted the Aramaic script for their South
Levantine Systems
79
R
N
178 = aaa\\NA\\J\J\J\\I\a
10
20
100
Semitic language, including a variant of its numerical notation. This system was
used from approximately 100 bc to 350 ad in inland areas of the Levant (modern southern Syria and Jordan) including the cities of Damascus and Petra, and
even as far south as the port of Aqaba on the Red Sea. Its signs are indicated in
Table 3.6.
As with all the Levantine systems, Nabataean is decimal, cumulativeadditive
below 100 and multiplicativeadditive above, with additional signs for 4, 5, and 20.
Unitstrokes are grouped in threes where necessary and are sometimes joined together
at the base in cursive writing. The sign for 4 is used only in some inscriptions, and then
only in numeralphrases for 4 (never 5 through 9); 8 is expressed as 111N (5 + 3) or
11\111\111, but never (to my knowledge) as RR. Lidzbarski (1898: 199) argues that its
shape represents four unitstrokes placed in a cross, strictly on graphic principles,
but this is unproven. Its historical connection with the identical Kharoh sign for
4 is still unclear, but some link seems probable, as the Nabataeans were frequently
engaged in commerce with peoples to the east. However, Gibson (1971: 13) notes
that the eighthcentury bc Samaria ostraca, in which the Hebrew variant of the
Egyptian hieratic numerals (Chapter 2) predominates, contain a + or Xshaped
sign for 4, which would antedate either the Nabataean or Kharoh symbol by
several centuries. Finally, Cantineau (1930: 36) and Lidzbarski (1898: 199) believe
the signs for 4 and 5 to be quite late inventions, possibly independent of any other
system.
The Nabataean sign for 10 is a more arched version of the hooked horizontal
stroke used in most Levantine systems, while the 20sign can easily be shown to
derive from the Aramaic form. In one inscription from Egypt, the yearnumber 160
(ad 266) is written irregularly as 100 20 10 20 10 instead of the expected 100 20
20 10 10; this seems unlikely to be a scribal error but is otherwise unexplained (Littmann and Meredith 1953: 16). The sign for 100 is not obviously related to that of any
other notation, though Cantineau (1930: 36) argues for its possible derivation from
Phoenician . The sign for 100 combines with signs for 1, 4, and 5 multiplicatively.
Accordingly, the 4sign is used to express 400, as in an inscription from Dumr (near
Damascus) from 94 ad in which the number 405 is expressed as NIR (4 100 + 5)
(Cooke 1903: 249). Such numeralphrases make the system look less cumulative
than it actually was. No Nabataean writings contain numbers higher than 1000.
80
Numerical Notation
Hatran
A variant Aramaic script was used in the region around the city of Hatra (modern
AlHadr, in northern Iraq), an outpost of the Parthian Empire and later the capital
of the small autonomous state of Araba. The Hatran script, for which inscriptions
have been found dating from about 50 bc to 275 ad, possessed a distinct numerical
notation system with signs for 1, 5, 20, and 100, as shown in Table 3.7.
As with all Levantine systems, the Hatran numerical notation is decimal, cumulativeadditive for numbers less than 100, multiplicativeadditive above 100, and written
from right to left. The precise relation of the Hatran system to the other Levantine
systems is unclear, but it is descended in some way from the Aramaic system used
around Hatra in the centuries prior to the development of the Hatran script, given
the similarity of signs for 1, 10, and 20 to earlier Aramaic forms. The sign for 5 is
identical to that of the Old Syriac script used around Edessa at that time. Finally,
the 100sign is of entirely mysterious origin, though a case could be made that it
is related to the Phoenician .
Levantine Systems
81
10
20
>
100
697 = 11>AJJJJ1>
The Hatran numerals were probably ancestral in some way to the Middle
Persian numerals, at least giving rise to some of the Middle Persian numeralsigns. Hatran numerical notation was used widely on ostraca, on inscriptions on
stone, and in economic documents. Unlike the Palmyrene and Nabataean states,
which were subjected to Roman political and economic domination for most of
their history, Hatra remained independent from both Roman and Parthian control until 272 ad, when the Middle Persian king Shapur I conquered the region.
After this time, Hatran inscriptions are more rarely encountered, and the Middle
Persian script and numerals (and its successors) replaced them.
Old Syriac
A consonantal script was used at the ancient city of Edessa (modern Urfa, in
southeast Turkey) in the early years of the Christian era. Based on an Aramaic
model, this script, which resembles the estrangela Syriac script that emerged in
the later manuscript tradition, was used to write the Old Syriac language, a close
relative of Aramaic. A large number of Old Syriac inscriptions on stone, mosaics,
and parchment have survived, dating from the beginning of the Christian era to
500 ad, and largely found in northern Syria and southern Turkey.1 The Old Syriac
numerical notation system used signs for 1, 5, 10, 20, and 100, and possibly also 2
and 500, as shown in Table 3.8 (cf. Rdiger 1862, Duval 1881, Segal 1954).
Numeralphrases, like the script itself, are always written from right to left. The
system is decimal (with a special sign for 20), cumulativeadditive for numbers
less than 100, and multiplicativeadditive for higher values, all of which points to
its membership in this phylogeny and its close relationship to Aramaic. However,
it has some curious features. The sign for 2, as defined by Duval (1881: 1415), is
simply a ligatured form of two unitstrokes, a paleographic convenience that was
never used consistently or regarded as a structural feature of the system. The sign
for 5 is identical to that of the Hatran system, and the sign for 20 to one variant
form used in Phoenician. The sign for 5 was not consistently used in numeral
1
The earliest dated Syriac inscription is from ad 6, although Drijvers and Healey (1999: 17)
argue that it may have originated earlier.
Numerical Notation
82
10
>
A
697 = >A@@@@
20
100
500
phrases for 5 through 9 (Segal 1954: 35). For instance, in three separate inscriptions
dating to ad 165, the date 476 of the Seleucid era is written in three different ways,
as indicated in Table 3.9 (Drijvers and Healey 1999). The earliest dated Old Syriac
inscription (As55, dating to ad 6) expresses the yeardate (317) using seven unitstrokes (Drijvers and Healey 1999: 141). This suggests that the sign for 5 may have
been a later development, and at best an occasional one.
Duval (1881: 14) argues that the 100sign is a slightly modified form of the 10sign, resembling the Palmyrene numerical notation in this respect. The Old Syriac
symbol for 500 is rare, partly because numbers of this magnitude are infrequent
in Syriac writings. Duval (1881: 14) insists that it ought to be understood in many
numeralphrases where it is not written, such as in yeardates. However, in the
Old Syriac slave sale contract, P. Dura 28, written at Edessa in ad 243, 700 is written multiplicatively as seven unitstrokes (some ligatured) followed by the powersign for 100, rather than with a sign for 500. The notion of a sign for 500, if not its
form, may have been borrowed from the Roman numerals, given the cultural and
political dominance of Rome in Syria throughout the period.
Old Syriac numerical notation was used on numerous inscriptions on stone
around Edessa, but not in the shortlived tradition of mosaic inscriptions from the
third century ad. It evidently originated as a variant of the Aramaic system, but
its specific relationship to the other Levantine numerical systems remains unclear.
Three Old Syriac legal texts written on parchment survive, dating from ad 240243;
of these, only P. Dura 28 (just discussed) contains Syriac numerals (Drijvers and
Healey 1999: 232235; Goldstein 1966). In many inscriptions and manuscripts,
lexical numerals were used instead of numerical notation. Although the Edessan
Table 3.9. Old Syriac yeardates for 476 / ad 165
Inscription
Date
As29
a>A\\\aaaa
As36
aaaaaaA\\\aaaa
As37
aaA\\\aaaa
Transliteration
(1 + 1 + 1 + 1) 100 + 20 + 20 + 20 +
10 + 5 + 1
(1 + 1 + 1 + 1) 100 + 20 + 20 + 20 +
10 + 1 + 1 + 1 + 1 + 1 + 1
(1 + 1 + 1 + 1) 100 + 20 + 20 + 20 +
10 + 1 + 2 + 1 + 2
Levantine Systems
83
Christians were subject to Roman imperial authority throughout most of the history of their script and numerical notation, and although both Greek alphabetic
and Roman numerals were widely used in Syria, the Old Syriac numerals were not
displaced by either system.
By the fifth and sixth centuries ad, however, the older numerals began to be
replaced by the cipheredadditive Syriac alphabetic system (Chapter 5), which
assigned numerical values to the twentytwo letters of the Syriac consonantary. This
gradual obsolescence corresponds to the development of the indigenous Syriac Christian manuscript tradition, particularly oriented toward liturgical subjects. Many texts
use the two systems side by side. A seventhcentury Syriac religious commentary (BM
Add. 14,603) contains several lines of Old Syriac numerals that are incomprehensible
until the numeral values are converted into their corresponding values in the Syriac
alphabetic numerals; the resulting alphabetic signs can then be read as the authors
epigraph (Wright 1870: II, 586587). While this demonstrates that the numerals were
still in use, the cryptographic nature of the note suggests that they may not have been
well known. The very latest evidence of the Old Syriac system is from the eighth century, after which only alphabetic numerals were used (Duval 1881: 15).
Kharos.t.h
The Kharoh script was used in the region of Gandhara in eastern Afghanistan and
northern Pakistan from around 325 bc to 300 ad and, from the second century ad
onward, in parts of Central Asia. Given that this region was under the control of the
Achaemenid Empire (for which Aramaic was a lingua franca) from 559 to 336 bc, the
similarity in form and value of many of the signs in the two scripts, and their common righttoleft directionality, Kharoh is clearly descended from Aramaic. During
the earliest periods of its use (before about 100 bc), Kharoh inscriptions containing
numerals are quite rare, being found in only a few royal inscriptions of the Mauryan
King Aoka, who reigned from about 273 to 232 bc. Only the numbers 1, 2, 4, and 5
are represented, and they are always formed using simple unitstrokes. In the later Saka,
Parthian, and Kusana inscriptions (dating from about 100 bc onward), a more complex
system was used, and much larger numbers were represented. This system possessed
unique signs for the numbers 1, 4, 10, 20, 100, and 1000, as shown in Table 3.10 (cf. Das
Gupta 1958: Table XIV; Salomon 1998: Table 2.6; Glass 2000: 139143).
In common with all the Levantine systems, Kharoh is purely cumulativeadditive up to 100 and multiplicativeadditive thereafter. As in the script as a whole
(and in other Aramaicderived scripts), the direction of writing is always from right
to left. Unlike other Levantine systems used at the time, Kharoh has no special
sign for 5; numbers from 4 through 9 were always expressed through combinations
of units and 4signs. Unitsigns in the cursively written texts of Central Asia are
Numerical Notation
84
20
100
J
697 = aaa\R\\JJJJ\\aaR
10
1000
usually ligatured together in groups of two and three. The sign for 1000 is found
only in the late (perhaps fifth century ad) texts from Inner Asia (Das Gupta 1958:
259; Mangalam 1990: 48; Glass 2000: 143). It is most likely a variant of the similar Aramaic sign, which was a conventionalized version of the lexical numeral for
1000, LP (Salomon 1998: 64). The signs for 100 and 1000 combine multiplicatively
with signs for units less than 10, with the units to the right (before) the powersign.
Figure 3.2 is the obverse of a leather text written in cursive script found at Niya by
Sir Aurel Stein; the numerals 3 and 25 (20 + 4 + 1) are readily visible at the bottom
of the text (Boyer, Rapson, and Senart 1920: 120, Plate V).
The Aokanperiod system of vertical strokes may or may not be of Aramaic
origin, though the geographical proximity of its users, coupled with the obvious
relation of the Kharoh alphasyllabary to the Aramaic consonantary, suggests
that it was. In its fully developed form, however, it is definitely part of the Levantine phylogeny, and not related to the Brhm numerals (Chapter 6) used in India.
Kharoh shares with the other systems the righttoleft direction of writing, the
use of vertical strokes for units, similar forms for the numeralsigns for 10 and 20,
and the use of the multiplicative principle for 100.
The use of X\ for 4 is common to both Kharoh and Nabataean, and this is
unlikely to be coincidental, since they share a common sign for 20 as well, and
both systems developed around 100 bc. These are the only two cumulative systems worldwide ever to use a special sign for 4. As mentioned earlier, the Hebrew
hieratic ciphered sign for 4 was + or X, suggesting transmission from west to east.
However, Datta and Singh (1962: 23) argue that the sign may have developed by
rotating the Brhm sign for 4 (+) by fortyfive degrees, and may then have been
transmitted westward to the Nabataeans. Buhler (1896: 73), in turn, contends that
the Nabataean and Kharoh signs were invented independently of one another.
The Kharoh numerical notation system was used primarily on inscriptions on
stone and on copper, but there are also surviving documents from Inner Asia written on wood, palm leaf, birch bark, and leather (Salomon 1996: 378). Throughout
its history, it was in competition with its rival, Brhm numerals (Chapter 6),
the system used on the Mauryan inscriptions of the Indian heartland. The use
of Kharoh was tied to the political independence of the Greek, Scythian, and
Parthian kingdoms, which looked to Bactrian and Iranian traditions rather than
85
Figure 3.2. A Kharoh leather text found by Sir Aurel Stein at Niya. The numeral 25 is clearly visible at the bottom left of the page (20 + 4 + 1).
Source: Boyer, Rapson, and Senart 1920: Plate V.
86
Numerical Notation
to Indian ones. By the late third century ad, the Bactrian and IndoScythian polities
of the Kharoh heartland were seriously weakened, and the advent of the Gupta
Empire in the fourth century ad heralded the predominance of Brhm throughout
the Indian subcontinent. However, Kharoh survived longer in the small states of
Inner Asia. Inscriptions on wooden documents from the city of Niya date to as late as
the seventh century ad, and contain a variant of the Kharoh script and numerals.
Middle Persian
The Persians of the Achaemenid, Seleucid, and Parthian Empires used the Old Persian numerals (Chapter 7), the Greek alphabetic numerals (Chapter 5), or the common Aramaic numerals already described. The Sassanian dynasty began in ad 228
when Ardashir I destroyed the Parthian Empire, which had ruled much of the territory of modern Iraq and Iran for several centuries prior. For several centuries, the
Sassanian empire was a rival of Rome and later Byzantium to the west, and of the
Gupta Empire in India to the east, and was the dominant power in Mesopotamia
and Persia until the Islamic conquest. The Middle Persian language (the ancestor of
modern Farsi), the language of Sassanian administration and commerce, was written
in an Aramaicderived consonantary, reflecting the legacy of Achaemenid, Seleucid,
and Parthian rule in the region. The numerical notation system associated with the
Middle Persian script is shown in Table 3.11 (Frye 1973).
Like its ancestor, the Aramaic system, the Middle Persian system was cumulativeadditive and decimal for numbers below 100, and written from right to left with
the highest powers at the right. The signs for 1, 10, and 20 resemble closely the
signs used in the other contemporary Levantine systems, while the sign for 100
resembles only that of the Hatran system. Unlike most of the later variants of
Aramaic, however, there was no sign for either 4 or 5, and units from 5 to 9 were
written using grouped sets of three or four unitstrokes. The sign for 1000, as in
Aramaic and Kharoh, is a reduced version of the Aramaic lexical numeral LP
thousand. For numbers above 100, the signs for 100 and 1000 combined multiplicatively with cumulative numeralphrases, as in the other Levantine systems.
Thus, in the Qaba inscription of Kartir, the number 6798 is written as shown in
Table 3.11 (Frye 1973: 4).
It is impossible to determine the precise historical affiliation of the Middle
Persian system to the other Levantine systems, other than to note that it is most
definitely descended from the Aramaic system in some way. The lack of a sign
either for 4 or 5 is quite unusual for such a late descendant of Aramaic, as all of the
other contemporary Levantine systems have some such sign. The Middle Persian
script is most closely affiliated with the Hatran script, and the two systems share
similar signs for 100, suggesting a historical connection. Middle Persian numerals
Levantine Systems
87
10
20
100
1000
2
3
4
5
6
6798 = 1111\1111\24444\5\111\1111\7\111\111
were employed on silver bowls and plates to indicate weights, inscribed on stone
texts, and written in ink on ostraca.
As the Middle Persian period progressed, the script and numerals tended to be
written increasingly cursively, with signs ligatured together. The Middle Persian
Empire came to an abrupt end in ad 637 after the childking Yezdigird III was
overthrown by the Islamic Umayyad caliphs. By that time, the Middle Persian
script had diverged into several variants, one of the more important of which
was Book Pahlavi. The Book Pahlavi numerals are sufficiently different from their
Middle Persian ancestor to warrant separate treatment.
Sogdian
The Sogdian language was an Iranian language closely related to Middle Persian
but spoken further to the north, in modern Uzbekistan and Tajikistan. Sogdian
was written using three separate scripts: the Sogdian script descended from Middle Persian, the Manichaean script used by Sogdian followers of that religion and
descended from the Estrangelo Syriac script, and the Christian Sogdian script
used by Nestorian Christians and descended from Nestorian Syriac (Skjaerv
1996). This religious and scriptal pluralism greatly complicates the history of Central Asian Iranian scripts and numerals. The Sogdian script is first attested from
the Ancient Letters dating to ad 312313 found by Stein in Chinese Turkestan,
but may have originated in the third century. The Sogdian script proper and the
Manichaean script had distinct numerical notation systems of the basic Levantine
structure, which I will treat in turn, while the Christian Sogdian script used lexical numerals or Syriac alphabetic numerals (SimsWilliams, personal communication). There has been no systematic comparative treatment of Sogdian numerals
to date, and minimal paleographic work. The Sogdian numeralsigns are shown in
Table 3.12 (cf. Sundermann and Zieme 1981).
Table 3.12. Sogdian numerals
1
10
20
100
1000
697=
G\GGG\GGGHIIII\JGGG\GGG
88
Numerical Notation
The Sogdian system has numeralsigns for 1, 10, 20, and 100. It is cumulativeadditive below 100 and multiplicative for the hundreds and thousands, with
numeralphrases always written in descending order from right to left. The sign
for 1 is never used to write 1 alone but always phonetically as yw (SimsWilliams,
personal communication). Units from 2 to 9 are usually ligatured, although they
can be arranged in groupings of two to four units, or sometimes ungrouped. There
is no special sign for 5, in contrast to many of the Levantine systems (including Manichaean), but in common with Middle Persian, Kharoh, and Pahlavi.
While the sign for 10 is similar to the Sogdian letter that begins the word s ten,
the similarity is probably the result of later paleographic assimilation rather than
being indicative of an alphabetic origin for the sign. Paleographically, the sign for
20 was originally two superimposed signs for 10, and in the Ancient Letters 30
was occasionally expressed using three such signs (SimsWilliams, personal communication). By the seventh and eighth centuries the Sogdian letters dleth (d)
and ain () had become assimilated to the forms of the numerals 20 and 100,
respectively (Livshitz 1970: 259). The sign for 1000 is not a numericalsign per
se, but, as in Aramaic and Middle Persian, an abbreviated ideographic form of
the Aramaic word LP thousand; similarly, 10,000 is written using an ideogram
RYPW (Aramaic ribb) (SimsWilliams, personal communication). Fractions are
poorly understood, although Grenet, SimsWilliams, and de la Vaissire (1998:
96) suggest that there is a sign for 1/2 that had previously been interpreted as a
variant for 100.
The majority of the texts in which this system was used are religious in nature
(the socalled Sogdian sutra script), although the Ancient Letters are personal
correspondence, and there are a few inscriptions on stone from Pakistan (Skjaerv
1996: 517). Numerals are used ordinally and cardinally in texts in various ways.
Sundermann and Zieme (1981) discuss some fragmentary lists of sequential numbers in the SogdianTurkish word lists used as translation glossaries, one of which
simply lists numerals from 88 to 100, and another of which contains undeciphered
(nonSogdian) numerical symbols associated with the Turkish numeral words
one through five. Numeralsigns could be combined with lexical numerals, as
in the phrases 100 t 108 and wy 100 20 220 in the Padmacintmaidhrastra dating to around ad 700 (Mackenzie 1976: 1217). Multiplicative powerideograms could also be combined together, as in the phrase 100 1LPW RYPW
100,000 myriads (= billions) in the Dhyna text (Mackenzie 1976: 7273).
Following the conversion of most of the Central Asian peoples to Islam, the
Sogdian script was used increasingly infrequently. In the eighth and ninth centuries Sogdian writing was adopted by Buddhist Uygurs, around which time the
script was rotated ninety degrees counterclockwise, resulting in vertical columns
(Coulmas 1996: 472473). By this time, however, it appears that no numerical
Levantine Systems
89
symbols were used (numbers were written lexically). It had largely fallen out of
use by the tenth century ad.
Manichaean
The development of a specifically Manichaean writing system is usually attributed
to Mani himself in the third century ad, although this tale is likely mythical, and,
the script may indeed be older than the religion (Skjaerv 1996). Manichaean
writing is derived from the Estrangelo variety of Syriac, and its numerical system
owes much to its ancestor. It was used to write a wide variety of languages, including Middle Persian, Sogdian, and Uygur. The bestknown Manichaean texts are
those from the oasis of Turfan along the Silk Road, which date from the eighth
and ninth centuries ad. As with many of the Central Asian scripts, Manichaean
numerals remain understudied paleographically and comparatively; the signs
shown in Table 3.13 derive from the manuscript published by Mller (1912).
Manichaean numerals are always written from right to left with the highest
powers written first. The system is cumulativeadditive below 100, with signs for
1, 5, 10, 20, and 100, while the hundreds are expressed multiplicatively. Units are
generally ligatured together, with long flourishes at the end (left) of the phrase.
Figure 3.3 depicts a section of a hymn book (Mahrnmag) in Manichaean script
dating to ad 76162; the yearnumber 546 (dated from Manis birth) is found
on lines 12 (5 100 20 20 5 1), while the last line contains the yearnumber 162
(100 20 20 20 1 1), reckoning from the death of Mari Schad Ormizd (Mller 1912:
36, Taf. II). The presence of a distinct sign for 5 very similar to that of Syriac, as
well as similarities in the sign for 20, suggest that Iranian and Central Asian scripts
such as Middle Persian and Sogdian played little role in shaping the Manichaean
numeralsigns. However, all of the Manichaean numeralsigns, with the exception
of the upright stroke for 1, are assimilated to letters of the Manichaean consonantal script (5 = ; 10 = h; 20 = p; 100 = m), so reconstructing the diachronic paleographic history of the system is complex (SimsWilliams, personal communication). It is unclear whether Manichaean had any signs for 1000 or higher powers.
While the Manichaean religion flourished for several centuries after its peak,
texts in the Manichaean script became less numerous after the tenth century, by
which time it seems to have acquired a dignified and prestigious but also arcane
quality (Sundermann 1997). The Manichaean numerals do not appear to have left
any descendant systems.
Pahlavi
Following the Islamic conquest, the Arabic script was normally used for writing
the Middle Persian language. The Zoroastrian Persians, however, continued to use
Numerical Notation
90
10
20
100
BA
C
D
697 = BACDEEEEFAC
their own Aramaicderived script for their religious texts and for other purposes.
No Persian texts on papyrus survive from the early Middle Persian period, but late
in the Middle Persian period, and following the Islamic conquest, Persian began
to be written using a cursive, highly ligatured version of the earlier Pahlavi script,
known as Book Pahlavi. Alongside this script, a set of numerals was employed
(which I will call simply Pahlavi), shown in Table 3.14 (Abramian 1965: 285;
Mackenzie 1971: 145).
The Pahlavi system is decimal and written from right to left with the highest powers at the right. Frye (1973: 46) established conclusively that the Pahlavi
numeralsigns are cursively derived from those of the earlier Middle Persian system. The signs for 1 through 9 are ligatured and cursive reductions of unitstrokes,
and the Middle Persian practice of grouping unitsigns in groups of three and four
Levantine Systems
91
a b c d bc cc cd dd ccc
10s
f j l m n
o
p
q
r
100s
s
1000s u
4697 = cdrsccud
1s
strokes can also be seen in the phrases for 5 through 9. Like the Middle Persian system, the Pahlavi system is multiplicativeadditive for the hundreds and thousands.
Yet the Pahlavi numerals are structurally quite divergent from their ancestor. The
signs for the tens, in particular, show almost no trace of their cumulative ancestry,
and the unitsigns have largely become ligatured into single signs or, in the case of
5 through 9, into combinations of two or three signs. Combinations of tens and
units were usually ligatured together. The Pahlavi system is thus, for all intents,
cipheredadditive rather than cumulativeadditive below 100. This transformation
from cumulation to ciphering occurred when the epigraphic Middle Persian script
and numerals, written mostly on stone and metal, were transferred to papyrus,
which is more amenable to cursive and ligatured writing. This transformation
is directly analogous to the derivation of Egyptian hieratic numerals from their
hieroglyphic ancestor (Chapter 2), corresponding to the switch in medium from
stone to papyrus. These two instances, in fact, are the only two known cases where
a cumulative system directly gave rise to a ciphered one.
The Zoroastrian Persians continued to use the Book Pahlavi script for their
religious writings and for new pieces of literature into the tenth century ad. Most
surviving Pahlavi numerals are found in these papyrus texts, although there are
also epigraphic texts on stone and metal. After about ad 1000, the Middle Persian
language underwent a set of changes that led to its transformation into Modern
Persian (Farsi). By this time, the abjad numerals (Chapter 5) and Arabic positional
numerals (Chapter 6) had completely replaced the Pahlavi system.
Summary
The Levantine phylogeny is descended from the Aramaic and Phoenician systems
developed around 750 bc, based on the dual model of the Egyptian hieroglyphic
system and the AssyroBabylonian common system. Over the second half of the
first millennium bc, the Aramaic system and its descendants spread throughout
92
Numerical Notation
Assyria, Persia, Egypt, Asia Minor, and even into India and Central Asia. While
these systems were used for over a millennium, they ceased to be used once the
polities in which they predominated (most significantly, Achaemenid and Seleucid Persia) declined in importance. Levantine numerical notation systems steeply
declined in frequency of use after the rise of Islam, as the systems were replaced
by those of the alphabetic (Chapter 5) and South Asian (Chapter 6) phylogenies.
They persisted the longest in Central Asia, where Islam was somewhat slower to
take hold.
The central features of Levantine numerical notation systems are as follows:
a) a decimal base; b) a special sign for 20 (sometimes a combination of two 10signs); c) the use of vertical strokes for units and horizontal strokes (usually with
some degree of curvature) for tens; d) a cumulativeadditive structure for numbers
smaller than 100; and e) the use of multiplicativeadditive notation for expressing multiples of 100 (and also of 1000 and 10,000, where appropriate). Signs for
4 are found in Nabataean and Kharoh. The presence of a sign for 5 used in late
Aramaic, Palmyrene, Nabataean, Hatran, Old Syriac, and Manichaean helps to
clarify some of the relationships among the systems of the family. The late Pahlavi
system, which is heavily ligatured, is essentially cipheredadditive and thus somewhat anomalous, but it shares all the other structural features of this phylogeny,
and is clearly derivative of a cumulativeadditive ancestor.
While these systems were used extensively for administrative and mercantile
purposes, as well as on inscriptions, there is no direct evidence that any Levantine
numerical notation system was ever used as a computational aid. We simply do
not know by what means the users of these script traditions performed arithmetic, but there is no reason to assume that it was done with pen and paper. There
are issues relating to the survival of perishable materials such as wooden tablets,
papyrus sheets, and leather scrolls, none of which survive well in the archaeological record. Yet, even in surviving texts, numbers were often written out lexically in
religious and literary contexts and even occasionally in economic documents. As
such, numerical notation occupied a less significant role in the script traditions of
these societies than would otherwise have been the case. Moreover, in comparison
to the incredibly wide diffusion of Aramaicderived scripts throughout Europe,
the Middle East, and South Asia, Levantine numerical notations spread only sporadically, and their imprint was impermanent. The reason for this deserves careful
attention, and I will return to the question in Chapter 12, after looking at the history of these systems competitors.
chapter 4
Italic Systems
The Roman numerals are undoubtedly one of the betterknown numerical notation
systems, and have received a tremendous amount of scholarly attention. Nevertheless, they constitute only a part of a larger phylogeny of numerical notation systems
that originated, not among Romans, but among Etruscans and Greeks on the Italian
peninsula around 600500 bc. The name Italic refers only to this geographical origin, and thus does not reflect any shared linguistic or cultural affiliation. Italic systems
flourished between 500 bc and 500 ad throughout the Mediterranean region, Western
Europe, and North Africa, under conditions of Greek and Roman cultural hegemony and political domination. Ironically enough, however, the collapse of the Roman
Empire brought about the greatest expansion of one particular system the Roman
numerals in medieval Europe, and ultimately throughout the modern Western world.
The most common variants of the Italic numeralsigns are shown in Table 4.1.
Etruscan
The Etruscans were a nonIndoEuropean people whose civilization had its center
in north central Italy, in the region of modern Tuscany (whose name is taken from
the Latin Tusci, meaning Etruscan). The origins and civilization of the Etruscans
are poorly understood, and large parts of their language remain undeciphered.
Yet Etruscan civilization was the most potent political force on the Italian peninsula between around 800 and 300 bc, and significantly influenced Roman culture
93
Numerical Notation
94
10
50
Etruscan
1 Q R ; \ \
Greek archaic
Greek acrophonic
1 b c d e f g h
Greek Argos/Nemea
d\
Greek Epidaurus
Greek Olynthus
Lycian
1 < 1
Roman
1 P R S\ U W Y\ ,
. ~
Roman multiplicative
1 P R S U W a
d e g
ArabicoHispanic
Calendar numerals
A E J
South Arabian
1 ! @ # $
i j
throughout the Republic and even later. The Etruscan alphabet, developed in the
early seventh century bc on the model of the archaic Euboean Greek alphabet,
usually runs from right to left (Bonfante 1996).
The Etruscan lexical numerals were probably base10 with a special term for
20, zathrum (but not for 40, 60, 80 ...), and it appears that subtractive structures
formed the words for 17 through 19 (Lejeune 1981; Bonfante 1990: 22). However,
these irregularities are not reproduced in the Etruscan numerical notation system,
shown in Table 4.2.
Table 4.2. Etruscan numerals
1
10
50
100
500
1000
5000 10,000
1 Q R :\ \ ;\ VaU U /\
1378 = aaaQRR;;;
?\
Italic Systems
95
96
Numerical Notation
Italic Systems
97
Figure 4.1. The Etruscan abacusgem (CII 2578 ter) showing a figure seated at a board
working with Etruscan numerals. Source: Fabretti 1867: 224.
Figure 4.1 depicts the socalled Etruscan cameo or abacusgem (CII 2578 ter),
a small gem (1.5 cm high) dating from the fifth century bc that depicts a seated
individual working at a large board upon which rows of Etruscan numerals have
been inscribed, including the elusive signs for 1000 and 10,000, but not 500 or
5000 (Fabretti 1867: 224; Keyser 1988: 545). This demonstrates the association of
the numerals with pebbleboard computation at an early date. Similarly, Etruscan
numerals may have been used on wooden tallies and similar perishable materials,
despite the lack of evidence for such a function. There is no evidence, however,
for the use of the Etruscan system for performing arithmetical calculations as we
would (on papyrus or slate). Computations would have been done in the head,
with the fingers, or on a counting board. Large numbers and long numeralphrases
are very rarely encountered; even the sign for 100 is relatively uncommon.
The demise of the Etruscan numerical notation system was a direct consequence
of the rising fortunes of the Roman republic. The lack of Etruscan political unity
in the third century bc, coupled with the political advantages of association with
Rome, led to the slow but steady assimilation of the cities of Etruria into the
Roman political and cultural milieu. While the Etruscans remained a culturally
Numerical Notation
98
10
50
100
500
1000
1 Q P\U R O A ; Z
distinct people at least until the beginning of the Roman Empire, by 100 bc they
were entirely within the Roman political sphere. This inevitable trend was accompanied by the slow replacement of the Etruscan language, script, and numerical
notation with those of the Romans. Given the similarity of the two numerical
notation systems, there would have been little difficulty in making the change to
the new system. The last certainly dated examples of Etruscan numerical notation
are from the second century bc.
However, A. P. Ninni (188889) first presented the theory that the Etruscan
numerals survived into the nineteenth century. While studying the tally marks
used by fishers along the coast of the Adriatic Sea, near the town of Chioggia (near
Venice), Ninni discovered a numerical notation system that he called cifre chioggiotte (numerals of Chioggia). This potential vestige is cumulativeadditive, with a
mixed base of 5 and 10, like both the Roman and Etruscan systems. Its numeralsigns are shown in Table 4.3 (Ninni 188889: 680).
Ninni noted that the signs of the cifre chioggiotte more closely resemble the
Etruscan numerals than the Roman numerals, and on this basis proposed that
they were of ancient origin. Furthermore, in both the Etruscan system and the cifre
chioggiotte, several of the subbase numeralsigns are halved versions of the signs for
powers of ten (Ninni 188889: 680681). Could this system in fact be a survival,
over 2,000 years, of an Etruscan tradition among modern fishers? At present, there
is simply not enough surviving data to speculate on the possibility of such longterm cultural survivals, particularly in a region, such as Italy, that has experienced
immense social change over two millennia. On the one hand, certain signs of the
cifre chioggiotte (e.g., the first signs in Table 4.3 for 50 and 100) are identical to the
Etruscan numeralsigns for those numbers, but are quite dissimilar to the intervening Roman numerals. On the other hand, Chioggia is not in modern Tuscany,
and there is no evidence that the systems users believed the cifre chioggiotte to be
ancient. Since no information has come to light for over a century, perhaps we have
lost our opportunity to learn more about this system.
Greek Acrophonic
Between 750 and 500 bc, what we now call archaic Greece was a conglomeration
of small citystates in mainland Greece, the Aegean islands (including Crete), the
southern half of the Italic peninsula (known as Magna Graecia), and western Asia
Italic Systems
99
10
50 100
500 1000
5000 10,000
50,000
1 b
d e
f g
36,849 = iiihgfeeeccccbaaaa
Minor, sharing in common only the use of Greek dialects. A tremendous number
of local scripts, known as epichoric scripts (from Greek epi, upon, over, and chora,
place, country), were used during this period, all of which were based on the
model of the Phoenician consonantary around 800 bc. In their earliest phases,
some of these alphabets were written from right to left or in alternating directions (boustrophedon), although by around 500 bc all the epichoric scripts were
written from left to right. Adjoining these scripts were two very distinct types
of numerical notation: the acrophonic, to be described here, and the cipheredadditive alphabetic numerals (Chapter 5). For our present state of knowledge of
these two systems, we are greatly indebted to the tireless and unparalleled work of
Marcus Niebuhr Tod.1 The acrophonic system as used in classical Athens is shown
in Table 4.4 (Tod 191112: 100101).
The system is cumulativeadditive, uses vertical strokes for units, has a base of
10 with a subbase of 5, and is always written from left to right, with numeralphrases in descending order of numeralsign value. The acrophonic system is so
named because the signs for many numbers are taken from the first letter (akros =
highest, outermost; phone = sound) of the corresponding (classical) Greek word.
Other names for this system, now largely rejected, include Herodianic and decimal (Tod 191112: 125127). The signs for 50, 500, 5000, and 50,000 combine
the sign for 5 with the sign for the appropriate power of 10. Whether we choose
to see these subbase signs as single signs or as two ligatured multiplicative ones
is largely a matter of definition, and does not substantially affect how we classify
the entire system.
Similar acrophonic signs were used in large portions of the Hellenic world, the
only difference being that the appropriate letters from each epichoric script were
used in place of the letters used in the Attic inscriptions. Dow (1952) notes that
the variety of acrophonic Greek numerical notation systems stands in sharp contrast to the Greek alphabetic system (Chapter 5), which is remarkably consistent
1
Tods six papers on Greek numerical notation (Tod 191112, 1913, 192627, 193637,
1950, 1954) have been reprinted in one volume (Tod 1979). My citations are taken from
the original papers.
Numerical Notation
100
10
50
100
throughout its geographic and temporal range. This degree of variation among
local systems is far greater than the variety of lexical numerals used in the Greek
dialects. However, the differences in signforms were probably not great enough
to affect their comprehensibility (Tod 193637: 246).
For expressing monetary values, the acrophonic numerals were often modified to
reflect the forms of currency being expressed; for example, in Attica, (talanton = 1
talent = 6,000 drachmas), Z (mna = 1 mina = 100 drachmas), (1 stater), (1 drachma),
I (1 obol), (1/2 obol), or (1/4 obol), and R (1/8 obol) (Threatte 1980: 111). These
could sometimes be ligatured to the sign for 5, just as the ordinary acrophonic powers
of 10, to express multiples of units of currency. While there is some potential for confusion ( can mean 1 talent or obol; Z can mean 1 mina or the numeral 10,000, etc.),
numeralsigns are always listed in descending order, which averts most ambiguities. In
some regions, special signs were used to indicate monetary values that did not fit easily
into the standard system. For instance, a system found in inscriptions from Thespiae
(in Boeotia) uses numeralsigns for 30 and 300, which consist of a sign (for triobole,
or 3 obols) ligatured to the appropriate sign for 10 or 100 (Tod 191112: 109; Feyel 1937).
Other acrophonic subsystems used cumulative signs related to systems of weight or
volume, such as those described by Lawall (2000) on graffiti from the Athenian Agora
from the last quarter of the fifth century bc; for example, EEEE = 4 hemichoes.
Despite the name of the system, not all numeralsigns used in the Greek epichoric scripts are acrophonic, and in fact the earliest ones are nonacrophonic.
Johnston (1975, 1979, 1982) has found several instances of a very early Greek
cumulativeadditive but nonacrophonic system with a mixed base of 5 and 10
dating from the sixth and fifth centuries bc. The signs of the system are shown
in Table 4.5 (cf. Johnston 1979: 2930; Johnston 1982: 208). Johnston argues that
this system was built up systematically by cumulatively adding oblique lines to a
vertical stroke to obtain higher numeralsigns. Curiously, he does not note that
the signs for the subbase (5 and 50) are the right halves of the appropriate primary
bases (10 and 100). This structure parallels the halving of Etruscan numeralsigns,
which is notable because many of the examples of this preacrophonic system
are of South Italian provenance. Johnston (2006: 17) notes several sixthcentury
Greek inscribed vases where X = 10, paralleling the Etruscan practice but in contrast with later Greek acrophonic practice.
A very unusual numerical notation system used only to express monetary values
is found in five fourthcentury bc inscriptions from the Greek colony of Cyrene
Italic Systems
101
20
b 2 R
Z 2 c
(in modern Libya). These numeralsigns are nonacrophonic, and their interpretation is controversial (Tod 192627, Oliverio 1933, Tod 193637, Gasperini 1986).
Our best evidence comes from the temple of Demeter at Cyrene, where inscriptions list the prices of various goods and the temples revenues and expenditures
(Tod 193637: 255). They present a dual series of figures in which each numeralsign has both a higher and lower value; the specific amount must be inferred from
the context within the numeralphrase. Normally the higher is 5000 times the
value of the lower, but this breaks down for some of the lower signs. The relative
values of different units of currency used in Cyrene during this period (drachmas,
staters, minas, and talents, where 1 talent = 50 minas = 1250 staters = 5000 drachmas) help explain its unusual structure. The interpretation presented by Oliverio,
Tod, and Gasperini is derived from an analysis of the maximum number of times
each sign is repeated (and is thus open to question if more inscriptions are found).
This system is shown in Table 4.6.2
Still another aberrant acrophonic system is found in fourthcentury bc inscriptions from Olynthus (in the northern Chalcidice region). There, a system was used
that is nonacrophonic and lacks a subbase of 5 (Tod 193637: 248249; Graham
1969). The signs for 10, 100, and 1000 (R, , and , respectively) are the last three
letters of the western Greek alphabet used in the region. Of course, R = 10 is common to the Roman and Etruscan systems as well.3 On this basis, Graham (1969:
356) argues that the Roman/Etruscan system was borrowed from the Chalcidian colony at Cumae (in southern Italy). This theory, while attractive, has several
flaws, many of which derive from Mommsens (1965 [1909]) flawed lostletter
theory of the Roman numerals discussed later. Moreover, the fourthcentury bc
numeralsigns of Olynthus cannot have spread to the sixthcentury bc Etruscans
by means of a colony at Cumae that never used the numeralsigns in question.
I suspect that the Greek letters were borrowed for the higher powers, just as the
Romans began with nonalphabetic numeralsigns, but later modified their signs
into alphabetic ones for mnemonic purposes.
2
The numbers listed are amounts in drachmas, based on the assumption that the lower
Z sign represents one drachma, without which the absolute value of each sign would be
indeterminate.
No significance should be attributed to the fact that the sign _, a common Roman
numeralsign for 1000, is rotated ninety degrees from the Olynthian sign for 100.
Numerical Notation
102
10
Standard Acrophonic
1
1
c
d
None
g
2 None
\\ d\\b
Olynthus
Epidaurus
Argos and Nemea
None
None
None
50
100
500
1000
g
g
None
None
None
A similar system was used in Epidaurus, on the southern Greek mainland (Tod
191112: 103104). It is acrophonic for 100 and 1000 but not for the lower powers.
Nearby, in Argos and Nemea, a closely related system was used that apparently
had a sign for 50, but not for 5 (Tod 191112: 102103; Ifrah 1985: 235). The systems
of Epidaurus and Argos, alone among the Italic numerical notation systems, use
a dot rather than a vertical stroke for 1. Table 4.7 compares the numeralsigns of
these irregular systems to the standard acrophonic system.
Most scholars explicitly or implicitly assume that the acrophonic system was
invented independently of the Roman, Phoenician, and other systems used at the
time (e.g., Ste. Croix 1956: 52). Because of the use of the acrophonic principle,
the numeralsigns are often Greek letters, which makes reconstructing the history
of the system rather difficult. It could be argued that the acrophonic nature of
the system suggests that it could only have been invented in Greece. Yet like the
Roman numerals, the earliest Greek acrophonic numerals are not phonetic signs
at all, which provides crucial evidence allowing us to reconstruct their origin.
While the traditional and widely quoted dates given for the use of the acrophonic system in Athens are 454 to 95 bc (Heath 1921: 30), there is indisputable
evidence of an earlier origin for the system. Tod argues, solely on logical grounds,
that a seventhcentury bc origin is not unreasonable, as the system was fully developed by the middle of the fifth century bc (Tod 191112: 128). Mabel Lang mentions a seventhcentury bc decorated Greek amphora inscribed with three vertical strokes, but this does not prove that the numeral was part of the acrophonic
system; it might have been part of an unstructured tallying system or almost any
numerical notation system in use in the Aegean at the time (Lang 1956: 3). For the
second half of the sixth century bc, however, there is more promising evidence
of the acrophonic system. Johnston (1979: 2729) discusses three different variations of the preacrophonic system mentioned earlier, used in the sixth century
bc in southern Italy, Sicily, western Asia Minor, the Aegean islands, and various
parts of mainland Greece almost the entirety of Greek civilization during that
period. Several vases from southern Italy and Sicily, which Johnston dates to the
last quarter of the sixth century bc, bear marks used in commercial transactions
Italic Systems
103
( Johnston 1975, 1979, 1982). It is telling that so many variations of the acrophonic
system were used in the fifth and fourth centuries bc, suggesting an initial period
of experimentation followed by consolidation and agreement on a single form of
the numerals.
The Greek acrophonic numerals likely originated on the Italian peninsula
around 575550 bc, around the same time and in similar contexts as the Etruscan
system. As I accept Keysers (1988) contention that the Etruscan numerals developed relatively independently as an outgrowth of tally marks, the obvious conclusion is that the Greek system developed on the model of the Etruscan numerals
in southern Italy and Sicily, an area of considerable GrecoEtruscan commercial
and cultural contact. It is difficult to believe that two cumulativeadditive, quinary/
decimal numerical notation systems developed on the Italian peninsula in the
second half of the sixth century bc independently of one another. Yet because
the Etruscans owe their script to contact with the Greeks, it is counterintuitive to
think of the transmission of numerals moving in the opposite direction. In any
event, separating out questions of chronological priority of the two systems is virtually impossible. There may also have been some influence from the Phoenicians,
who were in contact with both the Greeks and the Etruscans in the sixth century
bc. In at least one document, tablet V from Entella in west central Sicily, the early
acrophonic numerals for 10, 50, and 100 were written with the smallest numbers
on the left and ascending to the right perhaps in emulation of the righttoleft
direction of the Phoenician system (Nenci 1995). The Phoenician system, however,
has a special sign for 20, is a hybrid multiplicative system above 100, and does not
have a sign for 50 at all.
In the early classical period, acrophonic numerals were used in Asia Minor,
the Aegean islands, North Africa, southern Italy, and Sicily, in addition to mainland Greece, but their spread to the nonGreek world was relatively limited. The
Lycians of southern Asia Minor used a nonacrophonic numerical notation system
in the late fifth and fourth centuries bc that is probably an epichoric variant of the
acrophonic system, although their language was not Greek (see the following discussion). The enormous cultural debt of Lycia to classical Greece is beyond doubt,
and its geographic and temporal proximity strengthens this hypothesis. Also likely
is the possibility that the South Arabian numerals, which arose in the fifth century
bc, derive from the acrophonic system. The South Arabian numerals are cumulativeadditive, base10 with a subbase of 5, and use acrophonic numeralsigns. However, more evidence of cultural contact is desirable before this hypothesis can be
proven.
Acrophonic numerals are found on inscriptions on stone, lead, and silver as
well as on potsherds; they may also have been used on wood or other perishable
materials, though evidence is lacking. Of the thousands of Greek papyri from
104
Numerical Notation
the fourth century bc onward, only a handful from Saqqara contain acrophonic
numerals (Turner 1975). Inscriptions on stone use acrophonic numerals far more
frequently, including accounts, inventories, lists, regulations, treaties, and boundary markers. As well, graffiti or other marks on pottery often indicate quantities
for commercial purposes. The acrophonic numerals expressed measures of volume
or distance, quantities of goods, or monetary values.
What is notable is the wide range of purposes for which acrophonic numerals
were not used, even compared to other cumulativeadditive systems used in the
Mediterranean in antiquity. Firstly, the numerals could only be used to express
cardinal numbers; ordinal numbers were expressed using lexical numerals or,
when available, alphabetic numerals (Tod 1911: 128). Similarly, with the exception of monetary amounts, there was no acrophonic numeral expression for fractions. The Greeks never expressed dates in acrophonic numerals. The practice of
expressing the age of the deceased at death on funerary inscriptions, a source of
much information on other numerical notation systems, was not customary in
Greece. The custom of dating using regnal years did not arise until the Alexandrine period. Documents in connected prose (decrees, for instance) rarely contain acrophonic numerals, except to indicate the price of executing the inscription (Threatte 1980: 112).
There is no evidence that the acrophonic numerals were used direcly for arithmetic or accounting. For these purposes, as with the Roman and Etruscan systems,
the Greek acrophonic system was complemented by the use of the pebbleboard
abacus, in which several grooves were labeled with the appropriate acrophonic
numerals. Lang (1957) has established that many of the mathematical errors made
by Herodotos demonstrate his use of the abacus to perform calculations, with
which certain types of errors (especially in multiplication and division) can occur
easily. All of the thirteen examples of abaci (and fragments thereof ) known from
classical Greece have the row values inscribed with acrophonic numerals (Lang
1957: 275276). Most notable among these abaci is the remarkably wellpreserved
Salamis tablet, which probably dates from the fifth century bc (Menninger
1969: 299303). The numerals on it range from T (one talent) to X (1/8 obol); the
monetary values of the numeralsigns suggest that it was used for practical commercial computations.
The decline of the acrophonic system is thoroughly entwined with the fate of
the Athenian state as a Greek power. As Athens ceased to be a dominant power
in Mediterranean affairs, acrophonic numerals were used less often; by the third
century bc, they had been supplanted by the alphabetic numerals for most purposes throughout large parts of the Hellenistic world, including Ptolemaic Egypt
and Seleucid Persia. Only in Athens and the surrounding areas did the acrophonic
system continue to flourish. There are only a handful of known firstcentury bc
Italic Systems
105
examples from Athens (Threatte 1980: 113). By this time, Greece was firmly under
Roman control. Yet there is no evidence that the acrophonic system was replaced
by Roman numerals except, as one might expect, in southern Italy, where Latinspeaking populations dominated. However, the use of acrophonic numerals did
continue in one very limited domain stichometry, or the enumeration of lines of
verse in classical texts (Tod 191112: 129130). This practice continued as late as the
third century ad with the writings of the Neoplatonist philosopher Iamblichus.
Such late examples are analogous to the use of Roman numerals in the modern
West, in contexts in which it is very useful to have two separate numerical notation systems for paginating introductory sections versus the body of a work, or
for distinguishing volume numbers from page numbers in certain texts.
Lycian
Lycia was a small state of southern Asia Minor in the middle of the first millennium bc, centered around the city of Xanthus. The Lycians spoke an incompletely
understood IndoEuropean language related to the earlier Luwian language, which
was spoken in the NeoHittite kingdoms of Asia Minor. Lycia occupied an intermediate position between the Greek and Persian spheres of influence, and was
intimately involved in interregional commerce and conflict. The Lycian alphabet,
which was developed around 500 bc, is an epichoric variant of the Greek script,
like many others used in the Greek peninsula and western Asia Minor, except that
the language of the inscriptions was not a Greek dialect. A few hundred instances
of the Lycian script have survived, mostly from inscriptions on stone and on coins;
they are written almost exclusively from left to right and date from the fifth and
fourth centuries bc.
The Lycian numerical notation system is still very poorly understood. The
Lycian system, like the Greek acrophonic, Etruscan, and Roman systems, is cumulativeadditive with a base of 10 and a subbase of 5. However, the exact values of the
numeralsigns are still in debate. The signs of this system are shown in Table 4.8
(cf. Shafer 1950, Bryce 1976).
There is also a sign, 2, that probably represents , although Shafer (1950: 260)
argues that it may represent an additional onehalf of any numeralsign that precedes it; 2 would be 15 and <2 would be 7 according to this theory. The
numeralsigns for 50 and 100 are found only on a few inscriptions. The value 50
is assigned to primarily by default; its value is certainly between 10 and 100 (it
is found after 1 but before ). I follow Frei (1976: 15) in assigning it the value of
50. We can be fairly certain about the value of the sign for 100, because it is found
in the Lycian portion of a trilingual GreekLycianAramaic inscription found at
Letoon and dating from 358 bc (Frei 1976: 1315). Shafer (1950: 258259) suggests,
Numerical Notation
106
Table 4.8. Lycian numerals
1
<
127 = 1<11
10
50
100
based on one inscription, that the Lycians may have used the subtractive principle
to express the number 4 as a<. However, in other inscriptions, 4 is expressed as aaaa.
The Lycian numerals are very likely a previously unidentified variant of the
Greek cumulativeadditive systems. Lycian numerals arise in the early fifth century bc at the time of the peak use of the Greek acrophonic numerals in Athens
and throughout the Hellenic world. Both systems were cumulativeadditive, had
a base of 10 with a subbase of 5, and were used in the fifth century bc in the
Aegean region. While the Lycian numerals are not acrophonic, this is true of many
of the epichoric numerical notation systems of the classical Greek world. Alternatively, the Lycian system may have been based on an Aramaic model, with the
sign for 100 (1) being in fact multiplicative (1 100) rather than constituting a
single numeralsign (Frei 1976, 1977). No numbers higher than 120 are expressed
in any Lycian inscriptions, so we do not know whether, for instance, the Lycian
numeralphrase for 200 was additive (1\ 1) or multiplicative (11). Because none
of the Semitic systems had separate signs for 50, but all of them had signs for 20,
we would need to modify the value of the Lycian to 20, which is consistent with
the numeralphrases known from inscriptions. There is little similarity, however,
between the numeralsigns of Lycian and either Phoenician or Aramaic, and the
Lycian system uses a sign for 5, which is very rare in Aramaic. As well, Lycian,
like the acrophonic numerals, is written from left to right, whereas the Levantine
systems are all written from right to left.
The Lycian numerical notation system apparently did not diffuse outside Lycia.
Although Shafer (1950) argues that the similarities between the Roman and Lycian
numerals are sufficient to indicate the derivation of the former from the latter, this
likeness is no greater than that between Lycian and the Greek acrophonic system.
Furthermore, while there is some similarity between the Lycian numeralsigns and
other systems of the Italic phylogeny (especially South Arabian), these similarities do not correspond to any plausible circumstances of cultural contact. In Asia
Minor, scripts such as Phrygian and Lydian, both of which are closely related to
Lycian and were used in the fifth and fourth centuries bc, used numerical notation
based on the PhoenicianAramaic model rather than on the Greek or Lycian.
Lycian numerals are found primarily in a single context sepulchral epitaphs
indicating monetary amounts, normally including a numeralphrase preceded
by the word ada, now considered to be a monetary unit (Bryce 1976: 175). The
Italic Systems
107
monetary values may have stipulated a penalty to be paid should the tomb in
question be violated (Shafer 1950), or they may indicate fees paid in advance by
the family for a tomb site (Bryce 1976). The only nonfunereal context in which
Lycian numerals are used is the trilingual inscription found at Letoon, a public
legal regulation (Frei 1976). As Lycian numerals are not found on coins or on
financial inscriptions, they are quite distinct from the numerals of the rest of Asia
Minor and the Aegean.
As the Lycians became increasingly caught up in imperial conflicts between the
Persians and the Greeks, their script was used increasingly infrequently. By 300 bc,
the Lycian script had assimilated to the Greek, and its numerical notation ceased
to be used, replaced by the Greek alphabetic numerals.
South Arabian
The Old South Arabian scripts are of a very ancient origin, first appearing around
the turn of the first millennium bc in the southern part of the Arabian peninsula
(modern Yemen). They are consonantal and are characterized by large, wellformed
letters and by their extremely varied direction of writing (lefttoright, righttoleft,
or boustrophedon, depending on the inscription). They were used to write South
Semitic languages such as Minaean, Sabaean, Qatabanian, and Hadramauti. During their early history, these scripts did not possess any numerical notation system,
but following the rise of the kingdoms of Minaea and Saba in the fifth century bc,
numerical notation began to be used in monumental inscriptions. The numeralsigns used are shown in Table 4.9, including both lefttoright and righttoleft sign
forms, where appropriate (Halvy 1873: 511512; Hommel 1893: 8).
The system is cumulativeadditive, with a base of 10 and a subbase of 5, and is
written in whichever script direction is used in the inscription as a whole. The sign
for 1 is, as in all Italic systems, purely iconic. The signs for 5, 10, 100, and 1000,
however, are acrophonic; each is simply the first letter of the appropriate South
Arabian lexical numeral (Beeston 1984: 8). The sign for 50 is nonacrophonic, but
is simply a halved version of the sign for 100. In one inscription (Biella 1982: 531),
the sign X is apparently used with the numerical value 4, possibly in imitation
of the Nabataean system (Chapter 3). There are no signs for 500 or 5000 known
from any South Arabian inscriptions; these numbers were written with five signs
for 100 and 1000, respectively (Biella 1982: 1, 265). Normally, numeralphrases
were placed between large hatched bars to avoid confusing numeralphrases with
words, given the use of the acrophonic principle (Halvy 1875: 78). Large sets of
unitsigns were not divided into smaller groups, which creates a legibility issue
because the subbase of 5 is not used throughout the system; one inscription from
Sirwah denotes 12,000 using twelve signs for 1000 (Ifrah 1998: 187).
Numerical Notation
108
10
50
1
!
@
#
Right to left 1
Left to right
100
1000
Italic Systems
109
derived from a South Arabian model, used numerals based on the Greek alphabetic system (Chapter 5).
Although some South Arabian cursive inscriptions on wood have been found,
these contain no numerals. The system just described is documented only in monumental contexts. The functions of the numerals included recording details of
sacrifices or offerings to gods, quantities of booty obtained, numbers of military
troops, and information on construction projects such as monuments and irrigation systems. The South Arabians did not use an enumerated dating system, nor
do South Arabian coins contain numerical notation of any kind.
By the second century bc, instances of the South Arabian numerals were normally preceded by the appropriate lexical numeral written out in full. While this
aids modern scholars in their interpretation, doing so also removed any incentive
to continue to use the system. By the first century bc, although the South Arabian
scripts continued to be used, the numerical notation system had become extinct,
and it was not replaced until the seventh century ad, when the Islamic conquest
brought alphabetic and later positional numerals to southern Arabia.
Roman
Despite the importance and continued use of Roman numerals, the early history
of the system was very poorly understood until Keysers (1988) study. There are
several different classical variants of the Roman numerals, while the sign forms
and structure of the Roman numerals used today are medieval in origin. The
Roman alphabet was developed on an Etruscan model around 600 bc at a time
when much of Italy was under Etruscan political domination; it was written from
left to right, as it is today. Like its Etruscan precursor, the Roman numerical notation system has a base of 10, with a subbase of 5, and is essentially cumulativeadditive; unlike it, the Roman numerals are written from left to right and are
sometimes used subtractively. The great variety of numeralsigns used throughout
two millennia of its history contrasts strongly with the highly static quality of the
equally longlived Babylonian cuneiform and Egyptian hieroglyphic numerals.
Table 4.10 presents the numeralsigns used during the republican period (Ifrah
1985: 132; Keyser 1988).
These numerals were cumulativeadditive in structure in most inscriptions.
Only the signs for 1 and 10 remain unchanged throughout the entire history of
the system. The inverted V sign for 5, Q, is found only in early contexts, and
is evidence of the systems indebtedness to its Etruscan ancestor, but P is used
exclusively thereafter. The signs for 50 in Table 4.10 are roughly in chronological
order; the inverted arrow forms are earliest, with the inverted T forms prevalent
until about ad 25, and L most common thereafter (Gordon and Gordon 1957: 181).
Numerical Notation
110
10
50
100
Q
P\
\ V
U
T
\
S
500
X
W
,
Y\\
_
\

19,494 = YYYYUUUUTRRRRaaaa
19,494 = YUXRUaP
.\
~\
There is no sign for 100 in any early Roman inscription although there surely must
have been one to complete the series. The V form for 100 is extremely rare; Ifrah
(1998: 188) lists only a single inscription where it is found. Keyser indicates that
the first Roman C = 100 whose date is secure is from 186 bc, but he postulates
a thirdcentury bc origin for the symbol as a reduction of the Etruscan ;, even
though there is no example of any sign for 100 at this early date (Keyser 1988:
542). The number 500 is expressed using X in all early contexts, with assimilation
to the alphabetic D occurring around the transition to empire. The familiar M =
1000 used from the Middle Ages to the present occurs only in one classical Latin
inscription as part of a numeralphrase, although it is also found in various places
where M is simply an abbreviation for mille or milia (Gordon and Gordon 1957:
181182; Gordon 1983: 45). The signs for 5000, 10,000, 50,000, and 100,000 are
rarely encountered, though they are all attested as early as the third century bc.
Adding arcs on either side of the most common sign for 1000, Y, indicates successive powers of 10, while the right half of the appropriate base10 sign represents
the quinary component.
In a very limited set of texts, is used to represent 500,000 (Mommsen 1965
[1909]: 788791; Gordon 1983: 45). The sign is probably derived from alphabetic
Q and is thus an abbreviation of quingenta milia. Its use was limited to the later
Republic, and it was certainly not familiar a century later to Pliny the Elder, who,
in his Natural History, wrote Non erat apud antiquos numerus ultra centum
milia or Among the ancients there was no numeral larger than 100,000 (Natural History 33.47.133). Alongside the Roman numerals, the Romans had a duodecimal fractional system based on the as of twelve unciae (Menninger 1969: 158162;
Cagnat 1964: 33).
Italic Systems
111
There was very little regional variation in the signs or the structure of the classical Roman numerals (due, no doubt, to the centralization of the empire). In
North African inscriptions, however, the signs for the subbase were sometimes
not used, for example, IIIII for 5 and XXXXXX for 60 (Cagnat 1964: 3031).
Similarly, the signs for C and L were often written cursively, even in inscriptions
on stone, to distinguish them from the corresponding letters (Salama 1999). This
practice of cursively writing Roman numerals on stone was also attested in Spanish inscriptions of the fifth and sixth centuries ad, borrowing the numeral forms
used on much earlier Roman paypri written in Egypt (Mallon 1948).
Starting in the late republican period, the subtractive principle was occasionally
used for writing multiples of 4 or 9 (or rarely 8) of powers of 10. As would later
become the rule with modern Roman numerals, placing a lowervalued power of
10 to the left of a higher numeralsign indicated subtraction of the former from
the latter (IX for 9, XL for 40, but never using subbase signs as the subtrahend
i.e., VC is unacceptable for 95). This reduced the length of numeralphrases four
or five numeralsigns were replaced by two. Similarly, in the Augustan period
and into the early empire, the use of the subtractive XIIX and XXIIX for 18 and
28 were common, and IIX and XXC for 8 and 80 are also attested (Gordon and
Gordon 1957: 176181). Addition was used almost exclusively in the early republican period, and is the more usual form even in later classical inscriptions (Sandys
1919: 5556).4 Subtractive numerals are more common where a numeral is at the
end of a line of an inscription, allowing the engraver to avoid crowding many
numeralsigns into a limited space (Cajori 1928: 31). They may be more common
in informal texts than in formal inscriptions (Gordon and Gordon 1957: 180181;
Cagnat 1964: 3031). Despite Guitels (1975: 202203) denigration of subtractive
notation because it lacks the simplicity of the pure additive principle, it is a very
economical way of structuring numeralsigns. The Latin lexical numerals use the
subtractive principle for 18 and 19 (duodeviginti, undeviginti), perhaps explaining
the origin of this practice. Note, however, that while duodeviginti and undeviginti
are subtractive, novem (VIIII/IX), quatuordecim (XIIII/XIV), nonaginta et novem
(LXXXXVIIII/XCIX), and all other Latin lexical numerals are not.
Around the same time, the multiplicative principle began to be employed to
write very large numbers. Even as early as the third century bc, the Roman republic had become a large centralized state, and the need to express large numbers was
acute, yet the highest numeralsign was 100,000. At times, this led to extremely
cumbersome numeralphrases, such as the inscription on the famed Columna
4
112
Numerical Notation
rostrata of the consul Gaius Duilius, originally erected in Rome in 260 bc, which
celebrated a naval victory over Carthage in which over two million aes worth of
loot was plundered. The column is inscribed with at least twentytwo signs for
100,000, and possibly as many as thirtytwo, as the inscription is fragmentary
(Menninger 1969: 4344). Although it was recut, probably in the Augustan period
when other means of expressing this number were available, the original structure
of the expression was retained (Stenhouse 2005: 5960). One is struck, in looking
at this inscription, at the sheer enormity of the numeralphrase, and thus by the
impressive amount of booty obtained, and in fact this may have had something to
do with the reason it was written at all. Gordon and Gordon (1957: 180181) suggest
that one of the factors working against the widespread acceptance of subtractive
notation was the desire of public officials to impress and indulge. This conspicuous consumption of numerals is crossculturally frequent, for example, in the
Narmer macehead (Chapter 2) and other royal commemorative inscriptions.
Starting in the late republican period, a horizontal bar (vinculum or virgula)
placed above a numeralphrase or some portion thereof indicated that the number
under the bar should be multiplied by 1000 (Smith 1926: 7678; Gordon and
Gordon 1957; Cagnat 1964: 3132; Gordon 1983). This principle was first used in
the Lex de Gallia Cisalpina written between 49 and 42 bc (Gordon 1983: 47). For
most numbers, doing so did not improve conciseness; both gedddda\SROOO and
....YSROOO for 191,063 require twelve symbols. The main advantage is
that one need no longer remember so many numeralsigns or invent new ones; the
signs for 1, 5, 10, 50, and 100 are sufficient to express any number up to 500,000,
whereas eleven different signs would be needed under the purely additive system.
One slight complexity is that starting around the same time, barred numerals
were often used to distinguish ordinal from cardinal numbers, as in abbreviations
such as aaVIR for duumvir (Gordon and Gordon 1957: 166176). In one irregular
inscription from Pompeii from the first century ad, the numeralphrase LXXXX
apparently indicated 90,000 using a regular sign for 1000 multiplicatively, a practice not encountered again until the Middle Ages (Smith 1926: 7).
Starting around the early Imperial period, three vertical bars enclosing a
numeralphrase on the top and sides signified multiplication by 100,000. Thus,
instead of the thirtytwo signs for 100,000 found on the Columna rostrata, one
would need only to write mdddaan. This technique was first used in the late first
century bc, according to Gordon (1983: 47), but was employed rather sparsely
until the second century ad.5 In this way, any number less than 500 million could
5
According to Suetonius (Galba, 5), the emperor Tiberius willfully read the will of Livia
in such a way as to read a threebarred numeral min as i in order to reduce by a factor
of 100 the amount of inheritance owed to the future emperor, Galba (Cagnat 1964: 32).
Italic Systems
113
Regular signs
Multiplicative
(1000)
10
50
100
500
1000
1000
5000
10,000
50,000
100,000
500,000
1,000,000
Multiplicative
(100,000)
man
mbn mdn
men
35,863,120 = mgggebaaan edaaa URR
mgn
mjn
mkn
be expressed with just the lowest numeralsigns plus two types of bar to express
multiplication. This revised system is still a decimal system with a subbase of 5;
however, instead of being purely cumulativeadditive, it is a hybrid system using
cumulativeadditive structuring for numbers up to 1000 and multiplicativeadditive thereafter. The entire system (up to 100,000,000) as used in the Imperial
period is shown in Table 4.11.
Gordon (1983: 47) claims that the largest number expressed using this hybrid
cumulative and multiplicative system is 35,863,120, though an inscription at Ostia
from ad 36 apparently indicates 100 million as mkn (Menninger 1969: 245). Most of
the higher signs are attested only rarely. Curiously, Guitel (1975: 215) regards Roman
multiplicative notation as an evolutionary dead end, because, she argues, they no
longer needed to develop a more efficient positional system. The teleology of this
argument is immediately apparent, as it regards this development only with respect
to its failure to lead to a superior system. The Romans themselves likely perceived
it as a means of improving conciseness, while reducing the number of signs one
needed to memorize. Although the use of the 1000 bar continued among some
postRoman scribes, the use of the 100,000 box did not outlast the empire.
While the origin of the Roman numerals is a common topic of inquiry, unfortunately, as Cajori (1928: 31) noted, the imagination of historians has been unusually active in this field. Fortunately, Keysers (1988) panoptic essay on the origin
of the Roman numerals, which examines a variety of theories, ranging from the
sixthcentury theories of the grammarian Priscian to the twentiethcentury theories of modern classicists, has firmly settled the issue. The popular belief that the
Roman numerals originated as alphabetic signs is false. While the modern Roman
numeralsigns are also letters, the signs I, V, and X mean 1, 5, and 10, rather than
U, Q, and D, for unus, quinque, and decem. C is the first letter of centum, but this
is coincidental, since C is a reduction of the older Etruscan sign ; or VaU (Keyser
Numerical Notation
114
Etruscan
Roman
10
50
100
1
1
Q
Q\P\
R
R
\:
U\\;
\\T U
500
1000
\
\X
\\\/
Y\\\
1988: 542). The signs for 50 and 500 were not associated with the letters L and D
until the late Republic, and M was not used for 1000 until the Middle Ages. It
is fortuitous that the older sign for 1000 (Y) could be easily transformed into an
M. Similarly, theories employing the pictographic principle (for instance, I = 1
from a single finger; V = 5 from an outstretched hand, and X = 10 from two hands
together) were proffered by many early modern antiquarians, and later by classicists such as Sandys (1919), but, while they are imaginative, there is no evidence
to support them. Until recently, the most widely accepted theory was that of
Theodor Mommsen (1965 [1909]), who argued that the signs for 50 (), 100 (U),
and 1000 (Y) were taken from letters of the Chalcidic Greek alphabet that were
not needed to transliterate the Latin language: chi, theta, and phi, respectively,
which somewhat resemble the numeralsigns. Unfortunately, this attractive theory
has several flaws: the sign for 100 does not really resemble the Chalcidic theta;
these lost letters were sometimes used in Etruscan and Roman inscriptions; and
special pleading is required to derive the origin of X = 500. While it was once
plausible, it is no longer a parsimonious theory.
In fact, the Roman numerals up to 1000 developed through direct diffusion
from the Etruscans. The astonishing similarity between the Etruscan and archaic
Roman numeralsigns, as shown in Table 4.12, ought to be enough to prove a relationship between the two systems. The Etruscan numerals have temporal priority
over the Roman numerals, which do not appear until well into the fifth century
bc, and are not frequently encountered until the third century bc. In fact, the
similarities between the two systems are so great that one could treat them as a
single numerical notation system; they are identically structured, and many of
their numeralsigns are similar or identical. I treat them separately because the two
systems are written in opposite directions and because the Roman system used
signs for much larger powers at an early date.
The Roman numerals were used in a broader range of contexts than any other
cumulativeadditive system. In its earliest forms, they were used on coins, on pottery, and on inscriptions on stone. Dates, monetary values, and measures were
all frequently expressed in the Roman numerals. The Roman numerals could be
employed to express both cardinal and ordinal values. Their use in administration
and literature was widespread from the republican period onward. In texts, Roman
numerals were used to enumerate page and line numbers. Accounts, inventories,
Italic Systems
115
and legal documents also occasionally provide us evidence of their use in commercial and institutional contexts.
While Roman numerals certainly were used in the contexts of arithmetic and
calculation, there is minimal evidence that they were ever used for calculation.
Glautier (1972) discusses the Roman account records, which he characterizes as
primitive from an accounting standpoint because of the lack of positionality. A
similar point is raised by Meuret (1996) in his discussion of the Lamasba tablet, an irrigation regulation from North Africa during the reign of Elagabalus
(218222 ad), which contains a multiplication table to enable quick calculation
of water supplies, thus overcoming the computational deficiencies of the system.
Maher and Makowski (2001) demonstrate persuasively, however, that the assumption that Roman mathematics was poor because of Roman numerals cannot be
correct, given the complexity of the arithmetical calculations attested in Latin
texts. Nevertheless, their stronger claim that Roman mathematicians actually
used written numerals to do arithmetic, in particular for calculations involving
fractions remains unproven.
Roman numerals were used for computational functions if not directly for
computation, then certainly to mark the rows on the abacus. While few Roman
abaci survive, Taisbak claims that the Romans did all their calculations with them,
and even that the notation of Roman numerals originates from the abacus reckoning (Taisbak 1965: 158). This finding is contradicted by the derivation of the
Roman numerals from the Etruscan system and ultimately from an older tallying system. Because cumulative systems use onetoone correspondence intraexponentially, just as one counter equals one multiple of a power on the abacus,
the Roman techniques of numeration (Roman numerals) and computation (the
abacus) complement one another. This correlation is confirmed by the quinary
(base5) component of the abacus (there are rows not only for the powers of 10
but also for their halvings). No row on the abacus ever would have contained more
than four counters, which would have facilitated reading and working with them.
However, because the original Etruscan system probably emerged from a system
of tallying, it is more likely that the structure of the abacus emerged out of the
structure of the numerals than vice versa.
Despite the enormous influence of Roman civilization on Europe, North Africa,
and the Middle East, and despite the extraordinary chronological duration of the
Roman numerals (almost 2,500 years), they produced relatively few descendants.
While other systems in use over similar periods, such as the Brhm numerals and
Greek alphabetic numerals, changed their form greatly as they spread across time
and space, the Roman numerals of antiquity spread largely unmodified throughout Western Europe and other areas where the Roman alphabet was used. While
the Indian and Greek systems spread throughout many different scripts, changing
116
Numerical Notation
the forms of signs as they were transmitted, Roman numerals were infrequently
adopted by users of other scripts.
Of the few descendants of the Roman numerals, I have already discussed the
hybrid multiplicativeadditive system used occasionally from the first century
bc onward. In the medieval period, Roman numerals were essentially the same
as classical ones, though with slight differences in form and structure. In Arabinfluenced Spain, certain variant Roman numeral systems were used starting in
the tenth century ad. Around the same time in northern Europe, certain types
of medieval calendars contained unusual Roman numerals. Finally, as the Roman
numerals came increasingly under assault from the rival Western system, certain
positional variants of the numerals were occasionally used, combining features of
both systems.
Italic Systems
117
118
Numerical Notation
mildly (or even greatly) useful for a function for which they were never known
to have been used. The issue is not simply a mathematical game to see whether a
system can serve some arithmetical function. By understanding the functions for
which the Roman numerals were not used, we may better understand the circumstances of their eventual replacement.
Medieval Additive
As early as the second century ad, but most prominently in the fifth and sixth centuries, Roman numerals inscribed on stone inscriptions frequently took on a cursive
quality, and some signs became ligatured together (Gordon 1983: 46). A special sign
for 6, , which is nothing more than a cursively written and ligatured vi or ui, occurs
on many inscriptions from late antiquity (Lassre 2005: 5759). On sixthcentury
Byzantine imperial coins, was as common as VI (Wroth 1966: cx). The use of
this sign probably ceased in the eighth century (Bischoff 1990: 176). Similarly, in
late antique and early medieval inscriptions, 40, 60, and 500 could be written cursively, even in inscriptions on stone. The Visigothic script used between the seventh
and thirteenth centuries was particularly characterized by these and other ligatured
Roman numerals, particularly for 40 (XL) (Schapiro 1942, Bischoff 1990). These
sporadic shifts toward ligatured notation would later play some role in the paleographic modification of Arabic signs into what would become the Western numerals
(Lemay 1982: 393; see also Chapter 6, of this volume). Yet these ligatures did not
represent an inexorable diachronic trend toward ciphering in Roman numerals, and
the majority of early medieval Roman numerals remained cumulativeadditive.
After the fall of the western Roman Empire, literate knowledge was distributed
rather sparsely for instance, among Western monks, Byzantine bureaucrats, and
Middle Eastern scholars. The great early medieval Mediterranean polities the
Byzantine Empire and the early Muslim caliphates mainly used cipheredadditive numerical notation systems such as the Greek and Arabic alphabetic
numerals (Chapter 5). In Western Europe during the Middle Ages, the Roman
numerals were the only ones in common use. Knowledge of other systems was
restricted to peripheral regions such as Spain and southern Italy, and to a tiny
welleducated elite. Even at the height of the Carolingian Renaissance (around
800 ad), arithmetic was the province of a learned few, and was acquired late in the
scholars education (Murray 1978). Still, while the need for largescale bureaucracy and the corresponding need to express large numbers had declined since the
height of the Roman Empire, Roman numerals were still frequently encountered,
and even expanded in the range of functions they served.
The medieval Roman numeralsigns differ from the classical ones in several
aspects. Rather than being written solely in majuscule characters, Roman numerals
Italic Systems
119
120
Numerical Notation
The range of functions for which Roman numerals were used expanded considerably in the Middle Ages. Astronomical texts, which in antiquity were almost
exclusively written using Greek numerals, often employed Roman numerals in the
medieval era. As mentioned earlier, the alphabetizing of the signs for 500 and 1000
to D and M allowed the creation of numberriddles such as chronograms in which
the numerical value of the Roman numerals in a phrase expressed the date of an
event. Evidence for their use in legal documents and account records increases
greatly, though this may be a function of the differential survival of perishable
materials from later periods.
Medieval Modifications
The replacement of the Roman numerals by the Western system was neither easy
nor uncontested. Instead, between the twelfth and seventeenth centuries, there
was great controversy throughout Western Europe regarding which system to
use, with cultural, sociopolitical, and practical considerations being invoked in
favor of one system or the other. Nor was the situation simply a choice between
two options. While the general trend was toward the Western system in the great
majority of contexts, as knowledge of the cipheredpositional system spread into
Western Europe, a few individual writers made idiosyncratic modifications to the
Roman numerals in response to the interloping newcomer. While none of these
modifications was adopted on a wider scale, they can enlighten us about the circumstances under which the Roman numerals were replaced.
The most complete positionalized version of the Roman numerals is one of the
earliest. Around 1130, the mathematician H. Ocreatus, a student of Adelard of
Bath, invented a positional numerical system using the Roman numeralphrases
for 1 through 9 (I, II, ... IX) along with a special sign (O or t, called cifra) to
indicate an empty position (Smith and Karpinski 1911: 55; Burnett 1996, 2002c,
2006). This system is attested only in one thirteenthcentury manuscript, a collection of arithmetical texts (Cashel, G.P.A. Bolton Library, Medieval MS 1), of
which Ocreatuss Helcep Sarracenicum (Saracen Calculation) is only one part. Positions were separated using a dot to avoid confusion; thus, 1089 was expressed
as I.O.VIII.IX.6 This system, which blends the Roman cumulativeadditive and
Arabic cipheredpositional systems, is cumulativepositional and base10 with a
subbase of 5. While Murray (1978: 167) characterizes Ocreatuss system as clumsy,
it should be noted that it is fully positional, far more so than later compromises
6
Smith (1926: 72) argues that this practice was attested in a few classical inscriptions as
well, such as XII.L.D for 1,250,500 (!), but I know of no inscriptions where this was
actually done.
Italic Systems
121
NumeralPhrase
Roman with
Multiplication
Western Source
122025
IIDCCCXIIII
IX.XX.XVI
aaWUUUROP
URUPO
ZUUSPOOO
ZZSROOO
SRRRPOOO
2814
1231
196
Smith 1926: 6
1258
2073
88
aaaUSRRROOO
3183
Cajori 1928: 33
USOR
ZWP
UU
UWSR
ZWROP
ggg
baWUUROOO
159
Preston 1994
1505
200
460
Cajori 1928: 34
1514
Smith 1926: 7
300000
6713
gddaaaUWSPO
123456
Cajori 1928: 33
1258
Mij lviii
1340
IILXXIII
1388
143738
IIIIxx et huit
M XX
III C IIII III
VIIXXXIX
1505
IVcV
1392
1514
II
IIIIC.LX
1514
XVCXIV
1550
CCCM
1554
1771
vi vii xiii
c m
c
i xxiij iiij lvj
made between the Roman and Arabic systems. Nevertheless, this system was not
taken up more widely in medieval scribal or mathematical texts. The speculation of Busch (2004) that a difficulty with seafaring travel times in the fifteenthcentury Icelandic Landnmabk can be resolved by interpreting VI (6) as V I
(e.g., 51) is highly dubious.
While the use of barred numerals to indicate multiplication by 1000 was of
early origin and continued throughout the Middle Ages, new multiplicative forms
began to be used starting in the twelfth and thirteenth centuries. Some examples
of such numeralphrases are listed in Table 4.13, along with the transcriptions of
the appropriate number both in the classical Roman system (including the use
of subtractive and multiplicative forms, where appropriate) and in the Western
system.
These numeralphrases primarily express multiplication by 100 or 1000 by juxtaposing C or M either immediately beside the appropriate multiplier, above it,
or in superscript, and sometimes interposed with a dot. Often, the number 20
occupies a special role as a multiplier; such phrases are almost all from France
and result from assimilation to the partly vigesimal structure of the French lexical
numerals and/or to the monetary system (Preston 1994). These numeralphrases
have a multiplicative component, but they are not positional the value of the
122
Numerical Notation
1482
202
1200
1504
1515
1624
1447
1502
numeralsigns does not change due to their position, but rather due only to their
juxtaposition with another sign. It might be thought that multiplicative forms
were adopted in order to write numerals more concisely or with a smaller set of
numeralsigns as was the case with the initial use of multiplicative forms in classical Rome. However, as seen by comparing the numeralphrases in Table 4.7 to
their equivalents in standard Roman numerals, there is no such benefit.
In addition to these multiplicative forms, we find many cases where the signs of
the Roman system were intermingled with the positional principle of the newer
Western system, as well as its actual numeralsigns. The earliest known example is from the late twelfthcentury Microcosmographia, in which the only known
copy (from the thirteenth or fourteenth century) is dated mclxxviii (1178) but in
another place as mc87 (an error for 1178) (Williams 1934: 107). Preston (1994)
describes a number of such mixed uses in Durandus of SaintPourains Commentary on the Sentences of Peter Lombard of 1336, such as xxx3 for 33 and xl7 for 47. In
one example from an English almanac from 1386, Western numerals both precede
and follow Roman numerals in a numeral phrase for 52,220, written as 52mcc20
(Halliwell 1839: 116). Menninger (1969: 287288) provides examples of such
admixtures starting in the late fifteenth century, as indicated in Table 4.14.
In these partly cipheredpositional numeralphrases, conciseness is greatly
increased over their Roman counterparts. It is unknown whether this was being
done consciously as a compromise between the two systems, as a misunderstanding of the Western system, or as numerical playfulness. In at least one case, the
blending is obviously erroneous; in an astronomical table, the scribe wrote MCC6
for 1269, then crossed out the entire phrase and wrote 1269 in Western numerals
(Steele 1922: xvii). Such combinations are not necessarily advantageous; the idiosyncratic nature of these numeral formations almost certainly decreased their
comprehensibility. None were used frequently or consistently enough to create a
true variant system. These hybrid formations no longer appear after about 1650.
The roughly contemporaneous expansion of Roman multiplicative notation
and the introduction of hybrid forms employing ciphering and positionality in
Italic Systems
123
late medieval European scribal traditions meant that any number could be written in several ways. The wide range of textual genres containing such variants,
and their geographical and chronological breadth, confirm that this variety was
not isolated, but neither was it standardized. In any event, a parallel process was
ongoing from the twelfth century onward that would eventually lead to the nearly
complete replacement of Roman numerals by Western ones throughout Europe
and eventually throughout the world.
This statement must be considered in light of the fact that a) it is a sample sentence
rather than part of an evaluation of the two methods and b) Palsgraves book was paginated throughout using Roman numerals.
124
Numerical Notation
inventories not by ciphers but by plain letters (Berggren 2002: 361). As late as
1494, the Frankfurt Brgermeisterbuch (a mayoral book of edicts) ordered reckoners not to compute using Western numerals (Menninger 1969: 426427). One can
certainly imagine the consternation of merchants and bookkeepers upon learning
that one can simply add zeroes to the end of a numeral ad infinitum to multiply
its value by ten each time, but the role of xenophobia and traditionalism in all
of these cases should not be underestimated. The Western numerals may have
been rejected in part because of practical considerations, but they were, after all,
a foreign and newfangled invention that involved the unusual principle of positionality and that (because of their novelty) could be used to conceal information
or deceive others.
The rarity of paper in the earlier Middle Ages may also have contributed to the
continued use of the Roman numerals (Smith and Ginsburg 1937: 29). Until penandpaper calculation became feasible, the Roman numeralfriendly abacus was
the computational technology of choice, whereas the switch to pen and paper
made the conciseness of Western numerals more attractive. This attractiveness
increased with the introduction of printing presses, where long strings of movabletype Roman numerals would have been unwieldy in comparison to Western
numerals. Yet despite Eisensteins contention that, [t]he use of Arabic numbers for pagination suggests how the most inconspicuous innovation could have
weighty consequences in this case, more accurate indexing, anotation, and crossreferencing resulted (1979: 1056), most incunabula contained no Western numerals. Even where the utility of Western numerals for computation was recognized,
the Roman numerals were not always abandoned entirely, as in Italian metrological documents studied by Travaini (1998), in which computations were undertaken using Western numerals, but totals were then written in Roman numerals.
William Cecil (Lord Burghley), Englands Lord High Treasurer from 1572 to 1598
and a chief advisor of Elizabeth I, would frequently transcribe economic documents from Western back into Roman numerals, and was decidedly uncomfortable with the new system (Stone 1949: 31).
Despite such resistance, the Western numerals had taken hold by around 1300
in the Italian citystates. Elsewhere in Western Europe, particularly in Germany
and England, the Roman numerals predominated until the late fifteenth century or even later. Jenkinson (1926) finds limited evidence for the use of Western
numerals in English archives before 1500, and notes that Roman numerals were
forbidden from use in state accounts only in the nineteenth century. Barradas
de Carvalhos (1957) study of fifteenth and sixteenthcentury Portuguese texts
showed that the Roman numerals were replaced there only around 1500. Several
important sixteenthcentury arithmetical texts continued to use Roman numerals in
preference to Western ones. Jacob Kbels (14701533) Rechenbiechlin, published in
Italic Systems
125
Figure 4.2. Tables of variant Roman numerals from the Mysticae numerorum significationis
(158384) of Petrus Bungus. Source: Smith 1908: 380383.
Augsburg in 1514, an important commercial arithmetic book intended for mercantile instruction, uses Roman numerals throughout and strongly advocates the use
of the counting board, although Kbel frequently used multiplicative phrasing for
hundreds and thousands, and his text is the source of the unorthodox 260/400
Roman numeral fraction in Table 4.13 (Smith 1908: 100107). However, by the
seventeenth century, the battle was essentially finished, and the Roman numerals ceased to be used in most contexts. Figure 4.2, from the Mysticae numerorum
significationis of Petrus Bungus (d. 1601), first published in 158384, depicts a wide
variety of archaic and curious Roman numeral phrases, but is essentially a text on
number mysticism and symbolism rather than a manual for practical use (Smith
1908: 380383).
Why, after over a millennium of essentially unchallenged use in Western Europe,
should the Roman numerals have ceased to be used for most purposes? The traditional answer given that Roman numerals were less efficient for computation
is correct but incomplete. The Roman numerals were used alongside the Greek
126
Numerical Notation
alphabetic numerals for well over a thousand years, despite the great increase in
conciseness that could have been achieved by abandoning the Roman system. In
medieval Western Europe, debates over the use of Roman numerals lasted into
the seventeenth century, with the relative usefulness of the Roman numerals for
calculation being greatly increased by the accompanying use of counting boards.
However, two developments were fundamental in rendering the Roman system obsolete in the West. Firstly, the development of the printing press and the
consequent rise in literacy after 1450 correlates very well with the rapid adoption
of Western numerals throughout Europe (particularly in northern regions). The
newly literate middle classes of Western Europe were learning to read and calculate, unconstrained by centuries of traditional use of the Roman numerals. At
the same time, the rise of mercantile capitalism in the Renaissance changed how
numbers were viewed and used (cf. Swetz 1987). The functional needs of Western
society to represent number changed dramatically between 1300 and 1700, and
the computational functions that are better served by Western than by Roman
numerals increased in importance. Nevertheless, Crosbys (1997: 41) insistence
that Roman numerals were adequate for the weekly market and for local tax collection, but not for anything grander is misplaced if intended as an explanation
for their replacement. In fact, as Crosby himself notes, Roman numerals persisted
for centuries in the record books of institutions as grand as the British Exchequer
(Crosby 1997: 115116). Doubleentry bookkeeping was facilitated to some degree
by the transition to Western numerals, but it was by no means a requirement.
The introduction of a more fully mercantile and urban economy in Western
Europe did, however, expose more people, particularly members of the newly wealthy
and increasingly literate middle class, to a new form of numeration. Rather than
examining this change in notation simply by comparing the two notations, it is more
productive to examine the question from a social perspective. If Roman numerals
were truly so inferior, how did they persist for two millennia without being replaced?
Why did it take six centuries even after the introduction of Western numerals for this
shift to take place? The most likely scenario is that the new users of writing and written numeration, unconstrained by past practices, provided a critical mass of users for
the Western numerals that facilitated certain mercantile functions.
Even today, the Roman numerals have not disappeared. The use of the Roman
system on clock faces, in the enumeration of kings and popes, on many dated
inscriptions, and to mark the copyright dates on films is rather archaic, but the
cultural importance of its symbolic connotations of antiquity, tradition, and prestige likely guarantee its future survival. In media where archaism is desired, Roman
numerals are often preferred; this is the case among the Cornish church sundials
from 16701850 studied by Burge (1994), for instance. An inscription on a gateway at Harvard uses archaic Roman numeralsigns, noting the colleges 264 years of
Italic Systems
127
existence in 1900 as ANN. DOM. _WUUUU. COLL. HARV. UUTROOOO (McPharlin 1942: 18). Pot (1999) notes the existence of several twentiethcentury postage
stamps in which additive Roman numerals rather than the ordinary subtractive ones
are employed for instance, XXXXI instead of XLI on a 1990 German issue. Presumably these variants are not errors per se, but are mobilized for aesthetic reasons,
just as on Roman numeral clock faces, 4 is normally written IIII instead of IV, even
though 9 is IX (Hering 1939). The fifteenthcentury accountant Luca Pacioli advocated the use of Roman numerals for the yeardates in accounting books for aesthetic
reasons (although Western everywhere else), recommending [u]se the ancient letters in making this entry, if only for the sake of beauty (Crosby 1997: 222).
At the other end of the class and prestige spectrum, Roman numerals were
often retained in peasant or rural traditions and trades long after they had
ceased to be used in formal texts and administration. Gmr (1917) describes
a wide variety of wooden tally sticks of Swiss provenience, dating from the
sixteenth to the nineteenth centuries, containing Roman numeralphrases. The
same principle applies to the much earlier calendarnumerals used on Northern European calendrical documents (see the following discussion). With
the inexorable spread of industrialization and capitalist commercial practices
throughout Europe, the use of wooden tallies containing Roman numerals has
essentially disappeared.
When used to paginate the introductions of books, or to enumerate items in
sublists or subsections, Roman numerals continue to serve the very practical function of providing an alternate form of enumeration wherever two different series
of things must be listed. I would not expect that any industrial society could function solely with a cumulativeadditive system such as the Roman numerals. Conversely, however, I would not expect them to be replaced entirely. The failure of
over five hundred years worth of predictions of the Roman numerals imminent
demise, and complaints regarding their utility, suggests that functional considerations, while important, do not tell the entire tale with respect to the history of
numerical notation systems.
ArabicoHispanic Variants
Arabic and Western European knowledge systems interacted intimately on the
Iberian peninsula between the tenth and sixteenth centuries the period of the
Reconquista and somewhat beyond. At this time, the Roman numerals were
well known in Spain and Portugal, as in the rest of Europe. Arab mathematicians and astronomers had, since around 800 ad, used a cipheredpositional
system much like the one used today (Chapter 6), along with older cipheredadditive abjad numerals related to those of the Greeks (Chapter 5). Medieval
128
Numerical Notation
Spain thus presents us with a remarkable case where three coexistent systems
each had a different structuring principle (cf. Labarta and Barcel 1988). But
rather than a simple case of the two Arabic systems replacing the Roman numerals, medieval Spanish astronomers, bookkeepers, and scribes experimented and
mixed principles, incorporating ciphering and positionality into the basic structure of the Roman numerals in innovative ways. It was, as described by Berggren (2002: 358), a promiscuous blend of systems, a mlange that persisted for
several hundred years.
In the tenth and eleventh centuries ad, Spanish astronomers wrote extensive
astronomical texts using primarily the Roman numerals. Even the use of the subtractive principle was very infrequent in Spain at this time. Despite the use of
Arabic positional numerals in the region for some time previous and familiarity
with works such as the Arithmetic of AlKhwarizimi, Spanish astronomers did not
adopt the system directly (Lemay 1977: 458). Instead, the Roman numerals were
modified to increase their conciseness. Instead of using the cumulative principle
to express the numbers 3 through 9 (III, IIII, V, etc.), they used new acrophonic
symbols based on the first letters of the Latin numeral words: 3 = t = tres; 4 = q =
quatuor; 5 = Q = quinque; 6 = s = sex; 7 = S = septem; 8 = o = octo; 9 = N = novem.
Of these, only the symbols for 4, 8, and 9 were common, probably because they
are the longest numeralphrases below 10 in Roman numerals (IIII, VIII, VIIII).
In addition, a special sign for 40 was employed: , a cursive ligatured version
of XL, and the only exception to the abandonment of subtractive forms (Lemay
1977: 459). Bischoff (1990: 176) reports that this sign for 40 is also found in some
Visigothic manuscripts, suggesting considerable antiquity for the sign on the Iberian peninsula. These changes altered the classical Roman numerals into a partly
cipheredadditive system. However, these modifications were not accompanied
by the adoption of positional notation, and the system still had a mixed base
of 5 and 10 and was purely additive. These developments increased the conciseness of numeralphrases considerably. The number 99, which would have been
LXXXXVIIII, could now be expressed as LN. This system was used in mathematical and astronomical texts until about the mid twelfth century, at which time
the Western numerals took hold in astronomy, as in the rest of Western Europe.
Yet while the Arabic positional numerals had firmly established themselves and
were later transformed into the Western numeralsigns familiar to us, the Roman
numerals did not cease to be used on the Iberian peninsula, though their use
became increasingly limited. Labarta and Barcel discuss two curious offshoot
Roman numeral systems found in early modern Spanish documents, as shown in
Table 4.15 (Labarta and Barcel 1988: 3234).
Both systems are cumulativeadditive, decimal systems with a subbase of 5,
unlike those of the earlier astronomical texts. Unlike most Roman numerals,
Italic Systems
129
10
50
100
numeralphrases are written from right to left (highest values at the right), which
is curious because even in the Arabic script, which has a righttoleft direction, the
numerical notation system has a lefttoright direction. The first system is found
in a few late sixteenth and early seventeenthcentury documents to indicate monetary quantities, and its numeralsigns are somewhat similar to letters of the Arabic script. The second system is found only in a single Inquisition document from
1576 (Labarta and Barcel 1988: 34).8
From the sixteenth century onward, Western numerals became increasingly
common throughout the Iberian peninsula, but certain anomalies remained in the
Roman numerals of the early modern period. Subtractive numerals remained rare
in comparison to the rest of Europe. A symbol known as the caldron was sometimes used; it meant 1000 but was used multiplicatively, preceded by unitsigns,
not additively (Cajori 1922, Lowe 1943). The caldron was most often shaped like
a U, sometimes with one or two diagonal bars across it (). Thus, a manuscript
from the Ponce de Leon papers from 1501 has the year dated as JDJ (1 1000 500
1), reminiscent of the use of M for that purpose in other parts of medieval Western
Europe (Lowe 1943: 9). Cajori (1922) hypothesizes that the caldron originated
in Italy in the fifteenth century, possibly as a modification of the older Y for
1000 that was not fully replaced by M until the modern period. Curiously, even
after the Roman numerals had been abandoned entirely, it continued to be used
multiplicatively with Western numerals, as in a contract written in Mexico City in
1649 in which a sum of 7291 pesos is expressed both as 7U291 and VIIUCCXCI
(Cajori 1922: 201).
Calendar Numerals
In the late Middle Ages, unusual numerals were used in some Northern European
documents and inscriptions pertaining to calendrical calculations. Known as calendar numerals, runic numerals, or peasant numerals, they are a variant of the
Roman numerals. I reject the term runic because few runic inscriptions contain
8
However, see Chapter 10 for a discussion of a potential descendant of this system in the
form of Berber numerals used in North Africa.
Numerical Notation
130
a b c d e f g
A B C D E F G
1 2 3 4 5 5 5
1 2
10 11 12 13 14 15 16 17 18 19 20
h
H
5
3
i j k l m
I J K L M
5 6 6 6 6
4
1 2 3
n o p q r
N O P Q R
6 7 7 7 7
4
1 2 3
s t
S 8
7 8
4
them.9 Likewise, the term peasant numerals tells us who may (or may not) have
been using them, but lacks the precise functional association of calendar numerals. Some examples of these signs are shown in Table 4.16 (cf. Kroman 1974: 121;
Ifrah 1985: 146147).
These systems are all cumulativeadditive and have a base of 10 with a subbase
of 5. Units are marked by strokes or dots; fives are marked by angled or curved
lines or loops to create U or V shapes, and tens are marked by transecting
the vertical line perpendicularly, creating a cross or X. Although these numeralsigns are often joined together into single figures resembling digits by using a
vertical stroke, the system is not ciphered. There are no signs for 50, 100, or
higher values, and I am not aware of calendar numerals being used for numbers
higher than 31 (the number of days in the longest months). Other than their
unusual numeralsigns, the calendar numerals are identical to ordinary Roman
numerals. While they were used in rural traditions that suggest to some that
they were part of the large cultural substratum of tallying and notching in medieval and early modern Europe (Menninger 1969: 249251; Ifrah 1985: 146147),
the calendar numerals are not tallies in this sense. A tally is used ordinally
one places marks as necessary on some material (wood, paper, stone, etc.) in
sequence in order to keep a running total, rather than, as with numerical notation systems, marking an already totaled value. The calendar numerals are simply a variant of Roman numerals in which numeralphrases are written vertically
and attached to a line.
By the late fifteenth century, the Western numerals were fairly well known
in Germany and were becoming much more common in England and Scandinavia. Calendar numerals were used in a very delimited set of contexts, namely
The medieval antiquarian tradition of manuscripts discussing runic writing contains minimal evidence of any specifically Runic numerical system. The clophruna (Old Norse
klapprnir) system of notating numbers by dots (a = 1 = .; b = 2 = ..; c = 3 = ) discussed
by Derolez (1954: 134) is of obscure origin and is not clearly related to any other numerical notation.
Italic Systems
131
132
Numerical Notation
Summary
The Italic numerical notation systems developed in the early sixth century bc with
the invention of the Etruscan and Greek acrophonic numerals, based on a previously existing but still poorly understood system of tallies, and spread eastward
from Italy during the period of classical Greek preeminence in the Mediterranean.
However, of all the Italic systems, only the Roman numerals had any extensive use
in the Christian era, as the Greeks had switched to the cipheredadditive alphabetic
numerals (Chapter 5) by the Hellenistic period. The remarkable persistence of the
Roman system and the swift decline of other systems are best explained by the
changing political fortunes of their users. The use of the earlier Italic systems was
mainly limited to inscriptions and commercial marks, though the Roman numerals
were later used for an enormous variety of functions in different social contexts.
All the Italic numerical notation systems are cumulativeadditive and decimal,
with a subbase of 5, and virtually all use a single vertical stroke for the units. Some
unusual structural features occasionally emerge in the higher powers of 10, such
as the use of implied multiplication in South Arabian and the hybrid multiplicative structure of later Roman numerals. Although the epichoric numerals of Argos,
Nemea, and Epidaurus use dots rather than vertical strokes for units, they are related
to the Greek acrophonic numerals and must be considered part of this phylogeny.
The cultural history of some Italic systems is intermingled with those of the
Hieroglyphic (Chapter 2) and Levantine (Chapter 3) phylogenies, making the
construction of accurate cultural phylogenies more difficult, particularly because
the systems of all three are cumulativeadditive, decimal, and used in the eastern Mediterranean. Often the three families can be distinguished on structural
grounds: the Hieroglyphic systems all lack a quinary component, while the Levantine systems all have special signs for 20 and are multiplicativeadditive above
100. This structural distinction can be confirmed independently by examining
known patterns of historical contact. Despite the jumbled state of our present
knowledge, we can identify phylogenies, not only because of similarities in their
systems structure, but also as a result of the attested cultural connections among
the societies in which they were developed. Nevertheless, to insist too strongly on
pure cultural phylogenies following a model of speciation would be erroneous.
In comparison to either the Alphabetic (Chapter 5) or South Asian (Chapter 6)
systems, however, the diversity of Italic systems is far less than might be expected,
given the preeminence and impact of the Roman Empire on European social life. The
Roman numerals gave rise to relatively few descendant systems when they were borrowed, they were copied rather than modified. We would do well to remember that
the Roman numerals are simply one numerical notation system among the worlds
diverse systems, despite their importance within the Western consciousness.
chapter 5
Alphabetic Systems
The systems discussed in the previous three chapters are primarily cumulative,
repeating signs within each power of the base to indicate addition. In contrast,
the next two families the Alphabetic and South Asian systems consist mainly
of ciphered systems, which use, at most, a single sign for any power to indicate
its different multiples: 1 through 9, 10 through 90, 100 through 900, and so on,
in the case of decimal systems. Ciphered numeralphrases are thus much shorter
than cumulative ones, but require their users to be familiar with many more signs.
Alphabetic numerical notation systems generally use phonetic scriptsigns, in a
specified order, to express numerical values, and thus mitigate the effort needed to
memorize both scriptsigns and numeralsigns. Despite the name, the scripts in
question are not always alphabets; some, such as the Hebrew and early Arabic, are
abjads or consonantaries, expressing primarily consonantal phonemes, and one,
the Ethiopic Geez script, is an alphasyllabary or abugida, expressing consonant +
vowel clusters.1
Alphabetic systems were used as far north as England, Germany, and Russia
and as far south as Ethiopia, and throughout Africa and the Middle East from
Morocco eastward to Iran. Their history spans over two thousand years, from the
development of the Greek numerals around 600 bc to the present, but in some
1
For more extensive discussion of script typology, which is not warranted here, see
Daniels and Bright (1996).
133
134
Numerical Notation
cases important historical questions remain unresolved. While they are mostly cipheredadditive, they are not structurally identical. We can learn much more from
these structural differences than from the paleographic curiosities of the signs of
various systems. I hope in this chapter to illuminate areas of study where our
knowledge is less than perfect in order to draw attention to the need for further
specialized research. Table 5.1 shows the most common variants of the major alphabetic systems.
Greek Alphabetic
In Chapter 4, I discussed the cumulativeadditive Greek acrophonic numerals,
which were given their name because the letters used are the first letters of the
appropriate Greek numeral words. Another Greek system, much more common, is sometimes called the Ionic or Milesian system due to its origin in
western Asia Minor, but is most commonly called the alphabetic system. While
the Greek alphabet was based on a Phoenician model, probably in the ninth or
eighth century bc, none of the earliest Greek inscriptions contains numerical
notation; thus, the debates on the time of the origin of the alphabet are not relevant to the origin of alphabetic numerals (cf. McCarter 1975, Swiggers 1996).
The first examples of the alphabetic numerals date to the sixth century bc and
are written using the letters of the archaic Greek script used in Ionia and the
Ionian cities of Caria, such as Miletus. Table 5.2 has representative examples of
these early signs.
The system was purely cipheredadditive and decimal, and was usually written from left to right, though righttoleft and boustrophedon (alternating direction) inscriptions are not unknown. The numeralsigns are archaic variants of the
twentyfour familiar Greek letters, plus three special signs called episemons: vau
or digamma (6), qoppa (90), and san or sampi (900), which were added to reach
a full complement of twentyseven signs for all the values from 1 through 900,
allowing one to write any natural number less than 1000.2 Vau and qoppa were
occasionally used phonetically in the Ionic script, with the values of [v] and [k],
while san appears to have been borrowed from Phoenician sd [ts], and was possibly used in archaic Greek with a similar phonetic value (Swiggers 1996: 265266).
There are very few examples of alphabetic numerals from this early period, and
all of them express values under 1000, so we do not know whether higher values
could be represented.
2
There was no general name for the three signs in classical Greek; the term episemon originally referred only to the sign for 6, but following its etymology, some classicists use the
term in a more generic sense (J. Kalvesmaki, personal communication; Foat 1905).
135
70
60
50
40
30
20
10
a
b
g
d
e
v
z\
h
q
i
k
l
m
n
x
o
Greek
Coptic
Ethiopic
.a.
.b.
.c.
.d.
.e.
.f.
.g.
.h.
.i.
.j.
.k.
.l.
.m.
.n.
.o.
.p.
Gothic Hebrew
g
h
i
j
k
l
m
n
o
p
Syriac
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
Arabic
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
Fez
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
a
b
c
d
e
f
g
h
i
j
k
l
m
n
o
p
(continued )
a
b
c
d
e
f
g
h
i
k
l
m
n
o
p
q
136
6000
5000
4000
3000
2000
1000
900
800
700
600
500
400
300
200
100
90
80
p
,
r
s
t
u
f
c
y
w
.
/a
/b
/g
/d
/e
/v
Greek
Coptic
Ethiopic
.q.
.r.
.s.
.t.
.u.
.v.
.w.
.x.
.y.
.z.
.{.
Gothic Hebrew
q
r
s
t
u
v
sv
tv
uv
vv
svv
/a
/b
/c
/d
/e
/f
Syriac
q
r
s
t
u
v
w
x
y
z
:
;
Arabic
q
r
s
t
u
v
w
x
y
z
,
Fez
q
r
s
t
u
v
w
x
y
z
!
@
#
$
%
^
&
q
r
s
t
u
v
w
x
y
z
1
2
3
4
5
6
7
q
r
s
t
u
v
w
x
y
z
/
1
2
q
r
s
t
u
v
w
x
y
z
{
r
s
t
u
x
y
z
137
1,000,000
100,000
90,000
80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000
9000
8000
7000
z
h
q
i
p
/z
/h
/q
a
b
g
d
e
V
Greek
Coptic
Ethiopic
Gothic Hebrew
f
g
h
i
j
s
/g
/h
/i
a
b
c
d
e
Syriac
Arabic
Fez
*
(
)
8
9
0

Numerical Notation
138
a
10s
j
100s
s
562 = wob
1s
b
k
t
c
l
u
d
m
v
e
n
w
f
o
x
g
p
y
h
q
z
i
r
,
In the classical and Hellenistic periods, the familiar Greek alphabet supplanted
the archaic regional (or epichoric) variants, and the alphabetic numerals developed along with them, retaining their order and numerical values but assuming
their modern (majuscule) forms. In addition, starting in the middle of the fifth
century bc, two new techniques were used to express higher values. For multiples
of 1000, a small slanting mark (known as a hasta) was placed to the left and below
a sign for 1 to 9 to indicate that its value should be multiplied by 1000; thus, G
means 3, but /G means 3000 (Threatte 1980: 115). Values above 10,000 are rarely
encountered except in mathematical works, and individual writers used different
methods to do so. The most common method, used by Aristarchus, involved placing a small alphabetic numeralphrase (less than 10,000) above a large M character (= myriades) to indicate multiplication by 10,000 (Heath 1921: 3941).3 Thus,
3,000,000 would be expressed with only two signs, as y. This allowed any number
less than 100 million to be easily expressed.4 The system thus appeared as shown
in Table 5.3, using the Athenian letters.
The system is thus cipheredadditive for values under 1000, and thereafter is
multiplicativeadditive at two different levels: firstly, through the use of a hasta
to indicate multiplication by 1000, and then through the use of an M to indicate
multiplication by 10,000. The alphabetic numerals were generally written in descending order, with the highest values on the left. Numbers between 11 and 19
were often written with the 10sign (I) following the unitsign, however, to correspond with the way in which the ancient Greek lexical numerals were formed:
3
Heath also discusses techniques such as that of Herons Geometrica, where two dots
placed over a sign indicate multiplication by 10,000; that of Apollonius, using tetrads,
turning the system into a mixed base10/10,000 cipheredadditive system; and that of
Nicholas Rhabdas, a fourteenthcentury scholar who used Herons technique, except
that additional pairs of dots above a number indicated successive powers of 10,000.
None of these systems was ever widely used.
For the ancient Greeks, ten thousand times ten thousand 100 million, the first uncountable number in the alphabetic system was of symbolic significance, as in Revelation 5:11: and the number of them was ten thousand times ten thousand.
Alphabetic Systems
139
Aa
Ii
Bb
Kk
Gg
Ll
Dd
Mm
Ee
Nn
Vv
Xx
Zz
Oo
Hh
Pp
Qq
<,
1000s
Rr
/A
Ss
/B
Tt
/G
Uu
/D
Cc
/E
Ff
/V
Yy
/Z
Ww
/H
>.
/Q
10,000s
1s
10s
100s
562 = CXB
hendeka, dodeka, treis kai deka, tettares kai deka, and so on. For instance, Threatte
(1980: 114) provides a number of examples from Attica where GI, ZI, HI, and
QI appear for 13, 17, 18, and 19. Starting in the Roman period, the signs became
more rigidly fixed in highesttolowest order. There was a mild taboo against the
use of theta (Q) for 9 in some texts because it is the first letter of death;
circumlocutions were used instead, such as writing the number lexically, or as an
additive combination of alphabetic numerals such as E (5 + 4) or AH (1 + 8)
(Smith 1926: 69; Thomas 1977).
Because alphabetic numerals could easily be confused with written words, classical Greek alphabetic numerals were sometimes distinguished from the rest of the
text with special signs, most commonly a horizontal stroke above the numeralphrase, but occasionally with dots placed to either side of it. One of the problems
in identifying earlier Greek alphabetic numerals is the lack of such marks, meaning that any single letter could be an alphabetic numeral or a nonnumerical label.
Even in later periods, numerals frequently appear without any indicator mark
whatsoever (Threatte 1980: 115).
In most Greek monumental inscriptions, the only fractions used are acrophonic signs for fractions of different monetary units (Tod 1950: 134). In mathematical
and literary texts, an entirely different system was used in which small accents
or strokes placed above and to the right of a numeral indicated unitfractions
(Thomas 1962: 43). Special signs existed for 1/2 (< and U) and 2/3 (), in addition to standard unitfractions (Thomas 1962: 45). From the second century
ad onward, the requirement of using only unitfractions was lifted, and fractions
were expressed with both numerators and denominators using alphabetic numerals. Finally, a special system for fractions was used in astronomy, combining the
alphabetic numerals with sexagesimal structures borrowed from the Babylonians.
There is no evidence for an early (eighth century bc or earlier) origin of the
numerals, although prominent classicists such as Larfeld (190207) endorsed this
140
Numerical Notation
theory, arguing for descent from the Phoenician alphabetic system, because the
Greek alphabet was borrowed from a Phoenician ancestor and because many Semitic scripts have alphabetic numerals (Brunschwig and Lloyd 2000: 388). Yet no
Semitic consonantal numerical notation systems existed before the second century
bc, and they were based on Greek rather than the other way around (Gow 1883).
None of the very earliest Greek alphabetic inscriptions contains numerals. Conversely, however, the oncepopular theory of a very late origin (late fourth or even
third century bc) cannot be sustained in light of evidence from earlier periods
(Gow 1883).
The first epigraphic evidence for alphabetic numerals comes from a vase, dating
to around 575 bc, found at Corinth, which contains the inscription SYM g, which
Johnston reads as mixed batch of 7 (Johnston 1973: 186). There is good evidence
from Attica and Corinth for the systems use on mercantile vases in the late sixth
and early fifth centuries bc (Hackl 1909). Yet it is unlikely that the numerals actually developed in either of these localities. Rather, the numerals probably developed in western Asia Minor, in the regions of Ionia and Caria, especially in the
cities of Miletus5 and Halicarnassus, where several early instances of the numerals
have been found (Heath 1921: 3233). Another early example is found in the tunnel of Eupalinos on the Ionian island of Samos, constructed around 550 bce, in
which distances are noted using letters that are probably alphabetic numerals (as
opposed to nonnumerical or ordinal labels) (Verdan 2007: 1213). All the early
examples of the alphabetic numerals, even those found outside Asia Minor, are
written using the Ionic script, which was used in Ionia, Caria, and various Ionic
colonies throughout the Mediterranean. This reflects the predominance of Ionia
in regional and international commerce during the sixth and early fifth centuries
bc. While vau and qoppa were used phonetically in the early Ionic script, san was
not, and so had lost its place between pi and qoppa in the numerical order (Jeffery
1990: 327). Later, it was reincorporated into the script for the purpose of providing
a twentyseventh sign for the numerals to function, but was placed at the end of
the system, with the numerical value 900.
The principle of the Greek alphabetic numerals was borrowed directly from the
Egyptian demotic numerals in the early sixth century bc, only using alphabetic
signs as numeralsigns (Chrisomalis 2003; see also Zaslavsky 2003). Boyer (1944:
159) regarded the similarity between the two systems as indicative of a historical
connection, but his paper was not primarily oriented toward such an argument.
Because the alphabetic numerals were the first to use phonetic signs as numeralsigns, there will be no paleographic similarity between the Greek numeralsigns
5
Miletus, from whence the adjective Milesian, was the most important Ionian city in
Caria, the region of Asia Minor immediately to the south of Ionia proper.
Alphabetic Systems
141
and any other system. Yet the alphabetic numerals are structurally similar to the
demotic numerals. They are both cipheredadditive, base10 systems. While it has
yet to be established whether the alphabetic numerals used multiplicative notation
at an early date, both systems are multiplicativeadditive above 10,000. Unlike the
demotic numerals, the alphabetic system is also multiplicative for the thousands.
The Greeks could simply have continued the series above 1000 using 10 through
90 and 100 through 900 (/I = 10,000; /K = 20,000; etc.), so the use of twostage
multiplication may be a clue to the alphabetic numerals history. It would be
reasonable for the Greeks to adopt the multiplicative principle at the same level
as in the demotic numerals, namely 10,000. However, it would not have been feasible to find nine extra signs for the values 1000 through 9000. Consequently, the
inventor(s) of the alphabetic numerals may have had the idea of using multiplication for the thousands values as well as the ten thousands. The only remaining
problem is to explain why the Greeks, recognizing this irregularity, did not then
abandon the higher multiplicative series.
Furthermore, Greek arithmetical techniques for dealing with fractions strongly
resemble the Egyptian unitfraction (1/x) tradition of computation (Knorr 1982,
Fowler 1999a). Both systems used unitfractions formed by placing a small mark
above numeralsigns to indicate the appropriate unitfraction. As well, both used
nonunitfractions for specific values such as 1/2 and 2/3. Historians of mathematics are unanimous that the Greeks borrowed the unitfraction technique from the
Egyptians, and I see no reason to doubt that the Greek use of special signs for 1/2
and 2/3 is also a result of Egyptian influence.
Turning to the historical evidence, the demotic numerals were the predominant ones in use in Egypt (especially Lower Egypt) in the early sixth century bc,
when Greeks were just starting to encounter Egyptians in large numbers for the
purposes of international trade. Most notable among the Greek traders in Egypt
were Ionian colonists from Miletus, who set up an important emporion (port of
trade) at Naukratis in the western Nile delta in the seventh century bc. Naukratis
quickly became the central locus for cultural contact between Greece and Egypt, a
position that it held until the Ptolemaic era. Inscriptions in the Ionic Greek script
dating as early as 650 bc have been found at Naukratis (Heath 1921: 33). It should
be noted, however, that no known inscriptions from Naukratis contain alphabetic
numerals, and there are later (fourthcentury bc) inscriptions with acrophonic numerals (Gardner 1888). Since the earliest examples of the alphabetic numerals are
from containers for commercial goods, the context of the systems development
was probably in mercantile activity.
The only alternative to the hypothesis of Egyptian origin is that the Ionians
independently developed a cipheredadditive, decimal numerical notation system
within a few decades of coming into contact with Egyptians in large numbers,
142
Numerical Notation
founding a colony at Naukratis, and being exposed to the demotic numerals used
throughout Lower Egypt. These connections are too significant to be coincidental.
This should not be taken as a denial of the Greeks inventiveness, however, because
the alphabetic numerals have several distinctive properties. Firstly, while some of
the demotic numeralsigns use the cumulative principle, the alphabetic numerals
use purely ciphered signs. Secondly, as mentioned earlier, the alphabetic numerals
use the multiplicative principle for 1000 through 9000, obviating the need for
nine more signs for those values, as one could simply write a hasta before a unitsign. Finally, correlating numeralsigns with the ordered set of alphabetic signs
meant that anyone who knew the order of the alphabet could determine the signs
values as long as the episemons were taken into account. The oftenmentioned
weakness of the alphabetic numerals, that too many signs needed to be learned,
is thus illusory. In learning to read and write, Western pupils must learn twentysix alphabetic signs (in their proper order) plus ten digits in order, making thirtysix total signs in two separate series, while the ancient Greeks needed to learn only
twentyseven alphabetic signs and two auxiliary signs (/ and :), and needed only
twentynine total signs in one series.
After a period of Ionian cultural dominance between 575 and 475 bc, when
alphabetic numerals were commonly found, alphabetic numerals are found only
rarely in a period starting in 475 bc and lasting around 150 years (Johnston 1979:
27). During this period, the height of Greek civilization, Athens came to the forefront as an Aegean power, while Ionias power waned after the Milesianled Ionian
revolt of 499 to 494 bc against Achaemenid Persia. The system did not disappear
entirely; at Halicarnassus, there is evidence of its continued use (Keil 1894; Heath
1921: 3133). A curious inscription from Athens (IG I2 760) from the middle of
the fifth century bc contains a long series of alphabetic numerals, written with
Ionic letters (Tod 1950: 137). There is, as well, good reason to believe that some of
the graffiti on amphorae from the Athenian Agora had weight and volume marks
notated in alphabetic numerals (Lawall 2000). In general, however, acrophonic
numerals were used in most Greek texts during this period.
The resurgence of the alphabetic numerals in Greece around 325 bc corresponds
precisely with the rise of the Ptolemies in Egypt. In this renewed period, some
of the earliest instances of the numerals come from Egypt. Figure 5.1 depicts a
portion of a GrecoEgyptian astronomical text including a calendar for the Saite
nome (Hibeh Papyrus i 27) dating to around 300 bc, one of these very early instances (Grenfell and Hunt 1906, Fowler and Turner 1983). Ordinary (unmarked)
Greek letters stand for whole numbers, while 1/x unitfractions are represented
by signs modified with long oblique strokes. This is the earliest attested example of
alphabetic numerals used in a scientific context. Coins dating to 266 bc indicating
the regnal year of Ptolemy II Soter are, to my knowledge, the first coins bearing
Alphabetic Systems
143
Figure 5.1. Hibeh papyrus i 27, with Greek alphabetic numerals indistinguishable from
letters, but with unitfractions clearly distinguished through the addition of long upward
slanting strokes. Source: Grenfell and Hunt 1906: Plate VIII.
any cipheredadditive numerals (Tod 1950: 138). It is interesting that while these
examples come from Egypt, we have no record of the alphabetic numerals use in
Egypt during the interlude of the fifth and fourth centuries bc. The evidence, at
present, is simply too scanty to conclude what stimulus caused the rejuvenation
of the alphabetic numerals.
From the third century bc onward, the alphabetic numerals were preferred over
the acrophonic numerals throughout most of the Greekspeaking world, with only
Athens retaining the acrophonic system until around 50 bc (Threatte 1980: 117).
While the acrophonic numerals used in different citystates varied quite widely,
the alphabetic numerals had no regional variants. As such, they could be used as
an effective instrument of crosscultural communication and trade among diverse
regions (Dow 1952: 23). Whereas Greece before Alexander was highly fragmented,
rendering the development of a universal Greek numerical notation system unlikely, Alexandrine and especially Roman Greece provided a suitable environment
for the development of a single panHellenic notation. That the system was a very
concise way to represent numbers, and that it relied on alphabetic symbols that
had become invariant throughout Greece by this period in history, cannot have
hurt this process.
144
Numerical Notation
In the sixth and fifth centuries bc, the alphabetic numerals were no more than
a system for labeling mercantile containers. All the early instances of the systems
use are from marked vases and potsherds. Even then, most numerals on vases are
acrophonic or other cumulativeadditive Greek numerals, not alphabetic ones. In
these very early contexts, the alphabetic numerals, like the acrophonic ones, were
used for cardinal quantities, particularly of money, weights and measures, and
discrete quantities of commodities, the sorts of numerical expressions likely to
be found in inventories and decrees. From the third century bc onward, though,
when the alphabetic system became the predominant one throughout the Greek
world, the numerals were used in a much wider range of contexts. In contrast to
the acrophonic numerals, which are found solely on ceramic vessels and stone,
alphabetic numerals are found, in addition, in manuscripts of various sorts as well
as on coins. As described by Tod (1950: 130134) and Threatte (1980: 115116), the
functions for the alphabetic numerals include:
a.
b.
c.
d.
e.
f.
Before the Hellenistic period, we have no evidence that the alphabetic numerals
were used for arithmetic or mathematics. Early writers such as Herodotus instead
used the acrophonic numerals to record the results of computations performed
with a pebbleboard or abacus (Lang 1957). This situation changed once the alphabetic numerals began to be used more widely. Most surviving Greek mathematical
manuscripts use the alphabetic numerals, which were used by all Greek mathematicians beginning with Archimedes and Apollinius (third century bc). Users
of alphabetic numerals had a wide variety of computational techniques available.
Schrlig (2001) suggests that while the earliest abaci are most conducive to working with acrophonic numerals, certain types of abaci were well suited to working with alphabetic numerals also. Furthermore, the Greeks developed a complex
system of computing on the fingers up to 9999, reminiscent of the alphabetic
systems structural shift above 10,000 (Williams and Williams 1995). To facilitate
multiplication and division, which were difficult to undertake using the abacus or
finger computing, the Greeks used multiplication tables written with alphabetic
numerals (Dilke 1987: 16; Sesiano 2002). Finally, Greek arithmetic was sometimes
done in a manner analogous to modern Western practice, by writing down numbers and manipulating numeralsigns directly; the primary difference between the
Alphabetic Systems
145
two techniques is that in the Greek case, because the numerals are nonpositional,
there was no need to line up numbers in columns (Smyly 1905).
One would think, as Boyer (1944: 160) comments, that the adoption of the
alphabetic numerals by such prominent Greek mathematicians would curb the
criticism of modern scholars that the system was dysfunctional for mathematics.
Yet many modern scholars denigrate the alphabetic numerals in comparison not
only to cipheredpositional systems but also to cumulativeadditive systems such
as the acrophonic numerals (cf. Boyer 1944: 160166). In particular, the need to
learn many signs and the lack of resemblance between numeralsigns for common
multiples of different powers (e.g., 5, 50, and 500) are perceived as serious weaknesses. In the only instance where a modern scholar actually attempted to learn
and use the numerals, however, the system fared very well. The classicist Paul Tannery found that calculation with alphabetic numerals took little more effort than
with Western numerals, with which Tannery, despite all his efforts, surely had far
greater experience and familiarity (Boyer 1944: 160161). In any event, only a small
fraction of the Greek texts containing alphabetic numerals serve mathematical
functions. Like virtually all numerical notation systems, numerals are primarily
for representation and only secondarily for computation.
Throughout their history, the Greek alphabetic numerals were used primarily
in Greekspeaking areas or in regions under the control of Greek speakers. During
its early history, the system was used in the eastern Mediterranean (particularly the
Aegean), Ptolemaic Egypt, and Seleucid Persia. Cursive varieties of the alphabetic
numerals on GrecoEgyptian papyri show a great deal of paleographic variability
(Foat 1902, 1905). A large number of variant multiplicative signs for 1000 and
10,000 were used in these papyri, most notably for 10,000, which was the
normal form starting in the second century ad (Brashear 1985). While alphabetic
numerals were not used for administration during the height of Roman power,
from the fourth century ad onward, they were used as the primary numerals of
administration, law, literature, and mathematics in the Eastern Roman Empire.
Whenever and wherever the Greek alphabet was used in the Middle Ages, the alphabetic numerals followed. Additionally, the Greek alphabetic numerals were the
most common system used in Arabic papyri for several centuries after the Islamic
conquest for recording the results of financial transactions (Grohmann 1952: 89).
Most of the descendants of the Greek alphabetic numerals substituted the
letters of their own scripts for the Greek signs. In the late second century bc,
the Israelites developed the Hebrew alphabetic numerals along a Greek model.
In the mid fourth century ad, the Goths developed alphabetic numerals along with
their Greekinfluenced script, while in regions of Africa under Greek influence,
the Coptic script of Egypt and the Geez script used in Ethiopia both developed
alphabetic numerical notation systems based on a Greek model. In Armenia and
146
Numerical Notation
Georgia, right on the border of the Eastern Roman Empire, scripts and alphabetic
numeral systems developed in the fifth century ad at around the time they were
Christianized. In the sixth century ad, the Syriac script abandoned an earlier
cumulativeadditive numerical notation system (Chapter 3) in favor of one based
on the Greek. Around the same time, the Arabic abjad numerals used following
the Islamic conquest of the Middle East were derived at least in part from the
Greek alphabetic system. The Glagolitic and Cyrillic numerals, developed in Slavic
regions in the late ninth century ad, under the auspices of the missionaryrelated
script development of Cyril and Methodius, closely resemble their Greek ancestor.
Finally, knowledge of the Greek alphabetic numerals in Western Europe led to the
development of a set of Latin alphabetic numerals in the twelfth century ad.
It is remotely possible that the Greek numerals are ancestral to the Brhm numerals (Chapter 6), which were used from the late fourth century bc onward in
India, and which eventually gave rise to Western numerals. The Brhm numerical
notation system is cipheredadditive and decimal, and used a variety of the multiplicative principle. The chronology of its invention, corresponding almost exactly
with the Alexandrine conquests and journeys in India, is also suggestive. However,
the Brhm system is more similar in structure and numeralsigns to the Egyptian
demotic numerals than to the Greek alphabetic numerals.
The large number of descendants of the Greek alphabetic numerals is due in
part to their longevity. This cannot be a full explanation, however; other systems,
such as the Egyptian hieroglyphs, were used over a much longer period, yet generated few direct descendants. Other longlived systems, such as the Roman numerals, spread very widely over large parts of the world due to Roman imperial power,
but they were often accepted by colonized or subordinate societies unchanged,
and did not replace indigenous systems entirely. Because the Greek system was
alphabetic, cultures borrowing the principle of alphabetic numeration tended to
modify the signs to fit their own scripts (whether alphabets or consonantaries)
rather than adopting the Greek alphabetic numerals directly, and also made minor
structural changes to the system.
In the early Middle Ages, when the Eastern Roman Empires fortunes were
prosperous, the numerals were widely used throughout Greece, the Balkans,
Egypt, the Levant, and Asia Minor, and were incorporated into the learning of all
European mathematicians. For instance, they were known to the English scholar
Bede, who described them in his De temporum ratione (The Reckoning of Time)
in the early eighth century ad (Wallis 1999). Many Western European manuscripts
include numerical glossaries, for instance, a tenth to eleventhcentury codex
(Vatican Library, Codex Urbinas Latinus 290) that lists (among other information
concerning nonLatin alphabets) the Greek lexical numerals, letter names,
and alphabetic numerals along with their translations into Roman numerals
Alphabetic Systems
147
(Derolez 1954: 106109). Such tables would have aided medieval scribes in understanding texts, but the Greek alphabetic numerals were not actively used except
in rare instances. For instance, in one ninthcentury Latin manuscript, the Greek
numerals 1 (A) to 21 (KA) stand for the letters A through X in a cryptogram to
indicate that it was written by Irish writers at the Welsh court of Merfyn Frych
(824844) (Derolez 1954: 9799). Such cryptographic uses confirm rather than
refute the obscurity of the system, however. In the Western European debate between the abacists and algorithmicists, the Greek alphabetic numerals, used by
all the great mathematical minds of antiquity, did not rate a mention.
The eventual fate of the Greek numerals was directly tied to the fortunes of the
Byzantine Empire. Byzantine mathematicians and astronomers used alphabetic
numerals exclusively until the twelfth century, and numerals such as regnal years
were often stamped on Byzantine coins. By 1300, however, the geographical extent
of the numerals use had become very limited, and mathematicians were using
Western numerals under the influence of Arab learning in Spain and Italy. In the
Byzantine Empire, mathematicians used Arabic positional numerals in marginal
notes on Euclids Elements in the twelfth century (Wilson 1981). The first major
Byzantine mathematician to recommend the switch to the Arabic numerals was
Maximus Planudes (c. 12601310) (Schub 1932). In some Byzantine astronomical
texts, Arabic numerals were used to write year numbers while Greek alphabetic
numerals were used for all other functions; even then, the Greek equivalent of the
Arabic numeralphrase was often written in the margin of the page (Neugebauer
1960: 5). In a few late Byzantine mathematical texts, combinations of Greek alphabetic numerals and Arabic positional numerals with a zero occurred, such as
(400 + 50 + 0) instead of for 450 (Neugebauer 1960: 5). The ultimate extension
of this principle was the use by some writers of only the first nine Greek letters
(A through ) to indicate the units, then adding a dot, a circle, or a special sign, ,
to indicate a zero position (Schub 1932: 59; Menninger 1969: 273274).
In 1453, with the fall of Constantinople, the Greek numerals ceased to be used
administratively. Nevertheless, they continued to be used thereafter for restricted
purposes, such as paginating the introductions to scholarly texts and enumerating
ordinal lists, just as the Roman numerals were used in Western Europe. Starting in the seventeenth century, the sign for 90 (qoppa), little used in alphabetic
contexts, underwent a palaeographic alteration, so that a new, roughly Zshaped
sign served only as a numeral, while the older Qshaped sign could serve both
alphabetic and numerical functions, but was regarded as a different sign (Everson
1998). Alphabetic numerals are still used today in many Greek Bibles to enumerate
chapter and verse numbers, in certain legal contexts, and in pagination. Given the
prestige associated with these functions, it is unlikely that the alphabetic numerals
will cease to be used entirely.
148
Numerical Notation
Coptic
The Coptic language is the last descendant of the Egyptian languages. It was
written using the Greek alphabet until the fourth century ad, when a distinctive Coptic alphabet originated based on a Greek model, but using six additional
characters taken from the demotic script to express uniquely Egyptian phonemes.
Unlike the earlier Egyptian scripts, it is written from left to right and has signs for
vowels. The adoption of the Coptic script was accompanied by the introduction of
a numerical notation system based on the model of the Greek alphabetic numerals
(Megally 1991, Messiha 1994). The Coptic numeralsigns are shown in Table 5.4.
The numerals, like the script itself, were written from left to right, and were
cipheredadditive and decimal. The numeralsigns closely resemble the Greek
uncial signs used between the fourth and ninth centuries ad. In addition, as in the
Greek alphabetic numerals, a horizontal stroke above the numeralphrase indicates
that it is a numeral rather than a word, and a slanted subscript stroke under a unitsign (the Greek hasta) indicates multiplication by 1000 (Megally 1991: 1821). There
is no known sign or multiplier for 10,000 or higher values for these numerals.
The classical age of Coptic lasted from the fourth to the tenth centuries ad,
during which time the script and numerals were used extensively, surviving the
seventhcentury ad Muslim conquest of Egypt. There may have been a geographical division in the frequency of their use, with northern Egyptian scribes using
them frequently, while southern writers tended to write out numbers using lexical
numerals (Till 1961: 80). While the Coptic numerals were generally used in formal
manuscripts, the ordinary cursive Greek alphabetic numerals were used for calculation and administration, possibly because the Coptic numerals, being uncials
without tails, were less practical for rapid writing (Megally 1991: 1821).
It is unclear whether the Ethiopic numerals (used to write the Geez language from
the fourth century ad onward) were based directly on the Greek alphabetic numerals
or derived through a Coptic intermediary. While the Ethiopic system is generally
said to derive directly from Greek, the Coptic uncial lettersigns are similar enough
to the Greek to render such a determination premature, and given the geographic
proximity of Egypt and Ethiopia, this possibility cannot be dismissed outright.
While the primary function of Coptic numerals has always been religious, in
the context of the Coptic Church, their administrative and arithmetical functions
should not be discounted. Despite the control of the population of Egypt by a succession of foreign powers, and despite the decline of Coptic as a spoken language
from the twelfth century onward, the use of Coptic numerals continued as late as
the fourteenth century. They are still used today in Coptic Christian liturgical texts
for pagination and stichometry, although for most ordinary purposes, either Western or Arabic positional numerals are preferred by those familiar with Coptic.
Alphabetic Systems
149
1s
10s
100s
1000s
6085 =
ZIMM Numerals
In Egypt in the tenth century ad, when the Coptic script was being replaced
by Arabic for most administrative purposes, a different Coptic cursive numerical
notation system developed, known as numerals of the Epakt (named after the
computation used to resolve the discrepancy between the solar and lunar calendars) (Messiha 1994: 26) or, more generally, as urf alzimm account book; register numerals. Often simply labeled Coptic numerals in modern scholarship,
it seems prudent to distinguish them from their uncial antecedents, particularly
since the zimm numerals are often found alongside Arabic or Hebrew texts, not
ones in the Coptic language or the Coptic alphabet. Their longstanding use in a
wide variety of contexts renders their importance far greater than the restrictedfunction Coptic uncials. While doubtless many users of the zimm numerals were
Coptic Christians from Egypt, this was certainly not an ironclad rule, and Egyptian Jews and Muslims writing in Arabic script frequently used Coptic numerals
instead of the Arabic abjad or Hebrew alphabetic numerals. The zimm system is
shown in Table 5.5.
This system is structurally identical to the classical Coptic system, but the numeralsigns are cursive minuscule letters rather than uncial ones. Many of these
signs bear little or no resemblance to the classical signs. Some of them are probably taken from the signs of the Arabic abjad numerals, which I will describe later,
while others are of indigenous development. This system also uses two stages of
multiplication (at 1000 and 10,000), like the Greek alphabetic system. A horizontal stroke and two dots placed below a sign indicated multiplication by 10,000
(Sesiano 1989: 64). A fifteenthcentury multiplication table includes instructions
on writing numerals of the Epakt, and indicates that this sign for 10,000 could
be used in conjunction with any of the twentyseven letters, thus allowing any
number less than ten million to be expressed (Sesiano 1989: 5455). It may be
that the twostage multiplicative principle at 1000 and 10,000 existed even in
Numerical Notation
150
Table 5.5. Zimm numerals
1
1s
10s
100s
1000s
7104 =
the earlier uncial numerals, and that we simply have no paleographic evidence to
confirm this.
The paleographic relation between the first and second Coptic systems remains
unclear; although some of the signs are related, more are not, and in fact many
are modifications of Greek minuscule letters rather than Coptic uncials. These
numerals were often used in early bilingual ArabicCoptic documents, suggesting
that the Arabs were making concessions to local administrators. This situation is
quite extraordinary, because many Egyptian Arabs by this time were employing
the cipheredpositional Arabic numerals used today. This suggests that the advantages of cipheredpositional systems over cipheredadditive ones may not have
been perceived to be important. In addition to Arabs, many users of the zimm
system were Egyptian Jews who used the system alongside their own alphabetic
numerals (Goitein 196788: II, 178). Their primary functions were commercial,
including contracts, bills of sale, payments, and accounts. There is no evidence
for or against the position that numerals were manipulated for arithmetic; this
surely would not have been done on permanent media, even perishable ones like
papyrus.
Although zimm numerals originated in commercial contexts and are rare in attested astronomical and mathematical texts, Lemay (1982: 384) has found one early
text dating to ad 938 (MS Garullah Efendi 1508, Sleymanie Library, Istanbul)
in which the numbers in the text are written lexically, the Arabic abjad numerals
used in astronomical tables, and the zimm numerals used in pagination. King
(1999: 7678) discusses an astrolabe made in Cairo in ad 128283 with an Arabic
inscription but zimm numerals. At this period, particularly in Egypt, the Coptic
calendar was used for various purposes by Muslims, potentially explaining this
artifacts unusual blend of representational systems. Zimm numerals were also
used very widely for pagination and foliation of Arabic texts written by adherents
of all three Abrahamic faiths, throughout Egypt and more broadly throughout
the Maghreb (Troupeau 1974). Mingana (1936: II, 183184) reports that an Arabic
Alphabetic Systems
151
The use of highly ligatured Roman numerals in late medieval Europe, and of special
complex numeralsigns in Chinese banking, are additional examples of the practice of
using unusual numerical systems to avoid fraud, and are somewhat paralleled by the
modern requirement of writing out numerals lexically in addition to using Western
numerals on checks.
152
Numerical Notation
what was by this time a highly cryptic numerical code (Hamilton 2006: 38). They
ceased to be used only in the seventeenth century, after which time Arabic positional numerals were used for these functions (Ritter 1936; Messiha 1994: 26).
Ethiopic
The Ethiopic script developed in the late third century ad, primarily on the model
of the MinaeoSabaean consonantary used in South Arabia, but also influenced
by the Greek and Coptic alphabets used to the north. It was used (and continues
to be used) for writing various languages of Ethiopia, especially Geez, the liturgical language of the Ethiopian church, and modern Amharic. While the earliest
Ethiopic script has no signs for vowels, by the fourth century ad, the script was
an alphasyllabary, in which each individual sign represents a consonant + vowel.
The direction of writing is always left to right. While the Ethiopic script is based
on South Arabian writing, its numerical notation is strictly Greek and Coptic in
origin. The earliest Ethiopic numerical notation, such as that found on third and
fourthcentury inscriptions from the kingdom of Aksum (on which the signs for
70 and 90 are not found) is shown in Table 5.6 (Drewes 1962; Ifrah 1998: 247).
The earliest inscription with Ethiopic numerals is on the late thirdcentury Anza
stela erected by Bazat, negus (local king) of Agabo, in which the highest number
expressed is 20,620 (loaves of bread) (Drewes 1962: 6567). The modern Ethiopic
script uses a system derived from these early inscriptions, structurally unchanged
but slightly modified, as shown in Table 5.7 (Fossey 1948: 99; Haile 1996: 574).
The Ethiopic numeralsigns are not letters of the Ethiopic script, but rather
are derived from the Greek or Coptic letters. Even though the signs have no nonnumerical meaning in Ethiopic, the signs normally have marks both above and
below them to indicate that their value is numerical. This practice was universal
only from the fifteenth century onward, and it is not found at all in the Aksum
inscriptions (Ifrah 1998: 246247). In addition to these signs, modern Amharic
texts use a sign for 1000 ( ), which lacks marks above and below it (Bender et al.
1976: Table I). It may be that the minimal demand for writing very large numbers
led to the abandonment of the higher multiplicative formations and the introduction of an indigenous 1000sign.
The system has a hybrid structure: it is cipheredadditive for the units and tens
but multiplicativeadditive for numbers over 100. For 10,000, a multiplicative
sign consisting of two ligatured 100signs was used, as shown in Table 5.7. Thus,
because there are special signs for 100 and 10,000, the Ethiopic system has a mixed
base of 10 and 100, with the base10 formations governed by eighteen ciphered
characters for 1 through 9 and 10 through 90 and the hundreds and ten thousands
governed by multiplicative powersigns that combine with the ciphered signs.
Alphabetic Systems
153
1s
10s
100s
1000s
etc.
10,000s
etc.
20,620 =
This flexibility allowed the Ethiopic system to express very large numbers with
a very limited set of numeralsigns. In theory, it could be extended infinitely by
adding additional signs for 10,000 as needed, although such large numbers rarely
occurred in practice (Guitel 1975: 272273).
The development of the Ethiopic system occurred under cultural contact and
Christianization by Egyptian and Syrian missionaries. While some scholars argue
that borrowing of the signs must have been from the Greek uncial script (Bender
et al. 1976: 124; Ifrah 1998: 246), Haile (1996: 574) raises the possibility that this
transmission might have taken place by means of a Coptic intermediary. Egyptian
missionaries were active in Ethiopia throughout the fourth century ad, which
might tip the scales slightly in favor of a Coptic origin. Saint Frumentius, generally held to be the first major converter of the Aksumites, was a Syrian by birth,
trained by Greeks, but his missionary work was based in Alexandria and focused
on establishing connections between the Aksumites and the Egyptian Copts. The
Table 5.7. Ethiopic numerals (modern)
1
1s
10s
100s
1000s
10,000s
647,035 =
100,000,000 =
154
Numerical Notation
South Arabian numerals were already extinct by this time and had no influence
on the Ethiopic system.
The use of base100 for the multiplicative component of the system is the Ethiopic systems most notable feature. No other alphabetic system begins using the
multiplicative principle until 1000 or 10,000. This innovation had a clear antecedent in the Greek and Coptic use of multiplication at 10,000, but it eschews the
extra nine signs for 100 through 900 needed for both Greek and Coptic numerals.
Instead, it employs the entire set of eighteen ciphered signs, then the same set again
beside the 100sign, then the same set beside the 10,000sign. While the Ethiopic
numerals were based on the Greek or Coptic systems, the Ethiopic script was not.
Of all the descendants of the Greek alphabetic numerals, the Ethiopic system is the
only one to use nonphonetic signs as numeralsigns. While this makes it more cumbersome, since one needs to learn all the scriptsigns as well as twenty numeralsigns,
there is also no impetus to use all the borrowed numeralsigns. In systems that assign
numerical values to an ordered series of scriptsigns, it is natural that one would
assign values to all the signs, rather than stopping at some arbitrary point. In this
case, however, the Aksumites borrowed the first eighteen symbols of the alphabet
and then, rather than adopting another nine signs for 100 through 900, which were
meaningless to them, took only the sign for 100 and used the multiplicative principle
thereafter, thereby reducing the number of new signs they needed to learn.
Many Aksumite inscriptions on coins (indicating regnal years) and stone monuments (indicating cardinal and ordinal quantities of various kinds) contain Ethiopic numerals. After the fall of the kingdom of Aksum and the Islamic conquest,
the Ethiopic script was used only rarely, and over time became an esoteric script
known only to scholars and Ethiopian Orthodox priests. It was used in the mathematical, numerological, and astronomical texts of medieval Ethiopia, both for
writing whole numbers and for writing fractions in a special sexagesimal notation
based on Greek traditions (Neugebauer 1979; see the following discussion). In Amharic texts, Ethiopic numerals were used for a wider variety of functions from the
fifteenth century to the present day; for instance, they were used in the personal
correspondence of Amharic elites in the nineteenth century (Pankhurst 1985). The
New Testament printed in Amharic in 1852 uses the Ethiopic numerals throughout
for page, chapter, and verse numbers (Novum Testamentum in linguam amharicam
1852). Today, the numerals are still occasionally used for writing dates, but have
largely been supplanted by the Western numerals (Bender et al. 1976: 124).
Gothic
The Gothic alphabet was developed around 350 ad by Wulfila, a bishop who translated the Bible into his native language. Gothic was an East Germanic language
Alphabetic Systems
155
a
b
10s
j
k
100s s
t
665 = .xoe.
1s
c
l
u
d
m
v
e
n
w
f
o
x
g
p
y
h
q
z
i
r
{
spoken by the Germanic tribes who migrated throughout Europe in the latter
years of the Western Roman Empire (Ebbinghaus 1996: 290). The script was
alphabetic and written from left to right. Along with the script, an alphabetic
numerical notation system was employed, as indicated in Table 5.8 (Braune and
Ebbinghaus 1966: 10).
Like the Greek alphabetic numerals, the Gothic numerals were usually distinguished from the rest of the text either through dots to either side of the numeralphrase or by placing a horizontal stroke above the phrase (Braune and Ebbinghaus
1966: 10). The Gothic numerals never expressed quantities higher than 1000, and
thus there is no evidence of the use of the multiplicative principle. Larger numbers
were always written out in full (Menninger 1969: 260). The system is therefore
cipheredadditive and decimal throughout. The numeralsigns are related to the
Greek uncial letters. Of the episemons, f (6) was the sixth letter of the Gothic alphabet, and had the phonetic value [kw]. The other two episemons, qoppa (r) and
san ({), had no phonetic value and were simply used to fill out the full complement
of twentyseven signs using Greek models (Fairbanks and Magoun 1940: 319). The
possibility has been considered, but now largely rejected, that the Gothic script
owes its ancestry at least in part to either the Latin alphabet or the Germanic runes
(Ebbinghaus 1996: 290291). However, since neither of these scripts uses alphabetic numerals, the Gothic numerals must be of Greek origin.
The Gothic alphabet is attested in only a limited set of documents, mostly
translations of parts of the New Testament, but also in a small number of secular texts. Most numerals in Gothic texts indicate chapter and verse numbers in
Bibles. Additionally, they were used within the text to indicate numerical values,
while in Greek Bibles such numbers were always written out in full (Menninger
1969: 260). There is no evidence that the Goths ever did arithmetic or mathematics using these numerals. The sign for 900 is only attested in the Codex
Vindobonensis 795, an antiquarian text of the ninth century, by which time the
script had already fallen out of use (Fairbanks and Magoun 1940, Ebbinghaus
1978).
156
Numerical Notation
Very little of what surely must have been written in Gothic has survived.
Most surviving texts date to the sixth century ad, although the assignment of a
fourth century ad origin to the numerals is undisputed. It is unclear exactly when
in the seventh or early eighth century the Gothic language died out, but around
that time the script and numerals ceased to be used, and were replaced by the Roman numerals that were coming to be used throughout Western Europe.
Hebrew Alphabetic
The earliest Hebrew scripts began to diverge from the earlier Phoenician consonantary in the ninth or tenth century bc. Then as now, Hebrew consisted of twentytwo consonantal signs, written from right to left, and placed in a customary
order. Early Hebrew inscriptions used a variant of the Egyptian hieratic numerals
(Chapter 2). Somewhat later, particularly in the fifth and fourth centuries bc,
some Hebrew speakers used the Aramaic numerals (Chapter 3) for administrative
and commercial purposes. Only at a much later date, probably in the second century bc, did a uniquely Hebrew set of numerals develop. This system is indicated
in Table 5.9.
The first twentytwo signs, indicating 1 through 400, are the letters of the Hasmonean Hebrew script circa 125 bc, at the time of the writing of the Dead Sea
Scrolls and also when the numerals were first used (Goerwitz 1996: 488). The
system is decimal and cipheredadditive, and written from right to left. Values
between 500 and 900 were represented using the sign for 400 in conjunction
with one or more signs for the lower hundreds (i.e., 500 = 400 + 100, 600 = 400 +
200 ... 900 = 400 + 400 + 100). This structural irregularity exists because there
are too few letters in the Hebrew consonantary to fill out the twentyseven signs
needed to extend the system to 900. The number 400 occupies a special structural
role, but is not a base of the system, as its powers (16,000, 6,400,000, etc.) do not
receive any special treatment.
While the very earliest Hebrew inscriptions contain no signs for numbers above
1000, the need to do so quickly arose, as the numerals began to be used for dating
on grave inscriptions using the Hebrew calendar. For multiples of 1000, a mark
either a small curved stroke to the left of a numeralsign or two dots placed above it
could be used to indicate multiplication by 1000; thus, would signify 9000
and 90,000. This feature is similar to, but distinct from, the Greek alphabetic
system, which adds a stroke above or below a numeral to indicate multiplication by
1000, but begins again at 10,000 by placing the multiplicand above the sign Z. Thus,
while the Greek system could express any numeral up to ten million, as opposed to
one million for the Hebrew numerals, the Hebrew system is arguably easier to use
because it has only one value at which the multiplicative principle is employed.
Alphabetic Systems
157
1s
10s
100s
369 =
The Hebrew alphabetic numerals were derived from the Greek alphabetic numerals between 125 and 100 bc for use on coins inscribed with the Hasmonean
script (Gandz 1933: 76; Millard 1995: 192). Theories of a much earlier (ninth to
seventhcentury bc) independent origin are no longer credible (Smith and Karpinski
1911: 33; Gandz 1933: 7576; Schanzlin 1934; Zabilka 1968: 176178). A bilingual
ostracon from the Palestinian village of Khirbet elQm and dating to 277 bc uses
both Aramaic and Greek alphabetic numerals to denote a financial transaction
(Geraty 1975). Although it is not a Hebrew text, the village was in Judah and hundreds of contemporary Hebrew inscriptions have been found there. If the Hebrew
alphabetic numerals had been in use at that time, they likely would have been used
instead of the Aramaic numerals in a situation where the Greek inscription a few
lines below used a similar system.
The first safely dated instance in which the use of Hebrew alphabetic numerals
is certain is on coins from the reign of the Hasmonean king Alexander Janneus
(103 to 76 bc), some of which were stamped with Greek script and alphabetic
numerals, others with the Hebrew script and alphabetic numerals, in both cases
using alphabetic signs (Naveh 1968). Avigad (1975) suggests that a clay seal reading
Jonathan high priest Jerusalem M) might refer to the fortieth (m = 40) year of
the Hasmonean kings, and thus dates it to 103 bc, early in the reign of Alexander
Janneus, whose Hebrew name was Jonathan. That these early Hebrew alphabetic
numerals would both be found in the context of the same man is strong circumstantial evidence.
Despite the special formation of Hebrew numerals from 500 to 900, the similarities between the Greek and Hebrew numerical notation systems are striking.
The two systems share not only a similar structure (decimal and cipheredadditive)
but also the alphabetic principle for forming the numeralsigns. The notion that
the Hebrew numerals were independently developed can no longer seriously be sustained, despite the agnostic attitude of some scholars, including Ifrah (1998: 239). That
Hasmonean coins were struck in both languages and using both systems provides
specific contextual evidence that the Hasmonean kings adopted the technique
158
Numerical Notation
from the Greeks. The cultural influence of the Ptolemaic and Seleucid kingdoms
in the Levant at this time was enormous; Greek alphabetic numerals were used on
coins from the Phoenician cities of Sidon, Tyre, Byblos, and Akon from the mid
third century bc onward (Millard 1995: 193). At the same time, the Hebrew use
of 400 as a steppingstone for representing the higher hundreds is an important
innovation, as it did not require that Hebrew speakers learn and adopt additional
nonphonetic signs.
Despite the development of the Hebrew numerals, most Semitic peoples of the
ancient Levant (Nabataeans, Aramaeans, Palmyrans) never used alphabetic numerals, but continued to use their own hybrid cumulativeadditive/multiplicativeadditive systems (Chapter 3), which coexisted with the Hebrew alphabetic system for several centuries, and were replaced only over a long period. Until the
seventh century ad, inscriptions on Jewish graves throughout the Mediterranean
region were often written, not with the Hebrew numerals, but rather in the Greek
alphabetic numerals (Ifrah 1998: 238239). Gradually, the Hebrew numerals replaced the Greek and Levantine systems, until they were firmly established as a
distinctive system peculiar to the Jewish populations of Europe, North Africa,
and the Levant. Also, in the sixth century ad, the Hebrew numerals were partly
or wholly used by the creators of the Syriac alphabetic numerals (to be discussed),
which are also cipheredadditive and decimal and have the same break at 400 as
the Hebrew system.
The modern Hebrew alphabetic numerical notation system has the same structure as the ancient system, but uses modern scriptsigns, as shown in Table 5.10.
Around the beginning of the tenth century, five additional Hebrew characters, the
signs used for kof, mem, nun, pe, and tsade in wordfinal position, were sometimes
used to complete the sign set for 500 through 900. These signs occur in some
Masoretic commentaries on the Old Testament, but were never the regular forms
used (Gandz 1933: 96102). The older formations using additive combinations of
hundredssigns are the usual means of representation. The newer forms probably
were not widely accepted because the wordfinality of these signs is inconsistent
with the principle of the numerical notation system that the highest values should
come first in a numeralphrase (Gandz 1933: 98). To put a wordfinal letter for
500 through 900 at the head of a numeralphrase would be inconsistent with its
original purpose; to put it at the end of the numeralphrase would be inconsistent
with the rule of decreasing ordering of the powers.
Also in the tenth century, Hebrew scholars became aware of positional numerals, and occasionally experimented using combinations of the alphabetic numerals
and the positional principle. In a Masoretic poem by Saadia Gaon (ad 882942),
numbers are written quasipositionally in two columns; the rightmost represents
the thousands, while the leftmost column represents the ones, so that 42,377 was
Alphabetic Systems
159
1s
10s
100s
OR
160
Numerical Notation
dating. However, the fact that such a ruling needed to be made at all shows the
continued health and vibrancy of Hebrew alphabetic numerals, albeit in a limited
set of religious contexts. The Hebrew alphabetic numerical notation system is
now over two millennia old, and is one of the oldest systems in continuous and
regular use.
Syriac Alphabetic
In Chapter 3, I described the numerals used alongside the Old Syriac script between about 50 and 500 ad. Around the fifth century ad, this script diverged into
two forms: an eastern variety, Nestorian, used by the Christians of Persia, and a
western variety, the Serto script used by the Jacobite Christians of Syria (Daniels 1996). This split was precipitated by the expulsion of the Nestorian Christians from the Byzantine city of Edessa and their subsequent migration into the
Sassanian Empire (Duval 1881: vii). Soon thereafter, both the Nestorian and Serto
scripts began to use an alphabetic numeral system akin to those used elsewhere in
the Middle East. The basic signs of this system (signforms are those of the Serto
script) are shown in Table 5.11.
The system is decimal and cipheredadditive, and written from right to left.
As in many alphabetic numeral systems, numerals were sometimes distinguished
from letters by placing a horizontal stroke above a numeralphrase, but often no
special mark was present. Like the Hebrew alphabetic numerals, values from 500
to 900 were usually written using the signs for the lower hundreds with the sign
for 400 in various additive combinations. Alternately, the upper hundreds were
occasionally expressed multiplicatively, by placing a small dot above the signs for
50 through 90 to indicate multiplication by 10 (Duval 1881: 15; Noldeke 1904:
316317). For values above 1000, a slanted stroke placed beneath a unitsign indicates multiplication by 1000, while a horizontal stroke placed beneath a sign indicates multiplication by 10,000 (Hoffman 1858: 67). In this way, any number
below ten million could be expressed. Even higher values could be expressed by
placing two small strokes beneath a sign to indicate multiplication by ten million,
and placing one small stroke above and one small stroke beneath a sign indicated
multiplication by ten billion; however, these techniques were used extremely rarely
and not followed uniformly (Duval 1881: 1516). As in several other alphabetic systems, placing an oblique stroke above a numeralsign indicated the appropriate
(1/x) unit fraction (Duval 1881: 16).
The Syriac alphabetic numerical notation system was probably invented in the
sixth century ad, and gradually replaced the Old Syriac system over the next two
hundred years (Duval 1881: 15). Independent invention of this system can be ruled
out, given its strong similarity to others used in the region. Two likely possibilities
Alphabetic Systems
161
1s
10s
100s
a
j
s
b
k
t
c
l
u
d
m
v
e
n
sv
w\
/e
e
f
o
tv
x
/f
f
g
p
uv
y
/g
g
h
q
vv
\z
/h
h
i
r
svv
,
/i
i
OR
/a
10,000s a
369 = iou
1000s
/b
b
/c
c
/d
d
are that it was modeled on the Greek alphabetic numerals prevalent in the Byzantine Empire or else on the Hebrew alphabetic numerals. The Hebrew and Syriac
numerals are the only two systems in which 400 occupies a special structuring
role, in that the higher hundreds are expressed using additive combinations of
400 and the lower hundreds. The ordering of the Syriac numerals follows the letterorder shared by the Syriac and Hebrew scripts, not that of Greek. Finally, the
Syriac system was written from right to left like the Hebrew system, but unlike
the Greek. On the other hand, the Syrians were closely affiliated with Eastern
(Greek) Christianity, and many Syrians lived under Byzantine rule. The hypothesis of Greek ancestry is supported by the shared feature of the two systems that
the multiplicative principle was used at two different powers of the base, 1000
and 10,000, whereas the Hebrew numerals did so only at 1000. The inventor(s) of
the Syriac numerals probably would have been familiar with both the Greek and
Hebrew numerals, so it is possible that features of both systems were blended in
the Syriac system.
The Syriac scripts were never used as the official script of any polity, and thus
Syriac numerals are rarely found on stone monuments or coins. However, their
use in religious texts is extremely prevalent from the time of their invention to
the present day, a situation afforded by the relative separation of the Jacobites and
Nestorians from both Roman and Greek Christianity. Most Syriac liturgical texts
are dated and paginated using the numerals, making it easy to examine paleographic changes. There is no evidence that the Syriac numerals were ever used in
mathematical texts, nor, in contrast to the Hebrew numerals, were they used for
lettermagic or numerology.
Although the heyday of the Syriac scripts came and went before ad 1000, both
the Nestorian and Serto scripts survive to this day, the former in Iraq, Turkey,
and Iran among a small number of Nestorian Christians, the latter in Lebanon
162
Numerical Notation
among the Maronite Christians of that country. Both scripts retain their distinctive numerical notation systems, albeit restricted in the contexts of their use to
the same liturgical functions for which they have been employed for nearly 1,500
years. Arabic or Western numerals serve most other purposes. Nevertheless, there
is no reason to think that the Syriac numerals are about to disappear, particularly
given the special status accorded to Maronite Christianity in Lebanons 1990 constitution.
Arabic Abjad
In preIslamic times, Arabic speakers used a variety of the Nabataean script and
numerals (Chapter 3). While the classical Arabic script is directly descended from
this ancestor, the Nabataean hybrid cumulativeadditive/multiplicativeadditive
numerals were abandoned in favor of a cipheredadditive system based on the
Arabic scriptsigns. The basic signs of this system are shown in Table 5.12 (Saidan
1996: 332).
The system is decimal and cipheredadditive below 1000 and, like the Arabic
script, is written from right to left with the signs in descending order. The numeralsigns shown are the unligatured signs of the Arabic consonantary; in numeralphrases, signs were ligatured to one another as appropriate for the letters in question. Often, a horizontal stroke was placed above a numeralphrase to distinguish
it from an ordinary word. Curiously, the signs are not valued according to the
normal Arabic letterorder, but rather according to the letterorder of the Hebrew
and Syriac scripts, which was also the order used early in the Arabic scripts history.
The first three signs of this order (alif, ba, jim) gave the system its most common
name, h.isab abjad.7 Because the Arabic script has twentyeight basic consonantal
signs, the remaining sign, ghayn (;), was assigned the numerical value of 1000.
Ghayn was not used as a single unitsign, but as a multiplier in numeralphrases
by placing another sign before it. In this way, any number up to and including one
million could be written. The system is thus multiplicativeadditive above 1000.
The values assigned to six of the abjad numerals were different among users of
the Arabic script in the Maghreb (North Africa and Spain). This ordering developed somewhat later than that used elsewhere, probably in the ninth century ad.
Other than the different values assigned to the six signs in Table 5.13, the system is
structurally identical to the regular abjad numerals.
It is possible that the abjad numerical system is of preIslamic origin, and that
it spread from the north. However, it is more likely to have originated around
650 ad, at or shortly after the time of the early Islamic conquests in Syria, Egypt,
7
The first three letters in the modern Arabic script are alif, ba, and ta.
Alphabetic Systems
163
b
k
t
c
l
u
d
m
v
e
n
w
f
o
x
g
p
y
h
q
z
i
r
:
1000
a
j
s
;
7642 =
1s
10s
100s
and Mesopotamia. Under Byzantine rule, this region used the Greek alphabetic
numerals for administrative and commercial functions; furthermore, both the
Syriac and Hebrew alphabetic numerals were used in their respective scripts. The
abjad numeralsigns have exactly the same order as the corresponding Syriac and
Hebrew characters up to 400 (above which point the other two systems are structurally irregular). The Arabic script adopted a different letterorder starting in the
eighth century ad, but the older order was retained for the numerals. The Greek
system follows an order similar to that of the Syriac and Hebrew numerals up
to 80, but diverges thereafter by putting sampi (equivalent to the Hebrew tsade)
at the end of the system, rather than in the middle. Thus, either the Hebrew or
Syriac systems inspired the development of the Arabic abjad numerals (Guitel
1975: 276278; Ifrah 1998: 243). Table 5.14 illustrates how the Arabic order is directly parallel to the Hebrew and Syriac, while the Greek numerals diverge from
them starting at 90.
Table 5.13. Arabic abjad numerals (Eastern vs. Maghreb)
Sign
o
r
u
z
:
;
Lettername
Eastern Value
Maghreb Value
sin
60
300
sad
90
60
shin
300
1000
dad
800
90
900
800
ghayn
1000
900
Numerical Notation
164
Table 5.14. Arabic, Greek, Hebrew, and Syriac numeralsigns and letters
Arabic
60
70
80
90
100
200
300
400
500
600
700
800
900
o
p
q
r
s
t
u
v
w
x
y
z
:
Greek
sin
ayin
fa
sad
qaf
ra
shin
ta
tha
kha
dhal
dad
a
X
O
P
<
R
S
T
U
F
C
Y
W
>\
Hebrew
Syriac
xi
samekh
omicron
ayin
pi
pe
qoppa
tsade
rho
quf
sigma
resh
tau
shin
upsilon
tav
phi
chi
psi
omega
sampi
o
p
q
r
s
t
u
v
w
x
y
z
,
semkat
e
pe
sode
quf
rish
shin
taw
However, the Arabic abjad was rarely used in the same texts as either the Hebrew or the Syriac numerals. In contrast, Greek alphabetic numerals are found in
Arabic documents from the seventh to the ninth centuries ad. In ad 706, Caliph
Walid I dictated that, although his Greek financial administrators in Damascus
should no longer use the Greek alphabet, they could continue to use the Greek
alphabetical numerals (Menninger 1969: 410). In an eighthcentury ad Arabic
tax record, numbers are expressed in both Greek alphabetic numerals and abjad
numerals (Cajori 1928: 29). Because many of the regions conquered by the Arabs
even those such as Syria and Palestine in which the Syriac or Hebrew numerals
were found were under Greek rule, Greek numerals were the normal system
used for administration, on coins, and in inscriptions. It is unlikely that the Greek
system played no role in the development of the abjad numerals.
The phonetic values of the final six Arabic characters do not correspond to the
Greek, however. Furthermore, the Greek system has only twentyseven rather
than twentyeight signs (lacking a sign for 1000). While the Arabs were aware that
the Greek system had signs for the higher hundreds, and may thus have attached
Alphabetic Systems
165
numerical values to the remaining signs in their own script, the use of a special sign
for 1000 is unique to the Arabic system among all four systems under consideration. More likely, the Arabic system was based on the Semitic letterorder but employed the structural advantages of a system, such as the Greek, with a full complement of numeralsigns. This feature would have been particularly important, since
the administrative needs of the new Islamic caliphate were growing rapidly.
By the late eighth century ad, the Arabic abjad numerals had spread throughout the Middle East and into the Maghreb, including southern Spain. They were
used on administrative documents, in literary and scientific texts, on dated coins,
and on monuments, though not for the most part on ephemeral media such as
ostraca. In areas in which an existing administrative apparatus was retained from
the Byzantine Empire (such as Egypt), the Greek and Coptic alphabetic numerals
were used much more frequently on administrative and financial documents than
were the abjad numerals (Grohmann 1952: 89). While the systems are structurally
similar to the abjad numerals, it would have taken some effort for Arabic writers
to learn an entirely new set of twentyseven signs, so its failure to be used more
widely is somewhat surprising. Issues of identity and ethnicity may have played a
significant role in determining the scope of their use. For instance, Ifrah describes
a ninthcentury ad Christian manuscript written in Arabic but in which the verses
are numbered in Greek (Ifrah 1998: 243). In this case, the writer was a Christian
who associated himself with Greek Christianity through the alphabetic numerals
even though he wrote in the Arabic language and script. Abjad numerals were
rarely if ever used directly for arithmetical calculation. Instead, a system of finger
computing borrowed from the Byzantine Greeks, and known as h.isab aluqd
(fingerjoint arithmetic), was employed (Saidan 1974: 367368), as well as computation on dustcovered boards.
Despite their use over a wide area, the abjad numerals did not give rise to a
large number of descendants. In part, this must be due to the remarkable stability
of the Arabic script itself. The Coptic zimm numerals used in Egypt from the
tenth century ad onward are an interesting blend of Greek and Arabic influences,
although relatively few of the numeralsigns are evidently based on those of the
abjad, and it is unclear to what extent the Arabic abjad affected the structure of
the system. A similar situation probably arose in Morocco, where Fez numerals,
incorporating elements of the Arabic abjad as well as GrecoCoptic alphabetic
numerals, were used until very recently (Colin 1933). Finally, like Greek and Hebrew scholars, Arabic astronomers used a very unusual system for writing fractions, which combines a cipheredadditive system with a base60 (sexagesimal)
positional notation.
Shortly after they were invented, the abjad numerals began to be supplanted
by the wellknown cipheredpositional Arabic system (Chapter 6) borrowed from
166
Numerical Notation
the Hindus, and known in Arabic as alh.isb alhind (Indian arithmetic). The
Islamic conquest of enormous territories to the east brought the Arabic and Indian
spheres of influence into close contact by the mid seventh century. As Islam spread
eastward throughout the eighth century ad as far as the Indus River, the Indian
style of numeration began to diffuse westward and supplant the Arabic abjad,
which itself was still a novelty in western regions such as North Africa. This replacement was hastened by the arrival in 773 ad of Hindu astronomers and astronomical knowledge at the court of Caliph alMansur in Baghdad, which formed
the basis for the early ninthcentury ad writings of the famed Arab mathematician
alKhwrizm, who popularized the Arabic positional system (Menninger 1969:
410411). By the late ninth century, the positional system was being used in administrative and financial documents, and by the late tenth century, on inscriptions (Grohmann 1952: 89). The latest Arabic papyrus in which abjad numerals
express a date is from 517 ah, or 1123/4 ad (Destombes 1987: 131).
Despite the introduction of positional notation, abjad numerals survived in a
variety of contexts. Astronomers continued to use ordinary abjad numerals much
later than other writers, probably because they had also adapted them for use in
the quasipositional sexagesimal fractions to be described later. Until the thirteenth
century all astronomical and astrological texts used abjad numerals alone, and for
several centuries thereafter, abjad and positional numerals were often used side
by side in any given text (Lemay 1982: 385386). Abjad numerals were commonly
used on most Arabic astrolabes (both for marking gradations on the instrument
and for recording dates of construction) until the sixteenth century (Destombes
1987). Mauritanian mathematicians did not adopt Arabic positional numerals
until the end of the eighteenth century (Rebstock 1990). That positional numerals would be adopted more rapidly in nontechnical manuscripts than in scientific
ones is seemingly paradoxical, but only because it is customarily assumed that
positional numerals would be accepted owing to their efficiency for arithmetic. In
fact, however, Arabic arithmetic was largely conducted using dustboards, flat surfaces covered with a layer of fine dust on which calculations could be made using
any numerical system, but of course no record of this would survive; in contrast,
permanent records were often made in more traditional and authoritative numerical systems like the abjad numerals (Lemay 1982: 383384). We do not know to
what extent premodern Arabic astronomers may have performed computations
using Arabic positional numerals.
The abjad numerals survive to the present for very limited cryptographic, literary, and magical functions, such as chronograms (described earlier for Hebrew
numerals), known in Arabic as h.isb aldjummal (Colin 1971: 468). Chronograms
using the abjad numerals were common throughout the Middle Ages, particularly in Persia and Islamized parts of India (Ahmad 1973). Babajanov and Szuppe
Alphabetic Systems
167
Astronomical Fractions
In addition to the cipheredadditive decimal systems peculiar to their civilizations,
ancient and medieval astronomers from various linguistic backgrounds and script
traditions used distinct systems for fractions, combining their ordinary alphabetic
numerals with a base of 60 and the positional principle, borrowed from Babylonian
astronomy. This ingenious system of astronomical fractions represents a curious
digression in the history of numerical notation. From about 2100 bc to 0 ad, Babylonian astronomers and mathematicians used a cumulativepositional numerical
notation system with a base of 60 and a subbase of 10 (see Chapter 7). Numbers
smaller than 60 were expressed through cumulative combinations of cuneiform
signs for 10 (l) and 1 (k), while the positional principle was used to express multiples of powers of 60 (60, 3600, 216,000, ...). By the third century bc Greeks firmly
controlled most of the lands formerly under Babylonian rule, under the potent Seleucid kingdom that came into existence after the Alexandrine conquests. At some
point between the third and first centuries bc, the Babylonian positional notation
and the sexagesimal base were married to the Greek cipheredadditive numerals and used thereafter by Greek astronomers (Jones 1999: 9). The first Greek
8
Curiously, one of these systems is known as elYunani (Ionian!), suggesting that its users
were aware of the Greek origin of such notations.
Numerical Notation
168
a
b
10s
i
k
/afie k ie
1s
1515
g
l
d
m
e
n
Alphabetic Systems
169
170
Numerical Notation
that can be used with any numerical notation system as a means of combining
numeralphrases for 1 through 59 into a positional notation system for fractions.
The sexagesimal fractions were used only by astronomers and mathematicians
working with astronomical problems and writing in manuscripts. Even so, when
writing nonastronomical material, or when paginating and dating astronomical texts, they used other numerical notation systems. Yet, on astronomical instruments and in astronomical manuscripts, sexagesimal notation alongside alphabetic numerals was employed continuously for well over a millennium. That
sexagesimal notation was used solely for astronomy suggests that the demands of
the discipline led to its retention. The division of the circle into 360 degrees (with
subdivisions of 60 minutes per degree and 60 seconds per minute) is very useful,
since 60 has a large number of divisors.10 Faddegon (1932) showed that this feature
enables quick and easy multiplication and division using sexagesimal fractions.
The utility of sexagesimal fractions must therefore be evaluated in relation to computations involving this specific metrological system. The tenthcentury Persian
mathematician alBiruni reported, however, that because it was inconvenient to
multiply using a 60 60 multiplication table, Islamic astronomers would convert sexagesimal fractions into decimal numbers, multiply them, and then convert
them back into sexagesimal notation (Berggren 2002: 362).
Although sexagesimal fractions are no longer used, modern astronomers still
use the sexagesimal division of the circle, and anyone who can read a digital
clock uses a kind of sexagesimal numeration. While we no longer mix additive
and positional principles in notating time and angles, astronomers continue to
restrict themselves to values under 60 for the division of the sky into segments,
just as everyone is able to realize that thirty minutes pass between 1:50 and 2:20.
These vestiges come to us, via Greek and Arabic sexagesimal fractions, from
the Babylonian custom of numbering by 60. In this way, a peculiar custom
of numeration, useful for astronomy but not much else, influenced how humans perceive and structure time throughout most of the world today. However,
these do not represent a sexagesimal numerical notation system, but simply a sexagesimal division of various metrological units (for angles and for time) that are
then represented with decimal numerals. The notation 11:05 does not mean
665 minutes (11 60 + 5), but simply 11 hours and 5 minutes. That we continue
to measure time and angles in this way is a fascinating issue in the history of
astronomy and of timekeeping.
10
Alphabetic Systems
171
a A
10s
j J
100s s
766 = foy
1s
b B c
d D e
f
k
l L m
n N o
t T u
v V w W x X
g G h H i I
p
q Q r
y
z
, <
Fez Numerals
In the western extremity of the Muslim world, first briefly in medieval Spain, then
around the city of Fez in modern Morocco starting in about the sixteenth century,
an alphabetic numeral system was used, distinct from both the Arabic and Greek
systems. This system was known as isb alqalam alFs (Fez signs) or rm signs
(Roman, the name given to the Byzantine Greeks) (Guergour 1997: 68). The
numeralsigns, including paleographic variants where appropriate, are indicated
in Table 5.16 (Colin 1933: 199201).11
The system is decimal and cipheredadditive, and is written from right to left
with the highest values on the right. The twentyseven signs are thus sufficient to
express any number smaller than 1000. The resemblance between the signs for
6 and 7 and the modern Western numerals is likely a coincidence. For higher
numbers, a stroke placed to the left of or beneath any of the twentyseven signs
indicates that its value should be multiplied by 1000. In some documents, two
strokes placed underneath a figure indicated multiplication by one million (1000
1000) (Guergour 1997: 69). The Fez system is thus a hybrid multiplicativeadditive
system for values above 1000. At least in later periods, there were signs for specific
fractions, and for specific monetary values (Sanchez Perez 1935).
The earliest version of the Fez numerals was used among the Mozarabs (Arabic
Christians) of Toledo, Spain, in the twelfth and thirteenth centuries (Colin 1933:
204). Levi Della Vidas (1934) study of these documents, which includes a table of
these numeralsigns, confirms that they are essentially identical to the later ones
except that they are written from left to right. The question then arises how these
numerals reached Spain in the twelfth century. Three possible ancestors the Arabic
abjad (Maghreb variant), the Greek minuscule alphabetic numerals, and the Coptic zimm numerals are depicted alongside the Fez numerals in Table 5.17.
11
See also Sanchez Perez (1935) and Guergour (1997) for different paleographic variants of
this system.
Numerical Notation
172
a
b
c
d
e
f
g
h
i
j
k
l
m
n
A
b\
G
D
E
V
z\
E
Q
I
k
l
m
n
a A
b B
c
d D
e
f
g G
h H
i I
j J
k
l L
m
n N
r
p
q
z
s
t
o
v
w
x
y
:
;
x
o
p
<
r
s
t
u
f
c
y
w
>\
o
p
q Q
r
s
t T
u
v V
w W
x X
y
z
, <
All four systems are written cursively and have an enormous amount of variation. It is possible that the Mozarabs numerals are paleographic variants of the
Greek alphabetic numerals, and thus came to the Arabs of Spain via direct diffusion from the Byzantines (Levi Della Vida 1934: 283). The attribution of these
signs as rm (Byzantine) would tend to confirm this origin. However, many of the
numeralsigns bear no resemblance to the Greek alphabet. Colin suggests, rather,
that the Fez numerals (and their Spanish antecedent) were borrowed, not directly
from the Greek alphabetic numerals used in the Byzantine Empire, but by way of
the EgyptoCoptic zimm numerals (Colin 1933: 213). The paleographic similarities
between several of the Coptic and Fez numeralsigns (e.g., 8, 80, and 500) suggest
that some connection must exist between the two. Yet while the Fez numerals are
multiplicative at only one level (1000), both of these candidates for its origin are
multiplicative at both 1000 and 10,000. An alternate ancestor is the Arabic abjad
system used for numbermagic and astronomy in the Maghreb at that time. Some
of the paleographic resemblances between the Fez numerals and the zimm numerals (e.g., the signs for 7, 30, and 90) can be explained by both systems connection
to the appropriate letters of the Arabic abjad. Furthermore, unlike the other two
Alphabetic Systems
173
systems, the Arabic abjad is multiplicative only in combination with 1000. Most
likely, the Fez numerals are an unusual blend of the Greek, zimm, and Arabic
alphabetic systems adopted among a very unusual group of users, highly educated
Arabized Christians living in Muslimdominated southern Spain.
The Fez numerals did not last long in Spain; I know of no texts from the fourteenth
century or later in which they are used. They were described by the Moroccan mathematician Ibn alBanna in the early fourteenth century, and later by the great historian
Ibn Khaldun in the late fourteenth century, both of whom lived and worked at Fez
(Colin 1933: 206; Guergour 1997: 68). Their use in Morocco began in earnest only in
the sixteenth century, however, after the expulsion of the Moors from Spain in 1492.
They were used frequently throughout the sixteenth and seventeenth centuries in accounting and other commercial contexts, after which time they began to be replaced
by the Arabic positional numerals. While they were known to scholars, their use was
always mercantile and legal, never mathematical or astronomical (Guergour 1997: 74).
The Fez numerals were still used in the early twentieth century for indicating monetary
values in wills and in related legal documents (Colin 1933, Sanchez Perez 1935). Because
the meaning of the numerals was known to only a few learned notaries by this time,
the system had become a cryptographic notation to prevent fraudulent modifications
and forgeries, and generally to restrict access to information (Colin 1933: 195). Political
changes in postcolonial Morocco have ended this systems use.
Armenian
Before the introduction of Christianity, there was no native Armenian script, and
the Babylonian, Greek, and Old Persian scripts were used for literary purposes. The
Armenian adoption of Christianity in the early fourth century ad was followed
by enormous influence from the Greekspeaking world. In the early fifth century ad (probably in 406 or 407), the Armenian scholarmonk Mesrop Mashtots
(c. 360440) developed the first uniquely Armenian script, an alphabet of thirtysix letters, in order to translate the Bible from Greek into Armenian (Sanjian 1996:
356).12 At the same time, the letters of the alphabet were assigned numerical values
as shown in Table 5.18.
These signs are the erkatagir ironforged letters preferred from the fifth through
thirteenth centuries ad, and still used for epigraphic inscriptions (Thomson 1989;
Sanjian 1996: 357). In the tenth century, cursive letters known as bolorgir began to
be used, and are the standard forms used in modern Armenian writing. The Armenian system is cipheredadditive and decimal, and is written from left to right.
12
The modern Armenian script has thirtyeight letters, the last two of which (o and f)
were introduced in the medieval period and have no numerical value.
Numerical Notation
174
A
10s
J
100s S
1000s 2
346 = UMF
1s
B
K
T
3
C
L
U
4
D
M
V
5
E
N
W
6
F
O
X
7
G
P
Y
8
H
Q
Z
9
I
R
1
0
Because the ancient Armenian alphabet had thirtysix letters, it had enough signs
to express the complete series from 1000 to 9000 as well as all the units, tens, and
hundreds. The system could thus denote any number less than 10,000. However,
unlike many cipheredadditive alphabetic systems, the Armenian system does not
use multiplication to express higher values, which were written in full using lexical
numerals.
Very little epigraphic or paleographic evidence survives from the earliest centuries of the systems use. The Armenian numerals were probably developed on the
model of the Greek alphabetic numerals, just as the Armenian script itself was derived from the Greek. Many other scripts have been suggested as possible ancestors
of the Armenian script, based on resemblances in the shapes of certain characters
(Gamkrelidze 1994: 37), while there are few resemblances to the Greek alphabet.
However, of these likely ancestors, only the Greek alphabet used appropriate alphabetic numerals. Thus, regardless of the origins of the scriptsigns, the principle of
alphabetic numeration was certainly borrowed from Greece. It is unclear whether
the Armenian alphabetic numerals were developed by Mesrop Mashtots himself (or
his assistants) in the early fifth century ad, or whether they were produced later in
the century. Figure 5.2 is a monumental grave inscription from the temple of Garni
east of Yerevan, which commemorates a ninthcentury Armenian Catholicos, also
named Mashtots; the first three signs are numerals (300 + 40 + 6), indicating his
deathyear to be 346 according to the Armenian calendar, or 897 ad.
Although a connection is sometimes asserted to exist between the Armenian
and Georgian alphabetic numerals, the evidence for this is too tenuous to suggest
any definite link. The primary similarities between the two are that they were used
in the same region and had distinct signs for 1000 through 9000. The only system
that is derived from the Armenian alphabetic numerals is the variant Armenian
system developed in the seventh century ad by Anania Shirakatsi. The Armenian
numerals did not spread beyond the limited area around Lake Van where Armenian was spoken, nor do they appear to have inspired the creation of any foreign
Alphabetic Systems
175
Figure 5.2. Grave inscription of the Armenian Catholicos Mashtots (897 ad); the first
three visible signs of the inscription are 300 + 40 + 6, the year of his death. Courtesy Gabriel
Kepeklian.
systems. After the development of the minuscule Armenian script, these signs
were also used numerically in the same manner.
Cipheredpositional numerals the Arabic system used by the neighboring
Seljuk Turks were first used in Armenia in the twelfth century (Shaw 193839:
368). Yet Armenian writers retained the alphabetic numerals for most ordinary
purposes long afterward. Only in the mid seventeenth century, when Armenia had
been firmly under Ottoman control for some time, did cipheredpositional numerals (Arabic, then later Western) replace the alphabetic system. Wingate (1930)
has published an unusual, undated, and unsolved magic square arithmetical
puzzle using both Arabic positional and Armenian alphabetic numerals, part of
a scroll contained within a Armenian family amulet designed to be worn upon
the person. The Armenian alphabetic system is still sometimes used for numbering chapters of the New Testament, although page and verse numbers are most
often written using Western numerals. Otherwise, the numerals used in modern
Armenia are the standard Western numerals.
Shirakatsis Notation
The Armenian astronomer, geographer, and mathematician Anania Shirakatsi13
was born ca. ad 595600 and was most likely a monk in the Armenian Church
(Hewsen 1968: 34). While littleknown today outside his native country, Shirakatsis contribution to Armenian learning is unparalleled, particularly his synthesis
13
Numerical Notation
176
1s
10
A
j
D
s
G
2
100
1000
of Persian, Arabic, Greek, and other scientific knowledge. In addition to these accomplishments, Shirakatsi developed a very interesting numerical notation system
in his Book of Arithmetic (Tuabantiwn), which is a collection of arithmetical
tables designed for the instruction of pupils. The basic form of this system uses
twelve signs, as shown in Table 5.19 (Shaw 193839: 270).
The individual signs are identical to those used for the appropriate numbers
in the traditional Armenian system. However, Shirakatsi showed how these signs
could be combined to express numbers through multiplication as well as addition.
In this system, a unitsign followed by one of the three powersigns (for 10, 100,
or 1000) indicates that the values of the two should be multiplied; these pairs of
signs were combined into numeralphrases through addition. Instead of writing
9642 as 0XMB (9000 + 600 + 40 + 2), as in the traditional Armenian alphabetic
numerals, Shirakatsi would write the same number as I2FSDjB (9 1000 + 6
100 + 4 10 + 2). Thus, where the traditional Armenian system is cipheredadditive, Shirakatsis system is multiplicativeadditive.
Any numeralphrase can be written more compactly with the traditional alphabetic numerals than with Shirakatsis variant so why would Shirakatsi advocate
its use? Firstly, it requires knowing fewer symbols (twelve versus thirtysix) in
order to express any number less than 10,000. More importantly, numbers greater
than 10,000 could be expressed using multiplicative combinations of two or three
signs. To do so, however, one needs the entire repertoire of Armenian numerals from 1 through 9000, as described earlier. For numbers from 10,000 through
90,000, Shirakatsi juxtaposed the signs for 10 through 90 with the sign for 1000.
Similarly, the numeralphrases for 100,000 through 900,000 combined the signs
for 100 through 900 with the sign for 1000. Alternatively, the hundred thousands
could be expressed using unitsigns followed by a 100sign and then a 1000sign.
Thus, one could write 460,000 as VO2 (400 + 60) 1000 or DSO2 ((4
100) + 60) 1000. This system is no longer a purely decimal system, but has a
mixed base of 10 and 1000. For values below 1000, it is purely multiplicativeadditive, but above 1000, the multiplicand that is juxtaposed with the sign for
1000 (2) is not a single sign, but rather a cipheredadditive numeralphrase. In
the Book of Arithmetic, numbers up to the ten millions are expressed relatively
compactly (Abgarian 1962: 46; Hewsen 1968: 42).
Alphabetic Systems
177
Shaw (193839: 369) believes that this system was not developed by Shirakatsi in
the seventh century, but was a commonly used variant system, of which Shirakatsis writings are the only surviving remnant. I do not believe there is any reason to
regard the system as anything other than the creation of Shirakatsi himself, since
its structure is never found in Greek, Syriac, Hebrew, or any other alphabetic system. Shirakatsi may have borrowed the notion of multiplicative structuring from
one of two foreign sources. The numerals developed by the fifth century ad Indian
mathematician ryabhata (Chapter 6) were multiplicativeadditive; it is possible that Shirakatsi, a mathematician with extensive knowledge of foreign writers,
knew of ryabhatas numerals and emulated them. Similarly, it is vaguely plausible
that Shirakatsi knew of the classical Chinese multiplicativeadditive numerical notation system (Chapter 8). Neither hypothesis has any direct evidentiary support.
Shirakatsis system is thus a structurally innovative local variant of the Armenian numerals designed to facilitate the representation of large numbers of the sort
that would be needed for his astronomical and mathematical calculations. There
is no evidence that his system was adopted by any later writers, or that it had any
effect on the development of other numerical notation systems throughout the
world. Instead, we should view this system as the creative invention of a single
individual, used only within his lifetime.
Georgian
Like the Armenians, the Georgians developed a script and numerical notation
system modeled after the Greek alphabet shortly after they converted to Christianity. While the creation of this first Georgian alphabet is often attributed in
folklore to King Parnavaz in the third century bc, there is no direct evidence of
Georgian writing until the fifth century ad, at which time the asomtavruli or majuscule script began to be used (Holisky 1996). More familiar to modern scholars,
however, are the mxedruli characters developed in the eleventh century ad, which
continue to be used to write the Georgian language today. The numerals associated with this script are shown in Table 5.20 (Holisky 1996: 366).
The system is decimal and cipheredadditive and, like the Georgian script,
is written from left to right. Like the Armenian script, the Georgian script had
enough letters to serve for all numerical values up to 9000. Some later inscriptions
even include a special sign for 10,000 (). There is no evidence that the Georgian
alphabetic numerals were ever used to express larger numbers than this, either
through multiplication or through additional signs. Presumably, such numbers
were always written out in full using lexical numerals.
There is an undeniable structural similarity between the Georgian and Armenian systems, which both, unlike the Greek alphabetic numerals, have enough
Numerical Notation
178
a
10s
j
100s
s
1000s 2
4808 = 5zh
1s
b
k
t
3
c
l
u
4
d
m
v
5
e
n
w
6
f
o
x
7
g
p
y
8
h
q
z
9
i
r
1
0
additional letters to represent the values from 1000 through 9000. However, while
the Georgian and Armenian scripts both use thirtysix signs for 1 through 9000,
the letterorder of the two scripts is vastly different. The Georgian letterorder
was modeled very closely on the Greek, with additional signs added as necessary
at the end of the series, while the uniquely Armenian signs in that script were interspersed randomly within the original Greek letterorder. It is unlikely that the
Georgian numerals would be modeled on the Armenian numerals but retain the
Greek letterorder for their values (Gamkrelidze 1994: 77). There may have been
some mutual influence between the two numerical notation systems, given certain
similarities in the signforms, but the direction of this influence remains unclear
(Gamkrelidze 1994: 8182). For now, the hypothesis of direct diffusion from the
Greek alphabetic numerals, while using distinctly Georgian signs and using distinct signs for the thousands, is most strongly supported.
The Georgian numerals were used in literary and religious texts throughout the
medieval period, particularly for pagination, dating, and stichometry, as well as in
monumental inscriptions. Their regular use ended in the sixteenth century, when
Georgia came under Ottoman control, after which Arabic positional numerals
were used for administrative and commercial purposes, although the alphabetic
numerals may have been retained for religious functions. However, Paolini and
Irbachs 1629 GeorgianItalian dictionary, the first book printed in Georgian, does
not contain any mention of the alphabetic numerals alongside its list of Georgian
letters. Since the eighteenth century, when Georgia fell under the Russian sphere
of influence, the Western numerals have been those normally used for all purposes
in written Georgian.
Glagolitic
The Glagolitic script was probably developed between 860 and 870 by the brothers Cyril and Methodius, who, while on a mission to the Moravian Slavs of what
is today modern Serbia, Croatia, and Macedonia, created an alphabet for liturgical
Alphabetic Systems
179
a
j
100s s
708 = yh
1s
10s
b
k
t
c
l
u
d
m
v
e
n
W
f
o
x
g
p
y
h
q
z
i
r
/
writings in the language now known as Old Church Slavonic (Schenker 1996:
166167). There may have been a preChristian script in the region, which might
explain why many of the Glagolitic letters have no correlation with the Greek
alphabet, but no preChristian numerical notation existed (Cubberley 1996). The
numeralsigns of the Glagolitic numerical notation system are shown in Table 5.21
(Vaillant 1948, Gardiner 1984).
As with the Greek and many other systems, Glagolitic numerals were frequently
distinguished from words in texts by placing dots to either side of a numeralphrase or by placing a mark of some sort above it (Vaillant 1948: 24; Schenker
1996: 182). In addition to these twentyseven signs, additional signs for 1000, 1,
and 2000, 2, were purportedly used in some texts. The system is cipheredadditive
and decimal, and is always written from left to right. However, for the numbers 11
through 19, the ordinary sign order is reversed (e.g., bj instead of jb for 12),
which reflects the Slavic lexical numerals for the teens (Schenker 1996: 182).
Yet none of the surviving Glagolitic manuscripts apparently use numerals higher than 1000 (Gardiner 1984: 15; Lunt 2001: 28). Schenker (1996: 182) contends
that the Glagolitic thousands were expressed by placing a small diagonal or curved
stroke (like the Greek hasta) to the left of a numeralsign to indicate that its value
should be multiplied by 1000; if so, Glagolitic is a hybrid multiplicativeadditive
system above 1000. Gamkrelidze (1994: 3940) and others contend that, because
the earliest Glagolitic script had thirtysix characters, the last nine letters of the
alphabet (most of which were later dropped from the script) originally had the
values 1000 through 9000. The issue remains unresolved and contentious.
The Greek alphabetic numerals were the sole and direct ancestor of the Glagolitic
numerals. The similarities in structure between the Greek and Glagolitic systems
are considerable, and Cyril, Methodius, and their followers were Greeks. Among
other possible ancestors, the Gothic numerals were long defunct by the ninth
century ad, and the Cyrillic numerals were not invented until later in the century.
Nickels (1973) suggestion that the Glagolitic numerals may have originated from
the tamgas (clan identifiers) used by Turks and Iranians in southern Russia as early
as the first century is problematic. While a variety of scripts, such as the Latin,
180
Numerical Notation
Greek, Samaritan, and Hebrew, may have been used as the model for one or more
of the letters of the alphabet, many other signs have no obvious correlates in other
scripts (Schenker 1996: 168172). Regardless, the Glagolitic letters must have been
assigned their numerical values under the influence of Greek Christianity.
Manuscripts were written in Glagolitic throughout the medieval period in the
region of modern Croatia, Serbia, Slovakia, and even into the Czech Republic
and Poland. Yet, even during the Middle Ages, Catholic or Westerninfluenced
areas began to prefer the Roman numerals to the Glagolitic, while areas under
Bulgarian or Serbian control tended to adopt the Cyrillic numerals and script. By
the fifteenth century, almost all the Slavs had adopted either Roman or Cyrillic
numerals. Only in Croatia, particularly along the Adriatic coast (Dalmatia), where
they were retained for the Croatian Roman Catholic liturgy (Cubberley 1996:
350), did the Glagolitic script and numerals flourish. They were also used in monumental inscriptions in Croatia from the eleventh century onward, a context not
seen elsewhere. Yet, even in Croatia, the Glagolitic script and numerals declined
greatly in use after the Ottoman conquests of the sixteenth century, and were used
only rarely from the seventeenth century onward (mostly in religious texts). It is
not clear whether the Glagolitic numerals survived as long as the Glagolitic script,
which persisted until the beginning of the twentieth century in the Quarner archipelago in northwestern Croatia.
Cyrillic
Like Glagolitic, the Cyrillic script was developed under the guidance of the missionaries Cyril and Methodius. It is quite likely that Cyrillic was developed in ad 890
900, after the deaths of Cyril and Methodius, by Cyrils followers and disciples in
Bulgaria, who then named the script after their deceased mentor. Cyrillic was originally used for writing the Old Church Slavonic language, but later was adopted for
writing a variety of Slavic languages, most notably Russian. An alphabetic numerical
notation system14 was developed around the same time. The Cyrillic numeralsigns
are shown in Table 5.22 (Gardiner 1984: 1617; Cubberley 1996: 348).
The system is cipheredadditive and decimal, and is normally written from left
to right. For the numbers 11 through 19, the ordinary sign order was often reversed
(e.g., bj instead of jb for 12), which reflects the structure of Slavic lexical numerals
(Vaillant 1948: 24). Numeralphrases were often distinguished from ordinary letters by placing a bar or other mark (titlo) above the phrase, and sometimes also by
14
While some scholars call this system Slavonic, I use the term Cyrillic to prevent
confusion with the Glagolitic system.
Alphabetic Systems
181
A
10s
J
100s S
708 = yh
1s
b
k
t
c
l
u
d
m
/
e
n
w
f
o
x
g
p
y
h
q
z
i
r
0
placing dots on either side of the signs (Lunt 2001: 28). Placing a small stroke
to the left of a number indicated that its value should be multiplied by 1000
(Vaillant 1948: 24; Schenker 1996: 182). The Cyrillic numerical notation system is
thus a hybrid: purely cipheredadditive below 1000 and multiplicativeadditive for
higher powers. In some cases, higher Cyrillic numerals were expressed by using the
signs preceded by an unusual sign, [, to indicate multiplication by 1000 (Gardiner
1984: 15). Apparently in some cases the multiplierstroke could be repeated two
or three times to indicate multiples of 1,000,000 and 1,000,000,000, respectively
(Berdnikov and Lapko 1999: 16). In the earliest phase of the scripts history, there
were a number of ideograms for powers of 10 from 10,000 up to 1,000,000,000,
but these rare notations are poorly studied, and the range of their use is unknown
(Berdnikov and Lapko 1999: 17).
While there are only twentyseven signs listed in Table 5.22, there are more than
twentyseven signs in all varieties of the Cyrillic script; modern Russian Cyrillic
uses thirtytwo letters, and earlier Cyrillic scripts used a number of older signs
that have now fallen into disuse. The signs that are assigned numerical values in
Cyrillic are those that are directly derived from Greek, including the otherwise
rarely used signs for xi (o), psi (y), and theta (i). Yet numerical values were
never assigned to the commonly used but nonGreek characters (Gardiner 1984:
1415). Thus, the Cyrillic numerical values do not correspond to the customary
order of letters, remaining faithful to the Greek order instead.
The Cyrillic script and numerals originated around 890 ad, at which time Slavs
and Greeks who had been influenced by Cyril and Methodius were extremely
active in the Christianization of the Slavs in the region of modern Bulgaria and
Serbia. That this missionary work was undertaken under the auspices of the Byzantine Empire confirms what is clear from the paleographic evidence that the
sole external influence on the Cyrillic script and numerals was the Greek uncial
alphabet used at the time. The nonGreek signs for additional consonantal Slavic
phonemes were never assigned numerical values, further confirming the Greek
origin of the Cyrillic numerical notation system.
182
Numerical Notation
The Cyrillic numerical notation system spread to Kievan Rus in the tenth century. The earliest printed books in Cyrillic script, those printed in Krakw from
1491 onward, were paginated using Cyrillic rather than Roman or Western numerals (Zimmer et al. 1983). The Balkans fell under Ottoman influence in the fifteenth
century, after the fall of Constantinople, and the alphabetic numerals largely ceased
to be used there by around 1500. In Russia, the Cyrillic numerals were used much
longer. The late sixteenthcentury English Slavist, Christopher Borough, authored
a Russian copy of the pseudoAristotelian Secret of Secrets (Bodleian MS Laud
misc. 45 (SC 500)), which was paginated in Western numerals but used Cyrillic
numerals and Arabic positional numerals elsewhere in the text (Pennington 1967:
681682). Western numerals were known in Russia in the seventeenth century; a
seventeenthcentury sundial from Mangazeia (in Siberia, near the Arctic coast)
is numbered using Western numerals, showing that they were known even in
the Russian hinterland (Ryan 1991: 375). Yet the popular Schitanie udobnoe ready
reckoner arithmetic text of 1682, intended for merchants, contained only Cyrillic
numerals (Okenfuss 1973: 329). For most purposes, the schety or Russian beadabacus, which had been used in Moscow since the eleventh century, was perfectly
adequate for the computational needs of prePetrine Russia (Ryan 1991: 373374;
Simonov 1993).
Not until the reforms of Peter the Great around 1700, and the introduction
of technical training in mathematics by scholars from Western Europe, were the
Western positional numerals introduced into Russia on a widespread basis (Hans
195960, Fedosov 1995). Magnitskiis Arifmetika of 1703 used both systems side by
side (Ryan 1991: 373). In the same year, logarithm tables published for the use of
students at the Moscow School of Mathematics and Navigation were printed in
Western numerals only. In 1710, however, Peter deferred to the clergy in decreeing
that church texts were to be printed using the traditional Cyrillic numerals (Cracraft 2003: 103). Throughout the eighteenth century, aristocrats and literate officials
would have needed to be familiar with Roman, Western, and Cyrillic numerals in
order to read the full range of textual genres used in Russia (Billington 1968: 209).
As late as 1918, Tsaritsa Alexandra (Alix of Hesse) was learning Cyrillic numerals
and using them to paginate her final diary, demonstrating that their use was still
relevant, if increasingly formal, in the late tsarist period (Kozlov and Khrustalev
1997: 23). Many texts were paginated using the older system even as Western
numerals increasingly were employed for a wider variety of functions. Unlike the
transition in Western Europe, where the Western numerals took centuries to be
adopted fully, the change from the Cyrillic to the Western numerals took place
relatively smoothly and rapidly, however. Today, the Cyrillic numerals are occasionally used in modern Church Slavonic texts (especially for numbering chapters and
verses in Bibles), but never in ordinary Cyrillic writing (Gasparov 2001: 1718).
Alphabetic Systems
183
Latin
From the fifth through the twelfth centuries, knowledge of the Greek exact sciences in Western Europe was relatively limited. Starting in the eleventh century,
as Arabic translations of Greek astronomical works began to be retranslated into
Latin, Western European scholars became increasingly aware of the Greek numerals. In a handful of texts, an attempt was made to convert the Greek alphabetic
numerals or Arabic abjad numerals for use with the Latin alphabet, as shown in
Table 5.23 (cf. Lemay 2000, Burnett 2000c).
The system is cipheredadditive and written from left to right, and simply employs twentytwo letters of the Latin alphabet as it existed at the time to represent
1 through 9, 10 through 90, and 100 through 400 (e.g., xlh = 328). Additionally,
in the Dresden Almagest of 1121, z, , and represent 500 through 700 (Burnett
2000c: 61). There is, however, no known way to write 800 and 900, and there is
no known multiplicative technique to express numbers higher than 1000. In one
of the texts containing such an alphabetic system (MS Cambrai Bibliothque Municipale 930, a copy of Hermann of Carinthias Astronomia), even the sign for 400
is not used (Lemay 2000: 378379). Yet, because the texts containing the Latin
alphabetical numerals are astronomical treatises, in which numbers higher than
the 360 degrees of the circle are rarely needed, this was not a serious weakness of
the system, which was perfectly adequate for such values. While minuscule letters
were used most of the time, in one text, majuscule letters for K, M, and N, and
occasionally for G, R, Q, and L were used, in order to prevent confusion with
similarappearing Western numerals (Burnett 2000a: 82).
Burnett (2000c: 6162) briefly discusses a few tenth to twelfthcentury astrolabes and manuscripts with Latin letters serving as numerals, but these are direct
transliterations of Arabic or Hebrew numerals that follow the Semitic alphabetic
order rather than the Latin one. The earliest documents containing the Latin alphabetic numerals in Table 5.23 date to 1121 and 1127 and were copied in Antioch
in the Crusader States, probably in association with the work of Stephen of Pisa,
an early twelfthcentury translator of Arabic scientific texts. These contain no key,
suggesting that the system would already have been understood by its intended
audience (Burnett 2000a: 76). The copy of Hermann of Carinthias Astronomia
similarly contains no key, and simply uses Latin alphabetic numerals to annotate
astronomical diagrams (Lemay 2000). A later manuscript (MS London, British
Library, Harley 5402), dated 1160 and written in a mixture of Latin and Italian,
does, however, have a key for translating Roman numerals into the Latin alphabetic system (Burnett 2000a: 76). All of these manuscripts use Roman numerals
in the text and for nonastronomical purposes such as column labels, and some of
them use positional numerals as well. The Liber Mamonis uses Roman numerals
Numerical Notation
184
1s
1
a
2
b
3
c
4
d
5
e
6
f
7
g
8
h
9
i
10s
100s
for single digits and some low compounds, Latin alphabetic numerals for many
two and threedigit compound numbers, and represents larger numbers using
Arabic positional numerals (not Western numerals) (Burnett 2000c: 6465).
The Latin alphabetic numerals represent a brief and abortive attempt to adapt
the Greek and Arabic systems to the Latin alphabet for the sole purpose of translating astronomical documents efficiently. They did not, however, lead to a consistent or longlived tradition of Latin alphabetic numeration. By the time the
Latin alphabetic system was developed, cipheredpositional numeration was already being widely adopted by Western European mathematicians and scholars,
rendering the alphabetic system obsolete. I know of no thirteenthcentury or
later documents that use the system described here. However, two late thirteenthcentury manuscripts probably written in Flanders use the letters A through I in
place of 1 through 9, in conjunction with 0, in a cipheredpositional alphabetic
system written from right to left (e.g., B0G = 702), parallel to but probably independent from ibn Ezras use of Hebrew alphabetic numerals in the same way,
discussed earlier (Burnett 2000b: 200; 2000c: 6263).
In some medieval Latin texts from Western Europe, a set of unusual lettersymbols were associated with numerical values in a sporadic and nonsystematic
way, of which one example is shown in Table 5.24 (Cappelli 1901: 413421).
In some respects, this notation resembles the cipheredadditive Latin notation.
These lettersymbols occur in very different contexts, however in the same texts
as Roman numerals rather than in translations of Greek or Arabic astronomical
texts. The only letters that were not assigned values in this system were I/J, U/V,
X, L, C, D, and M, for obvious reasons these already had numerical values in
the Roman numeral system. Like the corresponding Roman numerals, placing a
bar above any of these abbreviations indicated multiplication by 1000; e.g., =
11,000. Unlike the Latin alphabetic numerals or the Roman numerals, however,
these abbreviations could not be combined to form numeralphrases that is, one
Table 5.24. Medieval Latin numeral abbreviations
A
B
E
F G
H K
N O P
Q
R S T
Y
Z
500 300 250 40 400 200 151 90 11 400 500 80 70 160 150 2000
Alphabetic Systems
185
could not write HN for 290 and be understood. Some numbers are represented
twice (G and P both equal 400); many multiples of powers of 10 (30, 60, 600)
are not represented; and strange numbers such as 151 are assigned letters. In other
manuscripts (e.g., the tenthcentury De inventione litterarum in Strasbourg, Bibliothque Nationale et Universitaire MS. 326), the numerical associations of the
letters are completely different than the ones in Table 5.24 (Derolez 1954: 332335).
In fact, this form of notation is not related to either the Greek or Latin alphabetic
numerals, but instead is a complex but unstructured quasinumerical abbreviatory
system.
Summary
Alphabetic numerical notation systems originated with the Greeks in the sixth
century bc, who combined the structure of the Egyptian demotic system with
the idea of using phonetic signs as numeralsigns. The political and ecclesiastical authority of Greek speakers, coupled with the brevity and adaptability of
cipheredadditive numeration, led to the development of other alphabetic systems
using numeralsigns specific to each script. This phylogeny expanded tremendously
between the fourth and seventh centuries ad (the time of greatest Eastern Roman/
Byzantine power), with eight new systems arising during this period. Yet most
systems of this phylogeny had died out, or had at least been greatly reduced in the
contexts of their use, by the sixteenth century ad, during which time the Arabic
and Western positional numerals replaced them. Many alphabetic systems are still
used today, but only in limited contexts such as liturgical texts, numbered lists,
and divinatory magic.
There is no one feature common to all the alphabetic systems. Most are cipheredadditive and decimal, with or without the use of multiplicative structuring for the
higher powers. However, the Armenian notation of Shirakatsi is multiplicativeadditive and sometimes uses a base of 1000, while the astronomical fractions are
positional and involve a sexagesimal base. The number of signs used, the degree
to which multiplication is used, and the correspondence of numeralsigns with
scriptsigns are all variable.
One of the great advantages of alphabetic systems is that, if the signs are ordered using a local script, one need not learn both a set of scriptsigns and a set
of numeralsigns; one merely superimposes the decimal structure of the numerals
onto the script, thereby lessening the mnemonic burden on both new learners and
experienced users. The Greek alphabetic, Coptic, Gothic, Hebrew, Syriac, Armenian, Georgian, Glagolitic, and Latin systems all take advantage of this feature.
However, the values assigned to Arabic and Cyrillic letters do not correspond to
the customary letterorder, thus reducing this benefit. The Fez numerals and the
186
Numerical Notation
Alphabetic Systems
187
systems users, but in others, such as Hebrew and Armenian, the systems users
have largely been politically and culturally marginalized. That these systems could
survive in such circumstances and where, in many cases, cipheredpositional systems were available, requires explanation. Structural features may partly explain it:
alphabetic systems alphabeticity means that one need not learn a set of numerals in addition to a script, and cipheredadditive systems are more concise than
cipheredpositional systems for any Western numeralphrase containing zeroes,
the corresponding cipheredadditive numeralphrase will be shorter.
A more satisfying explanation, however, is that alphabetic numerical notation
systems, like scripts, can be important markers of cultural identity. In many cases
(e.g., Coptic, Gothic, Armenian, Georgian, Glagolitic, and Cyrillic), a group of
people developed a unique set of alphabetic numerals and developed their own
script at the same time. The point of alphabetic numerals is not to be comprehensible translinguistically, but rather for each system to serve for one script alone.
Under these circumstances, an alphabetic numeral system becomes an integral
part of a script, and thus marks ethnic identity. Even when these systems ceased to
be used regularly, many of them continued to be used in restricted functions, particularly in the domain of religion (e.g., Hebrew, Syriac, Coptic, Greek, Cyrillic).
Of course, it must not be forgotten that ultimately, in the face of massive globalization over the past five centuries, Western and Arabic positional numerals have
become the earliest and standard systems learned and used by almost everyone.
The future of alphabetic numerals seems likely to be one of increasing vestigiality
and obsolescence.
chapter 6
The South Asian numerical notation systems include all systems that derive from
the Brhm numerals used on the Indian subcontinent, including Western numerals. With the possible exception of China, South Asian numerical notation systems
are predominant throughout the entire world today. While most of the modern
South Asian systems are cipheredpositional, the earlier systems were cipheredadditive or multiplicativeadditive. An important evolutionary development was
the shift from cipheredadditive systems, such as the early Brhm numerals, to cipheredpositional systems. Despite attempts to postulate the origin of the important cipheredpositional structure elsewhere (Greece, China, or Mesopotamia),
this development came out of South Asia.
Brhm
The Brhm script came to prominence in the mid third century bc, during the
reign of the Mauryan emperor Aoka, although inscribed potsherds from the site
of Anuradhapura in Sri Lanka have been dated (controversially) to 400 bc (Coningham et al. 1996). Brhm script was probably derived from a Semitic prototype
(Aramaic, South Semitic, or Phoenician), although many South Asian scholars
still support the theory that the script was indigenously developed (Salomon 1996:
378379). Along with the slightly earlier Kharoh script used in the northwestern regions of India, it was the first script used in India after the collapse of the
188
189
1s
10s
100s
1000s
A
J
S
1
B
K
T
2
C
L
U
3
D
M
V
4
e
N
w
5
F
O
G
P
H
Q
I
R
Most Western epigraphers and archaeologists accept that, aside from the Harappan
script, there was no preMauryan writing in India (Salomon 1995, 1998).
190
Numerical Notation
Figure 6.1. Rubbing of an inscription from Nsik, Maharashtra state, dating to 120 ad,
bearing a variety of Brahmi numerals. Source: Senart 19056: 82.
from Nsik, as in Figure 6.1, dated 42 aka (ad 120) (Senart 19056: 82; Sircar 1964:
164167). The Vktaka grants (fifth century ad), one of the latest texts containing signs for the thousands, denote 8000 as x, a ligature of the signs for 1000 (y)
and 8 (z) that occur on the same grants (Datta and Singh 1962 [1935]: Tables IV,
IX). Despite paleographic changes, the basic structure of the Brhm numerals was
always cipheredadditive below 100 and multiplicativeadditive at both 100 and
1000. No special sign for 10,000 was used; 10,000 and 20,000 are written as
1000 10 and 1000 20 rather than 10,000 and 10,000 2. In the Nana Ghat
inscriptions, 24,400 is written as *%@ (1000 20 + 1000 4 + 100 4).
The Brhm numerals appear on some of the earliest Aokan inscriptions,
dating to the middle of the third century bc, but not in the early Sri Lankan
writings. These early inscriptions contain only a few signs (for 1, 2, 4, 6, 50, and
200), but already the hybrid cumulativeadditive/multiplicativeadditive structure
of the system was in place.2 Most of the signs are recognizably ancestral to later
ones, such as the more complete sets of numerals found at Nana Ghat and at
Nsik Cave. While there is no paleographic evidence of Brhm numerals prior to 300 bc, Datta and Singh (1962 [1935]: 37) claim that, because the Aokan
2
This opinion contradicts that of Guitel (1975: 605), who sees the Aoka numerals as being nonmultiplicative on the basis that the sign for 200 does not sufficiently resemble a
ligatured multiplicative 100 2.
191
inscriptions are found all over India, the Brhm system must have been developed
much earlier than the paleographic evidence would indicate, perhaps between
1000 and 600 bc. This is a spurious use of the discredited agearea method
(determining the age of features by their geographical distribution). In the early
Mauryan Empire, an enormous region was quickly encapsulated within a single
polity, so it is unsurprising that Mauryan administrative inscriptions are widely
distributed. While it was certainly plausible for nineteenthcentury Indologists to
hope to find earlier paleographic evidence for the numerals, such hopes now seem
remote. I agree with Salomon (1996, 1998) and many other Indologists that a midthirdcentury origin for the Brhm numerals and script is probable.
The question of the ultimate origin of the Brhm numerals specifically,
whether or not they constitute a case of independent invention, and if not, on
which ancestor(s) they were modeled is unresolved, and is made more complex
by the politicization of the matter. Previous scholars have emphasized the paleographic comparison of individual signs. I believe that the consideration of the
systems structural features and historical context of origin supplemented by
paleography, where appropriate will be a more fruitful approach.
One set of theories regarding the origin of the Brhm numerals derives them from
existing representational systems used in South Asia. Borrowing from the letters of
the Brhm script to create an alphabetic numeralsystem, while once a popular
theory, is not really sustainable (Prinsep 1838, Woepcke 1863, Indraji 1876, Datta and
Singh 1962 [1935], Gokhale 1966, Verma 1971). While a few Brhm numeralsigns
resemble phonetic signs, if one accepts certain paleographic transformations, these
correspond neither to the standard Brhm letterorder nor acrophonically to any
languages lexical numerals. Renou and Filliozat (1953) note that in texts containing
both purported letternumerals and the corresponding signs used phonetically, the
two varieties are quite different. The derivation of the Brhm numerals from the
Kharoh letters is even more improbable, and has not been seriously proposed for
over a century (Bayley 1882). Ifrah (1998) proposes but discards the theory that the
Brhm numerals derive from the Kharoh numerals, which can easily be dismissed
by noting the temporal priority of the former. Finally, a more recent set of theories
derives the Brhm numerals from those of the Indus Valley civilization (Sen 1971,
Kak 1994), but there are no examples of any writing from India between the latest
Harappan inscriptions (around 1700 bc) and the first Brhm inscriptions (around
250 bc), and only limited and conflicting evidence for the nature of the Harappan
numerical notation system (Parpola 1994; cf. Chapter 10).
If not derived from any South Asian system, the Brhm numerals could have
developed independently. Woodruff (1994 [1909]: 5360) speculated that both the
Chinese and Brhm numerals derived from a hypothetical ancient set of cumulative tally signs for 1 through 9, which would then have spread to both China and
192
Numerical Notation
India. Kaye (1919) argued that the Brhm numerals developed independently during Aokas time, with their structural features representing three different stages
of development, but inexplicably then argues against Indian creativity. Ifrah (1998:
390), arguing that there are universal constants caused by the fundamental rules
of history and paleography, postulates nonattested cumulative signs for 1 through
9 which later became abbreviated and ligatured into the Brhm system. Salomons
position is more sensibly agnostic; recognizing the problems involved with many
other theories, he simply notes that numerical signs are sometimes cursive reductions of collocations of counting strokes, citing the hieratic and demotic systems
as examples (Salomon 1998: 60).
Finally, a number of theories argue for a foreign origin of the Brhm numerals.
Falk (1993: 175176), noting structural and paleographic resemblances between
the Brhm and the earliest Chinese (Chapter 8) numerical notations, argues for
a Chinese origin. However, there is little evidence of contact between the two
regions at this period, and the only paleographic similarity between the systems
is the common use of horizontal strokes for 1, 2, and 3. It has occasionally been
proposed that the Greek alphabetic numerals inspired the Brhm numerals, given
their appearance following the Alexandrine period, the strong trade ties with the
GrecoIranian kingdoms of Parthia and Bactria, and the structural similarities
between the two systems. However, the evidence for the alphabeticity of the
Brhm numerals is weak at best (see the previous discussion), and there is no
paleographic correspondence between the Greek and Brhm numerals.
It is most plausible that the Brhm numerals are derived from the Egyptian hieratic or demotic numerals. Burnell (1968 [1874]) argued for a demotic origin, while
Bhlers (1963 [1895]) much more prominent analysis argued for a hieratic origin.
The three systems are structurally similar: they are all decimal, hybrid cipheredadditive/multiplicativeadditive systems, and represent 200, 300, 2000, and 3000
by adding quasimultiplicative strokes to the signs for 100 or 1000. There are resemblances in around onethird of the signforms, and very close resemblances for a few,
such as 9 (hieratic = i; demotic = i; early Brhm = i) (Bhler 1963 [1895]: 115119;
Salomon 1995, 1998). While there was not tremendous EgyptoIndic cultural contact, Ptolemaic traders reached as far as the city of Muziris (modern Cranganore) on
the Malabar Coast, and Aoka is known to have sent Buddhist missionaries to Alexandria (Basham 1980: 187). Of the two Egyptian systems, I believe the demotic to be
a more likely ancestor, because in the Ptolemaic period the use of hieratic numerals
was very limited. Thus, although the demotic and Brhm systems differ in both the
power at which multiplication is used and the direction of writing, I believe that a
demotic origin should be adopted as a working hypothesis.
The Brhm numerals spread throughout the Indian subcontinent during the
Mauryan period. Only in the northwest, where Kharoh numerals predominated,
193
did the Brhm numerals fail to penetrate until around the fourth century ad.
They were used primarily for writing dates on stone inscriptions and copper land
grants. Thus, a full set of numeralsigns up to at least 1000 is attested, and the
numeralsigns can be assigned exact dates. Other functions for which Brhm
numerals were used include stichometry and the recording of financial transactions. While it is interesting to speculate on the use of Brhm numerals on other
materials than stone and copper, the Indian climate and geography are unsuitable for the survival of perishable materials. In Central Asia, manuscripts in the
Tocharian language were written using a variant of Brhm script and numerals
from the sixth to ninth centuries, and in this drier climate, Brhm numerals are
attested to have been used on wood tablets, palm leaves, and paper (Sander 1968).
There is no surviving evidence that the Brhm numerals were used for arithmetic or accounting; the rather substantial body of medieval Indian mathematical
works, sometimes attributed to very early dates, use either lexical numerals only,
or employ one of the alphasyllabic notations to be described here.
After the Kharoh script died out in the fourth century ad, variants of the
Brhm numerals were the only ones used in India until the late sixth or early
seventh century. They spread not only throughout the Indian subcontinent, but
also into Central and Southeast Asia, regions that were heavily influenced by India
during this period. In some Central Asian manuscripts, numeralphrases were
written from top to bottom rather than from left to right (Renou and Filliozat
1953: 702). There was enormous variation in the shapes of the numeralsigns from
location to location, suggesting that, as with the Indian scripts, no pressure existed
toward the formation of interregional standards. The primary regional division,
between northern and southern systems, began as early as the second century ad,
and these two basic variants diverged further in later centuries.
The end of the traditional Brhm numerals and the later local additive variants
was a gradual process, instigated not by external influences but by the invention of
cipheredpositional notation beginning in the late sixth or early seventh century
ad. Over the next couple of centuries, the older cipheredadditive forms became
increasingly rare, and by the ninth century ad, the Brhm numerals had been replaced by the modern cipheredpositional system throughout India and Southeast
Asia. Only on the southern tip of India and in Sri Lanka were additive systems
retained (though in an altered form) until significantly later.
194
Numerical Notation
While it would be teleological to portray the history of numerals in a linear fashion leading to our own system, the present nearuniversality of the cipheredpositional, decimal structure originating in India and spreading westward throughout the Islamic and Western spheres requires explanation.
Around ad 600, a change began in the writing of dates in the Brhmderived scripts of India and Southeast Asia. Instead of writing smaller numbers with
cipheredadditive notation and larger numbers with multiplicativeadditive notation, all numbers were written using paleographic variations of the nine Brhm
numeralsigns and a dot to indicate zero a purely cipheredpositional system.
The spread of the older additive systems throughout South and Southeast Asia
between the third century bc and the seventh century ad was followed by a second wave of diffusion of the positional principle and zero (seventh century ad
onward), wherein the additive systems changed into positional ones. The change
is actually a very simple one. The unitsigns were retained, but the powersigns
were replaced with a single sign for zero. This process is confirmed by comparing
the signs for 1 through 9 in the nonpositional (Brhmderived) systems with the
very similar unitsigns used with a zero in later periods.
This was not the first zero; the Babylonians (Chapter 7) and the Maya (Chapter 9)
had already invented the positional principle and zerosigns well before this time.
Some scholars have claimed that the Babylonian zero diffused eastward to India
just as it diffused westward to Greece (Fvrier 1948: 585; Menninger 1969: 3989).
Yet the Babylonian and Maya systems were both cumulativepositional, and used
a subbase in addition to a base (Babylonian 10 and 60, Maya 5 and 20). Similarly,
while Greek and Arab astronomers used positional fractions with a zero, these
had a sexagesimal (Babylonianderived) base and never used positional notation
for whole numbers. Thus, there is no apparent historical relation between these
systems and the later Indian one. The Indian positional system first combined
ciphering, a zero, and a single, decimal base.
It is frequently claimed that the earliest example of cipheredpositional numerals is found on the Sankheda or Mankani copper plate bearing the date 346 in the
Kalacuri era, which translates to 595 ad (Bhler 1896: 78; Smith and Karpinski
1911: 46; Das 1927a: 118; Kak 1990: 199). This plate is a donation charter of Dadda
III, used to certify a land grant. When discussing any land grant, the issue of a
later forgery always arises, as attempts to claim land by producing such evidence
were common in India (as elsewhere). Since much of the paleographic evidence
for early positional numerals comes from such land grants, we must be cautious
to avoid dating inscriptions simply by the date as inscribed, and also take paleography and historical context into account. We must also remember that texts containing positional notation that are transcriptions or translations of earlier works
must not be assigned an early date based simply on their putative earlier authors.
195
While we need not go so far as Kaye (1919: 346), who claimed that all positional
numerals in India prior to the ninth century ad were forgeries, most Indologists
are very wary of the Sankheda plate (Salomon 1998: 61). The fact that it is ninety
years older than any other positional numeralphrase suggests that we need to
question carefully any sixthcentury ad evidence for cipheredpositional numerals
in India.
The earliest surviving and unquestionable examples of cipheredpositional
numerals with a zero derive, not from India itself, but from Southeast Asia, in
Khmer, Old Malay, and Cham inscriptions from the late seventh century ad. A
calendrical inscription found at Sambor (in Cambodia) and written in a mixture
of Old Khmer and Sanskrit is dated 605 in the aka dating system, or 683 ad; the
zero appears as a small dot (Coeds 1931: 327). As the Sambor inscription is a calendrical passage rather than a land grant or financial document, it is unlikely to be
a forgery (Diller 1996: 126). Similar inscriptions from the Old Malay kingdom of
Sriwijaya have been found at Palembang and at Kotakapur on the nearby island of
Bangka dating to 683, 684, and 686 ad, or 605, 606, and 608 aka, respectively; in
these cases, the zero was written as a circle rather than as a closed dot (Diller 1995).
It is intriguing that the Old Khmer and Old Malay inscriptions appear in the same
year (Diller 1995: 66). Nevertheless, there is no reason to believe, as Kaye (1907)
did, that the existence of these inscriptions must mean that cipheredpositional
numerals actually originated in Southeast Asia and diffused from there to India.
The existence of intermediate additivepositional forms from the sixth and early
seventh centuries ad, which I will discuss later, coupled with the probability that
some of the earlier copper grant plates are authentic, make it likely that the invention of cipheredpositional numerals occurred around ad 600.
However, there is compelling evidence for something akin to the positional
principle and zero in certain earlier literary and religious texts. In texts using the
bhutasa khya or wordnumeral system, special cryptic words for one through nine
could be combined in a sort of positional fashion with a word for zero in order to
represent dates lexically in a fashion quite different from that of ordinary Sanskrit
numberwords (Datta and Singh 1962 [1935]: 5363; Mukherjee 1977). The earliest attested text to use this system is the Yavanajtaka of Sphujidhava (ad 269),
a Sanskrit version of a Greek astronomical text (Yano 2006: 15). In it, the words
moon and earth mean one, eye and twin two, and so on, with sky and dot
meaning zero, suggesting a sort of positionality of words if not of graphemes,
while also allowing multiple words to represent the number, a key to producing
numeralphrases that are unambiguous and yet flexible enough to fit into strict
Sanskrit verse (Plofker 2007: 395). The bhutasa khya system is thus suggestive
of positionality but does not constitute a system of graphic numerical signs, nor
should its use be taken to imply the widespread use of decimal positional numerals
196
Numerical Notation
in Indian manuscripts. Notably, the regular name for Indian numeral symbols,
anka mark, is also a numerical word for nine in this system, which suggests
that there were originally only nine numerical graphemes that is, that the zero,
and placevalue, were latecomers to Indian numerical notation even though they
existed conceptually in the wordnumerals (Clark 1929: 229230).
The earliest Sanksrit word for the zerosign, nyabindu (literally, voiddot),
is first used in Subhandus Vasavadatta, written around 600 ad (Sen 1971: 175;
Salomon 1998: 63). This evidence suggests a correspondence between the early use
of numeralwords and the structurally identical later use of the ten numeralsigns.
Thus, the literary concept of a zerospace in Hindu thought, the use of chronograms, and the term nyabindu in the fifth and sixth centuries ad may have prefigured the eventual development of cipheredpositional numerals. Within the interlinked tradition of Indian religious and mathematical thought, the invention of
the zero is as much a metaphysical concept as it is an arithmetical tool, if not more
so. Almost all the attested early cipheredpositional numerals are nonarithmetical,
and are simply used to register dates and other numbers on inscriptions and copper plates. It would be erroneous to assume that the positional numerals originally
had an arithmetical function, and then to use this assumption to hypothesize an
ancient mathematical tradition of cipheredpositional numerals.
There is no evidence that true cipheredpositional numerals were used prior
to the middle of the sixth century ad, and any claim prior to the middle of the
seventh century ad requires careful examination. The Indian climate and topography are not particularly suitable for the survival of materials other than stone
and metal, and we certainly do not have as much evidence as we would like.
Nevertheless, there is plenty of inscriptional evidence for continued use of the old
additive numerals from the sixth through the eighth centuries ad, they decline
significantly in frequency only in the ninth century. It is unlikely that all the evidence for cipheredpositional numerals was lost where so much survives for the
additive system though not impossible, if positional numerals were used only
on perishable media at first.
Further evidence for the chronology of the shift comes from inscriptions dating from the late sixth to the middle of the eighth century ad from the Orissa
region, which are written with unusual mixed structures combining the features
of the older additive and newer positional notations (Datta and Singh 1962 [1935];
Acharya 1993; Salomon 1998: 6263). The earliest of these is from the Urlam copper plates of the Eastern Gang king Hastivarman, dated to 578 ad, in which the
Gang era year 80 is written as the additive sign for 80 followed by a zero, but
this date is questionable (Salomon 1998: 62). Acharya (1993) describes many Orissan inscriptions dating from 635 to 690 ad in which dates such as 137 are written
as 100 3 7 rather than 137. This series of dates leads directly into the first fully
197
I am uncertain what to make of Mukherjees (1993) assertion that the copperplate inscription of Devakhadga expresses the date 73 in the Harsha era (starting 606 ad) using
positional numerals, which would thus be dated to 678 ad.
Numerical Notation
198
Bengali
Devanagari
Gujarati
Marathi
Oriya
Punjabi
Nepali
to present the paleographic data concerning the development of Indian numeralsigns from 800 ad to the present day (cf. Salomon 1998; Ifrah 1998: 367385 for
more complete analyses of this issue). Nevertheless, a look at some of the more
important variations on this common pattern of cipheredpositional decimal systems is warranted. Many of these systems (or very close descendants thereof ) have
been employed for well over one thousand years and continue to be used. Most
major South Asian languages have their own alphasyllabaries and numeralsigns.
Their numerical notation systems are structurally identical to one another and to
Western numerals. Today, all these indigenous systems are in competition with
Western numerals, especially for commercial and scientific purposes. In religious
and formal contexts, the traditional numerals are still frequently preferred.
North India
The ancestor of the northern Indian numerical notation systems is the Brhm
system used in the Gupta Empire, which ruled most of northern India from the
Indus to the Ganges from the fourth to sixth centuries ad, and also significantly
influenced central India. These earliest Gupta Brhm numerals were nonpositional, but the idea of positionality and the zero sign spread quickly through the
systems of the region. The most common modern varieties of this subgroup are
the Bengali, Devanagari, Gujarati, Marathi, Oriya, Nepali, and Punjabi; they are
thus used in a swath across Pakistan, northern and central India, Nepal, and Bangladesh. The northern Indian systems are also directly ancestral to both the modern Arabic and Maghribi numerals associated with the Arabic script, and thus,
indirectly, to Western numerals. The similarities between Western numeralsigns
and many of the north Indian numerals, especially those for 0, 2, and 3, are quite
evident in Table 6.2.
199
Tibetan
Mongolian
Central Asia
The Gupta script also gave rise to a small number of scripts in the Himalayas and
Central Asia, of which the most important are the Mongolian and Tibetan. The
Tocharian script had used a variant of the Brhm additive numerals from the
sixth to eighth centuries ad, but the Tocharian language and script died out before the introduction of positionality. Tibetan writing and numeration developed
in the ninth century, and Mongolian numerals developed from Tibetan in the
thirteenth century.4 These systems are related to the northern Indian systems. The
classical Mongolian numerals were usually written from top to bottom in vertical
columns, but the forms listed here are those used when they were written from left
to right. These systems are shown in Table 6.3.
South India
The scripts of the southern half of the Indian peninsula diverged from those of the
north as early as the second century ad. There are five modern scripts in this subgroup: Telugu and Kannada, two closely related scripts of east central India, along
with Tamil, Malayalam, and Sinhalese. Of these five, the Telugu and Kannada numerals shifted from addition to position in the seventh and eighth centuries ad,
and are thus structurally identical to those of the northern systems (Syamalamma
1992: 51). The numerals of the Grantha script (sixth to twelfth centuries), which is
ancestral to modern Tamil, Malayalam, and Sinhalese, did not switch principle.
The Telugu and Kannada numerals are shown in Table 6.4, while the other three
systems are described later in this chapter.
Southeast Asia
Far from being a cultural backwater or simple recipient of positional notation,
South Asian societies used cipheredpositional numerals very early. Scripts such
4
Despite Ifrahs assertion (1998: 382) that each of the Agnean, Kutchean, and Khotanese
scripts of Chinese Turkestan would have used a set of ten positional numerals, I know of
no evidence that this was the case.
Numerical Notation
200
Telugu
Kannada
as Kawi (the ancient script of Java) and Cham (used in Vietnam until the thirteenth century) originally used hybrid additive numerical notation systems on the
Brhm model, but these transformed into cipheredadditive positional systems
in the seventh century. The modern descendants of these systems include Khmer,
Thai, Burmese, Lao, Balinese, and Javanese. Of these, Balinese and Javanese are
closely related to one another but paleographically distant from any other South
Asian systems. They use Javanese letters to represent certain numbers, while retaining older signs derived from Kawi for the others (0, 4, 5, and 6). The Southeast
Asian systems are shown in Table 6.5.
Tamil
The Tamil script used in the far southeast of India and parts of Sri Lanka is derived
ultimately from Bhattiprolu, the southern variety of the Brhm script that developed in the first or second century ad, and immediately from the Grantha script
of the sixth through twelfth centuries, which is ancestral to Tamil, Malayalam,
and Sinhalese. The Tamil script is alphasyllabic and similar to other Brhmbased
scripts, but has unique features, such as the ability to represent consonant clusters as
a sequence of individual consonant signs. Similarly, the Tamil numerical notation
system is rather different from those of other Brhmderived scripts. The Tamil
numeralsigns are shown in Table 6.6 (Pihan 1860: 113119; Guitel 1975: 614615).
Table 6.5. Southeast Asian numerical notation systems
Script
Khmer
Thai
Burmese
Lao
Balinese
Javanese
201
A
B C
10
100
J
6408 = FLDKH
Units
D
K
G
L
1000
The numeralsigns are ultimately derived from Brhm, and are thus related
to all the systems of India and Southeast Asia. Following the Indian pattern,
numeralphrases are written and read from left to right, but are structurally neither
cipheredadditive, like those of the Brhm numerals, nor cipheredpositional, like
those of most of the Indian systems. Rather, the traditional Tamil system is multiplicativeadditive and decimal. There is no power multiplier for the ones. Tamil
has no signs for 10,000 or higher powers of 10; large numbers were expressed by
placing an appropriate numeralphrase before the sign for 1000, then multiplying.
Thus, 800,000 would be written as HKL (8 100 1000). There is no ambiguity
in this phrases meaning, because phrases are always read strictly from left to right.5
This is the only instance where a lower power sign may precede a higher one.
The Tamil numerals acquired their distinct structure in the medieval era, although
it is not clear exactly when the divergence arose. The change from a hybrid cipheredadditive/multiplicativeadditive to a purely multiplicativeadditive structure is easily accomplished; because Brhm numerals are multiplicative above 100, all that
is required is that the nine individual signs for the decades 10 through 90 be
replaced by a single sign for 10. At this early period, we know them largely from
inscriptions on stone, although we cannot exclude the possibility that they were
used in other contexts. The numeralsigns are derived from those of the Grantha
script, and are closely related to others of southern India. It has sometimes been
claimed that the Tamil numerals are a uniquely Dravidian invention using letters
of the alphabet, and, indeed, there are resemblances between the numeralsigns
for 1 through 9 and nine Tamil phonetic signs (Burnell 1968 [1874]: 68). Nevertheless, since these resemblances can be found only by comparing the modern paleographic forms of the numbers and letters, this argument cannot be offered as a
theory of their origin. Rather, the similarity is probably due to a later assimilation
of the numeralsigns to the phonetic signs.
Only Tamil and Malayalam, of all the South Asian systems, altered the Brhm
cipheredadditive/multiplicativeadditive system to a purely multiplicativeadditive
5
Curiously, this system is structurally identical to the Armenian alphabetic notation of Anania
Shirakatsi (Chapter 5), but it would be an error to make too much of this resemblance.
202
Numerical Notation
one. (Sinhalese retained the Brhm structure, while all other Brhmderived systems became cipheredpositional.) The Chinese classical numerals (Chapter 8)
are multiplicativeadditive, so contact with Chinese Buddhists might have stimulated the development of the unique notations characteristic of areas of southern
India that were also Buddhist, or, more likely, made their retention more appealing. However, there is no paleographic similarity between the Chinese and
Tamil numerals, making diffusion from China improbable. Moreover, none of
this evidence explains why the Tamils retained their system even after they ceased
to practice Buddhism. Another possibility is that the users of the Tamil and
Malayalam systems did not switch to positional notation as an effort to maintain
their cultural distinctness from the north.
Late in the systems history, an abbreviated form of the Tamil numerals developed that, for some numbers, adds an element of positional notation by omitting
the powersigns for 10, 100, and 1000. For instance, Pihan notes that while 21
was traditionally written BJA (2 10 + 1), it could also be written BA, abbreviating the phrase without any loss of information (Pihan 1860: 117). Such
numeralphrases are purely cipheredpositional. While this presents no problem
for numerals that lack any empty positions, a zero sign is needed in other cases;
however, no zero appears in any pretwentiethcentury Tamil numeralphrases.
Sometimes, rather than using a sign for zero, Tamil writers used the powersigns
for 10, 100, and 1000 to eliminate ambiguity. In one case, 2205 is written as BBKE
(2, 2, 100, 5), which indicates that the second 2 is to be understood as a hundreds
value rather than as a tens value, and that therefore the first 2 must be understood
as 2000 (Guitel 1975: 614615). Such phrases combine multiplicativeadditive and
cipheredpositional notation. These mixed multiplicative and positional phrases
are no longer used, and appear to be a transitional product of the colonial period, when contact with the West began in earnest. Apparently, in the nineteenth
century some Tamil astronomers were computing using a mixed decimal and sexagesimal computation system by manipulating pebbles or shells on a flat surface
in a manner reminiscent of Greek or Babylonian techniques, but there was no
corresponding numerical notation (Neugebauer 1952). Today, some formal Tamil
writings use the traditional numerals, while for most commercial and informal
purposes an ordinary zero sign is used, making the system cipheredpositional.
Most literate Tamils today are familiar with and use the Western numerals.
Malayalam
The Malayalam script, like Tamil, is derived from the Grantha script of southern
India. It is used to write the Dravidian language of the same name used in Kerala
at Indias southwestern tip. It first emerged as a distinct script around 700 ad,
203
M N
10
W
6408 = RYPXT
Units
P
X
S
Y
100
1000
although its letters and numeralsigns are closely related to those of the other
Brhmderived scripts. The Malayalam numerical notation system, like Tamil
and Sinhalese, was unaffected by the cultural and political influence that rendered the northern Indian numerical notation systems structurally identical and
paleographically similar. The traditional Malayalam numeralsigns are indicated
in Table 6.7 (Pihan 1860: 122125; Ifrah 1998: 335).
The similarities between the Tamil and Malayalam systems are striking. Both
systems are decimal and multiplicativeadditive, and written from left to right.
There are many paleographic similarities between the numeralsigns of the two
systems, thus refuting the claim that the Tamil numerals are phonetic in origin. As
in Tamil, there is no powersign for the units, and the 1 is understood in any numeralphrase with a units value. Furthermore, Malayalam numbers above 10,000
are also expressed using multiplicative combinations of the sign for 1000 with
those for 10 and 100, as necessary. None of these similarities is particularly surprising, given the close cultural and geographic proximity of these two Dravidian
peoples. The only structural difference between Tamil and Malayalam numeration
is that Malayalam numeralphrases were never expressed using the hybrid additive and positional notation that was occasionally used later in the Tamil systems
history.
There are few distinctly Malayalam inscriptions that date before ad 1000,
by which time it had already acquired its multiplicativeadditive structure. The
numeralsigns are derived from those of the Grantha script, showing that the Malayalam numerals are native to South Asia, although, as with Tamil, we cannot exclude
the possibility of some influence from Buddhist China. The fact that Buddhism was
maintained longer in the south than in northern India may be a partial explanation for the difference in structure. A millennium of trade with and domination
by other peoples of South Asia, most of whom used cipheredpositional notation,
did not affect the structure of the Malayalam system. Malayalam writers employed
this system regularly until the middle of the nineteenth century, at which time
European contact introduced the zero and the idea of positionality. A new sign
for zero was introduced (V), which, when combined with the nine regular unitsigns, produced a regular cipheredpositional system. Today, the older Malayalam
Numerical Notation
204
a
b c
10s j
k
l
100 s
1000 t
3684 = ctfsqd
1s
d
m
e
n
f
o
g
p
h
q
i
r
system is used rarely if at all (largely by those who need to understand old texts),
and is quickly becoming a historical curiosity.
Sinhalese
The Sinhalese (or Singhalese) script developed from the model of the southern
Brhm scripts for use among the speakers of the Sinhalese language, and was
influenced by the Grantha script that is ancestral to the Tamil and Malayalam
scripts used for writing the Dravidian languages of southern India and northern
Sri Lanka (although Sinhalese is an IndoEuropean language). It is an alphasyllabary, written from left to right, and is used today in Sri Lanka and the Maldives.
The traditional Sinhalese numeralsigns are indicated in Table 6.8 (Pihan 1860:
140141; Ifrah 1998: 332).
Sinhalese has distinct signs for each of the units, each of the decades, 100, and
1000. The numeralsigns for the units resemble many of those used in other South
Asian numerical notation systems, and many of the signs also resemble, but are
not derived from, the phonetic signs of the Sinhalese script. Numeralphrases
are written from left to right. The system is cipheredadditive below 100, but it
is multiplicative for the hundreds and thousands, combining the unitsigns with
the appropriate powersigns. It is unknown how 10,000 and higher numbers were
written, though Pihan (1860: 141) speculates that it may have been through multiplicative forms such as those used in Tamil and Malayalam.
The Sinhalese numerals are thus structurally identical to the old Brhm system. The Sinhalese did not adopt the positional system when the peoples of India (except Tamil and Malayalam speakers) and Southeast Asia did, between the
seventh and ninth centuries ad. Sinhalese inscriptions and texts used this system
throughout the medieval period, seemingly unaffected by the shift to positionality occurring in other Indian numerical notation systems. Pihan (1860) shows no
awareness of any structural changes in the Sinhalese numerals in use at the time
205
he was writing; although his knowledge of the numerals was limited, there is no
reason to believe that they were in significant decline in the mid nineteenth century. Modern Sinhalese writings normally use the Western numerals, although the
traditional numerals are retained for certain formal purposes.
Indian Alphasyllabic
The primary numerical notation systems of India were cipheredadditive before
the seventh century ad and cipheredpositional afterward, with only a few systems (Tamil, Sinhalese, Malayalam) remaining additive after that point. The numeralsigns of these systems are abstract and do not resemble closely the letters
of the Brhm script or its descendants. However, starting around 500 ad, Indian
astronomers and astrologers began to use a very different principle for representing numbers: assigning numerical values to the phonetic signs of various Indian
alphasyllabic scripts. The basic principle of the Indian alphasyllabaries is that a set
of consonantsigns are combined with a set of diacritic marks that indicate vowels
to produce a set of signs for CV syllables; unmarked consonantsigns denote the
syllable with the inherent vowel a.6 These numerical notation systems, known
collectively as varnasankhya systems, were considered to be distinct from the normal Indian systems that had abstract numeralsigns (Ifrah 1998: 483). The three
systems that I will now discuss ryabhatas numerals, katapaydi numerals, and
aksharapall numerals, represent an important side branch of the South Asian systems. Although used only by a limited group of initiates, they are very important
for understanding Indian astronomy, astrology, poetry, and numerology, as well as
serving important mnemonic functions by linking words and numbers.
ryabhatas Numerals
While it is sometimes claimed that the Indian grammarian Pnini used alphasyllabic numerals in the seventh century bc (Datta and Singh 1962 [1935]), this is a
highly dubious proposition given the lack of attested writing in India at that time.
The first attested alphasyllabic numerals appear about 510 ad, near the end of the
dominance of the Guptas over India. The system was very probably invented by
the mathematician and astronomer ryabhata, in whose works (later named the
ryabhatiya by his disciples) it first appears. ryabhata, who lived in Kusumapura
in modern Bihar, was renowned among Indian scholars of the Gupta empire and
later centuries, was known to Muslim scholars as Arjabhad, and was likely the figure
6
The number of consonantsigns and vowel diacritics varies from script to script, and
there are also signs for V syllables (isolated vowels) and CCV syllables.
Numerical Notation
206
&
%
&
G
ka
1
kha
2
ga
3
gha
4
a
5
ki
100
khi
200
gi
300
ghi
400
i
500
<
<
E
;
D
A
ca
6
cha
7
ja
8
jha
9
a
10
ci
600
chi
700
ji
800
jhi
900
i
1000
?
@
4
6
B
wa
11
wha
12
sa
13
sha
14
ta
15
wi
1100
whi
1200
si
1300
shi
1400
ti
1500
8
5
K
C
3
ta
16
tha
17
da
18
dha
19
na
20
ti
1600
thi
1700
di
1800
dhi
1900
ni
2000
!
F
pa
21
pha
22
ba
23
bha
24
ma
25
pi
2100
phi
2200
bi
2300
bhi
2400
mi
2500
'
'
7
#
ya
30
ra
40
la
50
va
60
yi
3000
ri
4000
li
5000
vi
6000
(
H
a
70
va
80
sa
90
ha
100
i
7000
vi
8000
si
9000
hi
10,000
207
ka
100 (1)
ki
102
ku
104
kri
106
kli
108
ke
1010
kai
1012
ko
1014
kau
1016
the vowel attached to one of the basic signs alters its numerical value. When combined with the vowel i, the signs take on the numerical values 1002500, 30009000,
and 10,000, as shown in the rightmost five columns. While this means that there are
two signs with the value 100 ha () and ki () this has little potential to cause
confusion. Each successive vowel diacritic multiplies the value of the sign by 100 with
respect to its predecessor, as shown in Table 6.10 (indicating only the combinations
of k + vowels). Using these signs in combination, any number up to 1018 could be
expressed, and ryabhatas system does not exhaust the available diacritics.
Numeralphrases were written with the lowest powers on the left, which reflects
the order of powers of the Sanskrit lexical numerals, but which is opposed to the
Indian numerical notation systems, in which the highest power was on the left. No
sign for zero was needed, and none used. The signs for 11 through 19 and 21 through
25 were not strictly necessary; 15 could be written as GA instead of B without any
ambiguity, but the latter was more concise. These extraneous signs normally were
not used in numbers such as 85, which was written as G (5 + 80) rather than B(
(15 + 70). In some cases, these rules were violated (we do not know why), so that in
one astronomical table, 106 is written as 16 + 90 and 37 as 16 + 21 (Guitel 1975: 587).
Table 6.11 indicates several numeralphrases written alphasyllabically.
Table 6.11. Alphasyllabic numeralphrases
Alphasyllabic
Representation Transcription and Signvalues
kha
va
62
%
2
60
ta
ha
ta
OR
116
8or8 16
100
16
ra
chu
70,040
7E
40
70,000
kha
ya
ghi
li
765,432
%'&#<( 2
30
400
5000
ma
vi
nu
jri
98,206,025
3;H
25
6000 200,000 8,000,000
phu
ghe
40,000,220,000 &
220,000 40,000,000,000
Value
ki
100
cu
u
60,000
700,000
sri
90,000,000
208
Numerical Notation
The best way to conceive of this system is as a base100, or centesimal, multiplicativeadditive system with a decimal and ciphered subbase. Unlike most multiplicativeadditive systems, however, there can be up to two unitsigns within each
power of 100, each of which combines with its own powersign. For instance, in
the representation of 9800 in Table 6.11, the signs for 8 (;) and 90 (H) each combine separately with the diacritic sign for 100 ( ). The system is slightly irregular below 100 in that the basic thirtythree signs include signs for 11 through 25. ryabhata
recognized that the system was centesimal rather than decimal, as he distinguished
the set of centesimal powers, or varga classes, from the intermediate powers, or
avarga (Das 1927a: 110; Jha 1988: 80). ryabhatas system is not positional, since
placing an unmodified consonantsign in the middle of a numeralphrase would
render it meaningless.
If ryabhata was unfamiliar with positional numeration, then he may have developed alphasyllabic numerals because of the insufficiency of the Brhm cipheredadditive system for writing large numbers, a task that could be done very concisely
using his own system. However, there is no evidence that the calculations that
ryabhata undertook were done directly with these numerals. Instead, their function was primarily to efficiently express very large numbers in verse, in the manner in which traditional Indian mathematics was written (Clark 1929: 232233).
The versification of mathematical expressions, while seemingly constraining and
complicating the expression of numbers, assisted the user mnemonically while
retaining the brevity of other numerical systems.
It is possible that ryabhata was familiar with the Greek alphabetic numerals in
addition to the Brhm numerals. ryabhatas work was inspired in part by Greek
astronomical writings, and Fleet (1911a), among others, has argued that both ryabhatas astronomy and his numerals are derived from Greek sources. However,
even if he borrowed the general idea of using scriptsigns as numerals from the
Greeks and there is no definite evidence either way this does not tell us very
much, because the two systems are radically different (Das 1927a: 111114). Even
if he did know the Greek numerals, they played little role in the invention of his
own system, whose idiosyncratic features, such as a base of 100, are not found in
other systems. Another possibility is that even though his system was not positional and lacked a sign for zero, ryabhata had a complete knowledge of cipheredpositional numeration when inventing his alphasyllabic system (Ganguli 1927). As
I discussed earlier, the concept of nya or emptiness existed in the fifth century
ad and may have prefigured the use of positional numerals in India, but no good
evidence survives for an actual cipheredpositional numeration system prior to the
seventh century ad, long after ryabhatas death. Ifrahs (1998: 450451) statement
that the use and adoption of ryabhatas system caused the Indian discoveries of
the placevalue system and zero, which took place before ryabhatas time, to be
209
irretrievably lost to history is typical of the confused antiempiricism of recent research on this system. The nonpositionality and relative complexity of ryabhatas
system argue against his having been familiar with positionality.
While ryabhatas numerals were known to Indian astronomers and mathematicians long after his death, they were used solely in the context of commentaries
on his work. The systems lack of facility for arithmetic, coupled with the difficulty
in pronouncing the resulting numeralphrases in Sanskrit, led to its abandonment
even among most of ryabhatas proponents (Jha 1988: 8586; Yano 2006: 149).
It was replaced, in part, by the regular numerical notation systems of India, but it
also gave rise to a variety of successor systems for correlating phonetic signs with
numerical values, most notably the katapaydi system. While these successors were
not as unusual as ryabhatas system, they were far more successful.
Katapaydi Numerals
When later scholars experimented with alphasyllabic numeration starting in the
ninth century ad, they immediately saw that an alphasyllabary could also be
turned into a cipheredpositional system. Known as katapaydi, the signs of this
system are shown in Table 6.12 (Fleet 1911b; Datta and Singh 1962 [1935]: 70).
In this system, each V and CV syllable is given a value from 0 to 9. Unlike
ryabhatas system, changing the vowel of the syllable does not change its numerical value, so that ka = ki = ku = 1. Two of the signs (a and na) take on the
value of zero, as did isolated vowelsigns (those representing a V syllable alone,
without any consonantal component), which did not have a numerical value in
ryabhatas system. CCV syllables do not have their own numerical values, but
are considered to have the value of the consonant to the left of the vowel, so that
tva = va = 4 and ntya = ya = 1. As a result, any sequence of syllables can be assigned a numerical value, and any number has a wide variety of possible phonetic
transcriptions. Katapaydi numerals were read with the lowest power on the left,
as in ryabhatas numerals. Thus, the word bhavati or F 8 is had the numerical
value 644. The name katapaydi itself is taken from the four syllables (ka, ta, pa, ya)
that are assigned the value 1 in this system. Although it is unusual in that each
digit from 0 to 9 has several alphasyllabic values that represent it, structurally this
system is an ordinary cipheredpositional and decimal system.
The earliest example of the katapaydi numerals is from the Grahacranibandhana
by the astronomer Haridatta, written in ad 683 (Sarma 1999: 274). Datta and Singh
(1962 [1935]: 71) place its invention around ad 500 and claim that it must have
been known to ryabhata himself, but there is no textual evidence to support this
assertion. Haridatta was a direct intellectual descendant of ryabhata, and used
his predecessors system as the basis for his own. Nevertheless, as the katapaydi
Numerical Notation
210
&
ka
1
kha
2
ga
3
gha
4
ua
5
<
ca
6
cha
7
ja
8
jha
9
a
0
wa
1
wha
2
sa
3
sha
4
ta
5
ta
6
tha
7
da
8
dha
9
na
0
pa
1
pha
2
ba
3
bha
4
ma
5
'
ya
1
ra
2
la
3
va
4
a
5
va
6
sa
7
ha
8
211
Several variants of the katapaydi developed, most of which changed a few numerical values or eliminated the values of certain categories of signs, such as the
isolated vowels (Datta and Singh 1962 [1935]: 7172). Some of these systems were
unique to one writer, while others were used in specific regions over a longer
period. Renou and Filliozat (1953: 708) claim that their use in paginating loose
manuscripts served a cryptographic function in that the pages of the text, once
jumbled, could be placed back in order only by initiates of the system. Katapaydi
numeration survived much more extensively in southern India, particularly in the
province of Kerala curiously, where the additive Tamil and Malayalam numerals were only recently and incompletely replaced by positional systems. Northern
Indian katapaydi numerals are rare, although Sarma (1999) discusses a sixteenthor seventeenthcentury astrolabe labeled in the northern Devanagari script that
uses them, perhaps in imitation of the Arabic abjad numerals. They were still used
in astrological manuscripts and horoscopes in South India even in the late nineteenth century (Burnell 1968 [1874]: 7980).
Aksharapall
A third variety of alphasyllabic numerals, sometimes confused with the cipheredpositional katapaydi in the scholarly literature, is known as aksharapall numeration (after akshara, the word for the CV syllableclusters that comprise the basic
unit of the Indian alphasyllabaries). Whereas ryabhatas system was multiplicativeadditive, and the katapaydi system was cipheredpositional, the aksharapall
systems are cipheredadditive and decimal, assigning the numerical values 19,
1090, and sometimes also the low hundreds (but never as high as 1000, to my
knowledge) to a set of phonetic signs. It was used very widely for paginating
books, and was written in the margins from top to bottom with the highest
power at the top.
Unlike the first two alphasyllabic numerical notation systems, there was never
a single regular system for correlating signs with numerical values in the aksharapall. Datta and Singhs (1962 [1935]: 73) search through old manuscripts revealed
no fewer than three signs for 1, twelve different signs for 4, and nine signs for 60.
This is less complex than it seems, because within specific traditions, there were set
sequences of signs that would be understood by anyone working within them. In
some instances, parts of these sequences may be comprehensible; for instance, in
Nepali manuscripts from the eleventh to the fourteenth centuries, the numbers 1
through 3 were represented by the syllables e, dvi, and tri, which correspond to the
Nepali lexical numerals (Burnell 1968 [1874]: 66). In other cases, the signs used appear to have been assigned almost randomly. Datta and Singh (1962 [1935]: 73) list
many signs for which they cannot even attach a plausible syllabic value. Dialect
212
Numerical Notation
213
alphabet had their own alphabetic numerical traditions. Thus, while the Indian
tradition of cipheredpositional numerals spread fairly readily into the Arab world,
alphasyllabic numeration remained a strictly South Asian phenomenon.
Arabic Positional
The Arabic script is written from right to left, and is basically consonantal, though
with some representation of vowel sounds. The earliest Arabic speakers used the
hybrid cumulativeadditive/multiplicativeadditive Nabataean numerals (Chapter 3);
this was replaced after the Islamic conquest of Greekspeaking regions by an
alphabetic system (the abjad numerals) akin to the Greek, Hebrew, and Syriac
systems (Chapter 5). Shortly after the introduction of abjad numerals, however,
users of the Arabic script became aware of Indian cipheredpositional numerals,
and developed their own system on this basis, whose modern numeralsigns are
shown in Table 6.13.
This system is written with the higher powers on the left, rather than from right
to left following the direction of the Arabic script. Thus, 26,049 would be written
. Burnett (2000b) discusses this phenomenon in light of the fact that
Western numerals are read from left to right but in this case in accordance with
the direction of the script, noting that this may have led to some confusion when
the Arabic numerals were borrowed into the West as to the direction in which the
numerals should be read. In the modern Arabic numeralsigns, there are alternate
signs used for 4 ( ) and 5 ( ), and the modern zero sign is written with a dot instead of with a circle because the circle was already assigned the value of five.
There are undeniable paleographic resemblances between the Arabic positional
numeralsigns and those used in medieval north India. Table 6.14 compares the
Arabic positional numerals found in eleventhcentury mathematical and astronomical treatises with the inscription found at Gwalior, India, dated to 876 ad,
containing the Nagari numerals used in medieval India. These signs are very similar, and it is thus safe to assume that the Arabic numerals have an Indian origin.
In some cases, as for 2, 3, 7, 8, and 9, the Nagari numeralsign became rotated or
inverted, which may have resulted from the scribal practice of writing from top
to bottom, then rotating the manuscript to read it (Ifrah 1998: 532533). The fact
that medieval and modern Arabic scholars are unanimous in attributing an Indian
origin to these signs, and call them isb alhindi (Indian numerals), abundantly
confirms the paleographic evidence.
The social context of the transmission appears to have been limited to the exact
sciences initially, specifically to astronomy. In 662 ad, a Syrian Christian bishop,
Severus Sebokht, noted the Hindu proficiency in astronomy, commenting that
as for their skilful methods of calculation and their computing which belies
Numerical Notation
214
description, they use only nine figures (Nau 1910). The meaning of this statement is unclear, as he does not mention the zero, but it is likely that Sebokht was
referring to cipheredpositional numerals; if so, the Arabspeaking world probably
would have had some such knowledge as well, possibly through Persian intermediaries (Kunitzsch 2003: 3). In ad 773, an Indian embassy visited the court of the
Abbasid caliph alMansur in Baghdad, among whose members was an astronomer
who brought with him a copy of a Hindu work of astronomy, which was translated into Arabic (Folkerts 2001: 15).
Within fifty years of this episode, the mathematician Muhammad ibn Ms
alKhwrizm wrote his Arithmetic (c. 825 ad) using cipheredpositional numerals
extensively, prompting later mathematicians and astronomers to follow his lead in
replacing the old cipheredadditive abjad numerals with the new positional system.
While alKhwrizms work does not survive in its original Arabic (the earliest surviving manuscript is a twelfthcentury Latin translation), alKhwrizm knew of
positional numeration and advocated its simplicity and functionality. We do not
know specifically, however, what numeralsigns he used, and, indeed, almost no
contemporary texts containing positional numerals survive. The earliest direct paleographic evidence for positional Arabic numerals comes from an Egyptian papyrus
(PERF 789) with a numeral 260 at the bottom, but only if that numeral is the date
260 a.h., or 873/4 ad, a point that Kunitzsch (2003: 5) disputes. The Trkh of alYaqb of 889 mentions the sign for zero as being a small circle, without describing
the system further (Kunitzsch 2003: 4). The manuscript MS Paris, BNF ar. 2457 by
the astronomer alSijzi, written between 969 and 972 ad, provides secure evidence
for the mid tenth century, but this is a considerable gap in our direct evidence
from the numeralforms (Folkerts 2001: 14; Kunitzsch 2003: 56). Nevertheless, the
textual nonpaleographic evidence demonstrates that some Arabs were surely using
them at least by alKhwrizms time, and possibly as early as ad 775.
Table 6.14. Early Arabic and Nagari positional numerals
0
Arabic
Nagari
215
From their origins in the late eighth and early ninth centuries ad, the numerals spread throughout the Islamic world, though not without resistance or
confusion. Many conservative scribes and bookkeepers resisted the new numerals in favor of older calculation on the fingers and with numeralwords.
In his Kitb almuallimn, the ninthcentury scholar alJhiz recommended
finger reckoning above the isb alhindi because it needed neither speech nor
writing, a position echoed a century later by the historian alSl in his Adab
alkuttb (Kunitzsch 2003: 45). Whether this functional explanation is complete, or whether other, ideological considerations came into play, is unknown.
Lemay (1977: 440444) questions the extent to which the Indian numerals
were known to the Arabs before the tenth century, and shows that there was
confusion among some Arabic thinkers over how they worked. While positional numerals began to dominate in both mathematical and nonmathematical
contexts starting in the eleventh century, astrological texts remained far more
conservative, retaining the abjad numerals solely until the fourteenth century
(Lemay 1982: 385386).
Contrary to the diffusion of most numerical notation systems, scientific functions rather than commerce or religion provided a significant impetus for the
transmission of the positional numerals from India westward to the Arabs. It is
nonetheless generally the case that medieval Arabic arithmetic did not distinguish
commercial, astronomical, divinatory, and other arithmetical practices unambiguously. The Arabs borrowed not only the Indian numerals, but also a host of computational techniques and devices, including the dustboard, a flat tablet strewn
with sand into which figures could be written for undertaking computations (Bag
1990: 290293; Folkerts 2001: 14). Other techniques available included a complex
Greekderived system of finger reckoning and the use of counters or shells; accordingly, the use of written penandpaper arithmetic was apparently not part of the
initial practice of Indianderived numeration.
The earliest major Arabic arithmetical text to advocate Indian numeration
instead was the Kitb alfusl f alhisb alhind of alUqldis the Euclidean,
written in Damascus in ad 952/3, the earliest extant copy of which dates to ad
1186 (Saidan 1966; Burnett 2006: 16). Yet alUqldis recommended concealing the
Indian origin of the technique, using the first nine letters of the Greek alphabet
or the Arabic abjad instead of the numeralsigns for 1 through 9 (Saidan 1966:
478479). Many of the earliest Arabic arithmetic texts had names such as kitb
altakht book of the board, confirming the association of the new numerals with
dustboard computation (Smith and Mourad 1927). Yet once the isb alhindi
had been firmly established by the eleventh and twelfth centuries, Arabic mathematical texts began to advocate doing computations with paper and ink instead
of the dustboard, which of necessity involved rubbing out intermediate steps of
216
Numerical Notation
computations and thus could lead to error. Clearly, Arab mathematicians sentiments toward the innovation were complex both recognizing it as highly useful
in comparison to earlier techniques, yet desiring to improve it to suit their own
purposes and prejudices.
Despite their prominence, the Arabic positional numerals have given rise to few
structurally distinct descendants. Some of the cryptographic systems of the
Ottoman Empire (Chapter 10) may have been inspired by the Arabic positional
system, although most of these systems are additive and may be more closely related to the abjad numerals. A number of subSaharan African cipheredpositional systems developed in the twentieth century (Chapter 10) were created by
inventors who knew the Arabic and/or Western positional numerals.
While the debt of Western numerals to Arabic positional notation is undeniable, they are directly related to the cipheredpositional Maghribi numerals
(known sometimes as ghubar numerals, and in medieval Latin as figure toletane
Toledan figures) described in the following section, and are thus at best phylogenetic cousins of the signs commonly used in the modern Islamic world.
Burnett (2000c) notes, however, that routes of transmission were complex, and
the standard Arabic numerals were described in several Western European texts,
through the medium of the Crusader state at Antioch in the twelfthcentury Latin
translation Liber Mamonis, and through the city of Pisa in the Latin translation of
the Hebrew works of Abraham ibn Ezra (see Chapter 5). These Eastern figures
(known in Latin as figure indice Indian figures) were, if never widespread, at
least known to Western mathematicians, but by the early thirteenth century, the
migration of Toledan translators to northern Italy marked the decline of Indian
figures in Western Europe (Burnett 2002a).
The Arabic numerals enjoy a degree of currency and use in the modern world
second only to the Western numerals. They are used regularly in most contexts
throughout all regions that employ the Arabic script, and are thus found regularly
from Morocco to Indonesia. While global commerce and the effects of mass media have introduced Western numerals into the Arabicspeaking world, and most
literate users of the Arabic script know them, it is unlikely that this will have any
longterm effect on the use of Arabic numerals.
217
Wc
Y e Z
[g
\ h
Following Kunitzsch (2003: 10), who argues persuasively that the term ghubar has
been misapplied, I use the label Maghribi, reflecting the systems geographical origin.
218
Numerical Notation
395; Miller 1933; Lattin 1933: 184185). Yet there is no evidence for cipheredpositional numerals in Byzantine Greece prior to the twelfth century (Wilson 1981).
The rejectionist view is further refuted by the fact that medieval Arabs, Western
Europeans, and Byzantines were in accord that the numerals were of Indian origin
(Lemay 1982: 382). It is best now regarded as an ethnocentric relic of the misguided notion that Indians and Arabs were uncreative.
The term ghubar, with its unusual meaning of dust or sand, has prompted
some comment as to the function of the number. Das (1927b: 358) and Gandz (1931)
assert that the ghubar tradition represented a sort of Arabic abacus, but Kunitzsch
(2003) shows that texts discussing the system refer to a takht board on which writings were made and from which items could be erased that is, boards covered
with dust or sand that were used as calculating boards by drawing figures on them.
Some of the variation between the Arabic positional and Maghribi numerals may
be explained by their use on differing media, the former in permanent media and
the latter for arithmetical calculations on sandboards. Their forms, thus fixed by
the separation of contexts, might have become entrenched through centuries of use
in disparate parts of the Islamic world. However, no actual dustnumerals in that
medium survive, of course. Kunitzsch (2003: 910) argues that despite the terms
isb alhindi and isb alghubar being references to different media and computational techniques, they do not necessarily imply two distinct sets of graphemes
associated with each. While the term ghubar seems to have originated in Tunisia
and is associated with Maghribi scholars such as ibn Khaldun (whose fourteenthcentury Muqaddimah mentions only ghubar, abjad, and zimm [Coptic] numerals, not hindi) (Lemay 1982: 384387), the distinction in signforms is thus better
understood as a geographical rather than a functional one.
In fact, the ordinary Arabic numerals and the Maghribi numerals were quite
similar until the twelfth century; their numeralsigns for almost all values are similar enough to be explained as graphic variations of a common system of Indian
derivation (the medieval Nagari cipheredpositional system). In a tenthcentury
manuscript written by the Persian astronomer Sijzi, the form of numerals used is
intermediate between the Arabic and Maghribi forms (Mazaheri 1974). Maghribi
numerals are thus a subset of the larger class of Indianderived positional, decimal
numerals (Lemay 1977: 437), both of which stand in contrast to the abjad numerals described in Chapter 5. This does not mean that the transmission of positional
numeration was a singular event, but it does suggest that it was not a matter of
two distinct waves of diffusion, one into the Maghreb and the other into the rest
of the Arab world, but rather a significantly more complex series of episodes that
resulted in two parallel systems.
While the Maghribi numerals began as a paleographic variant of the Indian
numerals, they eventually took on a distinct cultural meaning among the scribes,
219
astronomers, and mathematicians of the western Islamic world. This is partly because of the relative independence of polities such as the caliphate of Cordoba
from the Baghdadbased Abbasid caliphate. The traditionalism of the Maghrebi
and Andalusians may partly explain why the numerals persisted even after the
rest of the Islamic world had adopted the signs now used throughout the modern
world (Ifrah 1998: 539). They were still regularly used in Spain and North Africa
in the fifteenth and sixteenth centuries, and sporadically thereafter (Labarta and
Barcel 1988).
The Maghribi numerals would be little more than a paleographic curiosity, merely
one of many ways of writing Arabic numerals, if not for the fact that through them,
Western Europe adopted cipheredpositional numerals. For the past seventyfive
years, all major scholars have agreed that the resemblances between Maghribi and
Western numerals, coupled with the first appearance of the latter system in medieval Spain, demonstrate this origin. Yet even as the Western numerals developed
along their own trajectory within the Christian European context, the Maghribi
numerals survived for almost a full millennium. Ifrah (1998: 535) provides examples
of arithmetical texts written using the system from as late as the eighteenth century,
and suggests that the system may have survived into the nineteenth century before
being completely replaced by the standard Arabic numerals.
Western Numerals
From their origin as a foreign and suspicious novelty during the medieval period,
the ten Western numerals, structured by the use of the positional principle, have
become so familiar that it is easy for the nonspecialist to forget that there are
other numerical notation systems. The ubiquity and universality of the Western
numerals make understanding their origin and diffusion all the more important.
Unfortunately, no monograph has dealt systematically with the topic since Hill
(1915), whose work is rather outdated as a result of advances in paleography.
The first example of Western numerals is generally held to be the Codex Vigilanus, written in 976 in the monastery of Albelda near the town of Logroo in
northern Spain, in which the numerals are described (in Latin) as Indian figures
(Hill 1915: 29; Burnett 2002b: 241). The nine units are listed, in descending order,
but no zerosign, probably because the signs were intended for use with a counting
board. These signs are shown in Table 6.16.
These figures are very similar to the Maghribi numerals shown in Table 6.11,
and in fact there is no reason to consider them a separate system, except that they
are used in a Latin and Christian text from northern Spain rather than in an Arabic one from Andalusia. Toledo was a major center for the transmission of Arabic
knowledge to the Christian West in the tenth and eleventh centuries, and Lemay
Numerical Notation
220
A remaining puzzle is why the Roman numerals for 1 through 9 were never employed on
the counters; although Burnett (1997: 11) argues that this would have been impractical, I
do not see any reason why even a long numeralphrase like VIIII for 9 would have been
too long to have been used in this fashion.
221
The spread of Western numerals into the tradition of manuscript writing (in
both mathematical and other texts) began in earnest in 1202, at which time the
mathematician Leonardo of Pisa, better known as Fibonacci, promoted their use
in his Liber Abaci (Book of the Abacus). Despite its name, the purpose of Fibonaccis text was to promote not the use of the abacus, but rather the use of written
numerals for computation, with nine unitsigns and a zerosign (k), following
Arabic practices advocated by scholars such as alUqldis and alKhwrizm. Later
thirteenthcentury texts following in Fibonaccis wake, such as Alexander de Villa
Deis Carmen de algorismo and John de Sacroboscos Algorismus vulgaris, used the
term algorismus, a corruption of the name alKhwrizm, to refer to this new art
(Burnett 2006: 19). While he did not use the term himself, Fibonacci was thus the
forerunner of the algorithmists who, in direct conflict with the abacists, promoted
the use of written numerals for computation rather than the use of counting
boards (cf. Evans 1977, Murray 1978). This technique is the precursor to modern
computational techniques with pen and paper.
Despite the unquestionable importance of Gerbert, Fibonacci, and other mathematicians in introducing cipheredpositional numerals to the West and promoting their use, their eventual adoption is not a vindication of a great man theory
of history. Among the names given to the zerosign in a late twelfthcentury Latin
manuscript (Cambridge, Trinity College R.15.16, Fol. Av) was chimaera, suggesting that it was assimilated only with difficulty into the conceptual system of
medieval mathematical thought. Western numerals were not initially given the
same conceptual status as letters of the alphabet, or even as Roman numerals.
They were, instead, seen as characteres, signs to be made on physical artifacts, and
only gradually assimilated into texts as written signs (Burnett 2006: 29). The diffusion of the Western numerals from Andalusia and North Africa to the West occurred slowly, numerous times and by several different routes, some of which were
more fruitful than others (Gibson and Newton 1995: 316). In fact, Greek and Italian mathematical manuscripts contain the standard Arabic Eastern positional
numerals, not ghubarderived ones, through the twelfth century (Burnett 2002b).
A late twelfthcentury Latin manuscript from Bavaria contrasts the Toledan
figures and the Indian figures (Burnett 2002b: 241). Only in the thirteenth
century, when many of the important Toledan astronomers moved to northern
Italy, were the standard Arabic forms fully abandoned. Contact between the Arab
and Western cultural spheres followed several paths in the Middle Ages: through
Spain, to be sure, but also through Norman Sicily, along main trade routes from
African cities such as Tunis and Tripoli to Venice and Genoa, through the Crusader states such as Antioch, and through Byzantine Arabs (Burnett 2006).
Far from being an instantaneous adoption, then, the Western numerals were
used only by a small number of Western European scholars in the Middle Ages.
222
Numerical Notation
The ordinary populace of Western Europe used Roman numerals, if any, while
Eastern Orthodox regions used alphabetic systems such as the Greek or Cyrillic
alphabetic numerals. I have already discussed the various transitional and blended
versions of Western and Roman numerals used from the twelfth through seventeenth centuries, and the various medieval prohibitions enacted against the
use of Western numerals in Florence, Padua, and Frankfurt (Chapter 4). Most
notable among these are the Visigothic Roman/Western blended systems of
medieval Spain, where the forms of the Roman numerals affected the writing
of similarappearing Western numerals (Lemay 1982). Also as mentioned earlier,
mixed GreekWestern and HebrewWestern numerical structures were used in
some late medieval mathematical and astronomical documents (Chapter 5). In
astronomical texts, numbers up to 360 (i.e., degrees of the circle) continued to
be written in alphabetic numerals (usually Greek, sometimes Hebrew or Arabic),
while higher numbers were written in Western numerals (Burnett 2006: 20).
Whether we regard these blends as hidebound efforts to retain fragments of an
older notation or as progressive attempts to innovate, they illustrate the often
haphazard manner in which the Western numerals came to be introduced into
European scholarly circles.
Table 6.17 demonstrates the slow transmission of Western numerals throughout Europe, including both their first occurrence in each region and the period
in which they became more commonly known. In general, Latin and scholarly (particularly mathematical and astronomical) uses of the numerals preceded
their vernacular and commercial use by several centuries (Murray 1978: 193194).
Most of the earliest examples of the numerals in any given region are found in
mathematical treatises and texts designed specifically to explain the new numerals.
Where Western numerals were used in nonscientific contexts, they often served
cryptographic or secretive functions. One of the earliest such instances comes
from the legal documents of a notary from Perugia dating from 1184 to 1206,
in which the numerals indicated the numbers of lines of documents (Burnett
2006: 20). Similarly, the early thirteenthcentury Genoese notary Lanfranco used
Western numerals only in the margins of his private documents to make records
of payments made to him by clients, while retaining Roman numerals otherwise
(Krueger 1977).9 Western numerals were used, however, as assembly marks on
timbers of the roof of Salisbury Cathedral in the 1220s; we do not know why this
was done, and there is no obvious reason why they would have been preferred over
Roman numerals (TattonBrown and Miles 2003). Only when the audience for
9
The quasialphabetic Fez numerals used by notaries in the Maghrib (Chapter 5) served
similar functions, but in that case the nonpositional system was the obscure one and the
positional system the more commonplace.
223
Italy
France
England
Germany/
Austria
Greece
Scandinavia
Iceland
Portugal
Russia
Common Use
1490:a dating pages in texts
c. 1325: banking records and
account books in major cities
(Struik 1968; Menninger
1969: 428)
c. 1400: dating, accounting, etc.
15251550: archival records,
accounting books (Jenkinson
1926)
16001630: probate inventories
(Wardley and White 2003)
15251550 (Smith and
Karpinski 1911: 133)
c. 1400: Ottoman conquest of
most Greekspeaking areas
c. 1550 (books, manuscripts,
records)
c. 1550 (books, manuscripts,
records)
14901510: travelogues, scientific documents (Barrados de
Carvalho 1957: 125)
16th17th century: general use
(de Oliveira Marques 1996)
early 18th century (reforms of
Peter the Great)
Arabic documents from Spain used the ghubar numerals extensively from the tenth century ad onward; this date refers only to their common use in Christian Spain.
the numerals expanded from monks and astronomers did Western numerals begin
to replace Roman numerals more generally, however.
The common use of Western numerals in Europe was surely aided by the
transmission of doubleentry bookkeeping from Italy in the fourteenth and fifteenth centuries; this technique greatly benefits from cipheredpositional numerical notation (Jenkinson 1926: 267). Yet the Italian merchant Francesco Datini
(c. 13351410), an early adopter of doubleentry bookkeeping, switched his
224
Numerical Notation
A copper coin of Norman Sicily dated to 533 ah (1138 ad) is the earliest positionally
dated coin in Europe, but it is inscribed and dated using the Arabic script and numerals
(Hill 1915: 16; Menninger 1969: 439).
225
Curiously, the sixteenth century also marked the development of the attribution Arabic
to the Western numerals, whereas earlier they had always been seen (correctly) as an
Indian invention (Clark 1929: 217). It is possible that this occurred due to growing
awareness of Arabic learning among early modern scholars.
226
Numerical Notation
Summary
We end in the modern era with a host of local numerical notation systems and two
(the Western and Arabic) that spread enormously on the heels of political conquests, but we should not forget the origins of these systems in the South Asian regional tradition originating with Brhm. The common feature of the South Asian
systems is the set of nine Brhm unitsigns which persist, though greatly altered,
in the surviving numerical notation systems and a decimal structure. Only the
alphasyllabic systems use distinct signs, the letters of the Indian alphasyllabaries.
It is common practice to end studies of numerical notation with the Western
numerals (e.g., Guitel 1975, Ifrah 1998). Yet to do so portrays the spread of Western numerals throughout the world as the inevitable replacement of worse with
better systems, in continuous progress from primitive beginnings to the perfection
of the Western decimal positional system, an achievement that can never be surpassed. Most surviving modern systems are cipheredpositional, which indicates
that they are useful, but this does not demonstrate that they are the inevitable conclusion of a teleological historical process. If technology truly spread only through
the diffusion of what is functional and the replacement of what is not, one would
expect that structurally identical systems should expand with equal rapidity and
geographical reach. The South Asian phylogeny, with so many decimal cipheredpositional systems surviving and in use, provides a good testing ground for this
227
theory. Because only the Arabic and Western numerals (and, to a lesser extent,
Nagari) have spread extensively, their diffusion must be due mainly to sociopolitical factors. Furthermore, the geographic distribution of the surviving nonpositional systems is no less than that of many positional ones. Why would the additive Tamil system be as widespread as, say, the Khmer system, if functionality is of
supreme importance? Do we really expect Tibetan cipheredpositional numerals
to survive and Chinese multiplicativeadditive ones to decline?
I do not mean to suggest that functionality has nothing to do with the spread
of numerical notation systems, especially ones such as the Arabic and Western numerals that have been used extensively for accounting, arithmetic, and mathematics. Yet to proclaim the Western numerals spread as the triumph of functionality
and reason over illogic and unwieldiness is to ignore the many cipheredpositional
systems that have failed to spread or failed to survive. While, owing to the political might of nations that use them, the Western numerals are very important,
they are merely one branch of one phylogeny. In placing them in the middle of
my study, I choose to emphasize that their present triumph is neither inevitable
nor eternal.
chapter 7
Mesopotamian Systems
Numerical notation first developed in Mesopotamia around 3500 bc, contemporaneously with or slightly earlier than its development in Egypt. Scholars interested in the diffusion of Babylonian astronomy and mathematics to the Greeks
have long studied Mesopotamian numeration (Neugebauer 1957, van der Waerden
1963). Yet to depict the Mesopotamian phylogeny of numerical notation systems
as an archetypal case for the evolution of numerals, or to use it as the basis for a
universal evolutionary pattern, is dangerous. While Mesopotamian mathematics
is important for understanding later Greek developments (and, in turn, modern
Western mathematics), Mesopotamian numeration is nearly a historical dead end.
Although their history spans three millennia, the Mesopotamian numerals did not
spread geographically far beyond their point of origin, and did not survive when
placed under pressure from the numerical notation systems of later inhabitants of
the region.
The main signs of the Mesopotamian numerical notation systems are shown in
Table 7.1. There are several ways to classify them, depending on which features we
emphasize. Looking at the numeralsigns alone, the systems divide rather neatly
into archaic systems, used prior to 2000 bc and written using curviform symbols
made with a round stylus, and later cuneiform systems, written using wedgeshaped symbols. Both were written almost exclusively on tablets using a stylus to
impress signs onto wet clay. A second important distinction is between systems
that are primarily decimal and those that are primarily sexagesimal, or base60.
228
229
Babylonian
positional
Old Persian
Hittite
f
f
f
g
g
g
AssyroBabylonian
Mari
B
B
Z
A
A
S
10
Sumerian
Cuneiform
Systems
ProtoElamite
decimal
Bisexagesimal 2
Bisexagesimal
Sexagesimal
System
Archaic
Systems
C
C
P
60
f
i
w
100
M
Q
120
600
fm
gi
1000
N
R
1200
3600
7200
10,000
36,000
230
Numerical Notation
Mesopotamia is the only region of the world where sexagesimal numerical notation is attested.1 Finally, comparing the interexponential structures of the systems,
we can distinguish between additive systems, which include most of the systems,
and positional systems, of which the only true example is the Babylonian positional system.
Protocuneiform
Around 3200 bc or perhaps slightly earlier, the antecedent of the later Sumerian
script arose at the city of Uruk in southern Mesopotamia, during what is now
known as the Uruk IV period.2 This protoscript, which was probably read in Sumerian, lacked any means of expressing phonetic sounds. By the Uruk III period
(c. 3000 bc), it had spread from Uruk (the primary Mesopotamian city at the time)
to the north, to Jemdet Nasr, Khafaji, and Tell Uqair. The texts of this period of
Mesopotamian history do not represent a true literate tradition but rather a protohistoric system of bookkeeping and administration. In total, about 5,600 clay tablets have been recovered that record this script, known as protocuneiform. Around
sixty of the twelve hundred protocuneiform signs can be assigned numerical or
metrological values (Nissen, Damerow, and Englund 1993: 25).
Although Falkenstein (1936), who wrote the first comprehensive description of
the Uruk tablets, thought the protocuneiform texts from Uruk were both decimal
and sexagesimal, Friberg (197879, 1984) determined that there was no protocuneiform decimal system. In the 1980s, using computeraided analysis of the
entire corpus of texts, Nissen, Damerow, and Englund (1993) established that as
many as fifteen distinct systems (of which five were particularly common) were
used at Uruk, each used for enumerating a specific category of discrete objects
or metrological quantity.3 By examining the maximum number of times each
numeralsign is repeated, they determined the relative values between signs in
any given system (as if we were to infer that the Roman numeral V represents 5
by noting that I is repeated four times at most). This technique works because
cumulativeadditive systems order powers within each numeralphrase from
However, Price and Pospisil (1966) claim that the Kapauku of Papua New Guinea derived their sexagesimal lexical numerals from the comparable Babylonian numerical notation.
Over the past twenty years of research, the chronology of protohistoric Mesopotamia
has been shifted backward; older sources tend to regard the Uruk IV period as representing the early third rather than the late fourth millennium bc.
My discussion of the systems (including their functions) is derived almost entirely from
the work of Nissen, Damerow, and Englund (1993: 2529).
Mesopotamian Systems
231
highest to lowest and regularly replace lower powersigns with higher ones wherever possible. A difficulty is that a given numeralsign may be found in several of
the protocuneiform systems, but its value often varies from system to system.
Thus, B is equal to 10 A in some systems, but to 6 A in others.
Yet while we can identify the numerical ratio between the values of any two
signs within a system, we often cannot identify the specific quantity represented
by any one sign. For systems used for counting discrete objects, it is easy to identify the basic sign for 1, since fractions of humans do not normally occur in texts,
but for systems that measure area or capacity, we can never ascertain with certainty
which sign (if any) has the basic value of one unit. Following Nissen, Damerow,
and Englund (1993), I present the values for these metrological systems as ratios,
since we can only tell the value of a sign relative to the other signs of the system.
Despite having different numeralsigns and different numerical values, all the
protocuneiform numerical systems have much in common. All are cumulativeadditive, although some individual numeralsigns are formed multiplicatively
e.g., D (600) = C (60) B\(10). Groups of identical signs were sometimes sorted
into two or three rows for easy reading, but this is not a universal rule, and some
tablets contain long strings of signs. Numerals were most often grouped with signs
arranged from highest to lowest (although there are some rare exceptions, which
may be scribal errors). A single numeralphrase, together with one or more ideograms, was enclosed in a box in a section of the text.4
The protocuneiform texts are mainly accounting documents, often written on
both sides the obverse with a series of amounts of commodities, the reverse
with a single total. The greaterthanexpected prevalence of round or nearly
round numerals in protocuneiform texts allows the identification of hypothetical
problems that were used as training exercises and thus were not actual economic
texts (Friberg 1998). Learning the various signs, the ratios among them, and how
to construct texts would have required considerable scribal training, including
school mathematics associated with the temple economy at Uruk (Robson 2007:
6364). For instance, the late fourthmillennium tablet W20044,20 is an exercise in calculating the area of an irregular quadrilateral field (Robson 2008: 30).
Throughout the history of the various systems use, only a very small portion of
the populace would have had access to the training necessary to master the protocuneiform notations.
4
Note that the protocuneiform script was written vertically in columns reading from top
to bottom, but I follow Assyriological convention (and that used by Nissen, Damerow,
and Englund) in showing the signs rotated ninety degrees counterclockwise and thus
read horizontally from left to right. This convention reflects a similar change in the
direction of writing cuneiform signs around the middle of the third millennium bc.
Numerical Notation
232
Sexagesimal (S)
3600
=10 E
600
=6 D
60
1/2
=10 C =6 B
=10 A
( =6 &
=10 ^
Sexagesimal (S)
a
10
=2
The letters in parentheses in this table and the following ones are those assigned to each system by
Nissen, Damerow, and Englund (1993) in their research.
Sexagesimal Systems
The two sexagesimal systems shown in Table 7.2 alternate between factors of 6
and 10, and were the first and easiest to be deciphered because their structure is
identical to that of the later Sumerian numerals. The main sexagesimal system
(S) is employed in slightly less than half the Uruk texts (Damerow 1996: 292). It
was used to enumerate most discrete objects: humans, animals, finished products,
tools, and containers, which explains its frequency of use. The subsidiary S system
was used to enumerate a much smaller category of discrete objects, such as dead
animals and jars of some liquids.
Bisexagesimal Systems
The two bisexagesimal systems shown in Table 7.3 are so named because an additional factor of 2 is interpolated among the factors of 6 and 10 used in the sexagesimal systems. While the regular bisexagesimal (B) system is identical to the regular
sexagesimal (S) system up to 60, it has new signs for the values of 120 (60 2),
1200 (120 10), and 7200 (1200 6). It enumerated discrete numbers of grain
products, cheese, and fresh fish, and is the second most common system found in
the archaic texts. The function of the identically structured but much less common B* system is unclear, but it may have indicated discrete quantities of some
kind of fish. Both systems appear to have been part of a rationing system, one for
which a numbersign between 60 and 600 may have been useful.
1200
=6
120
60
10
=10 M
=2
C =6 B
=10 Q
=2
P =6 Z =10 S
=10 A =2
1/2
Mesopotamian Systems
233
=6
=10
=3
=6
=10?
GAN2 System
The GAN2 system shown in Table 7.4 is used to represent area measures. While
its signs are the same or similar to those of the common sexagesimal system (S),
the GAN2 signs values differ from those of the main system. For instance, where F
means 36,000 in sexagesimal numerals and is thus 3,600 times greater than B (10),
in the GAN2 system it is only ten times greater. This similarity probably has something to do with the use of roundended writing styli in all the protocuneiform
numerical systems.
EN System
This uncommon notation system, shown in Table 7.5, is known from only
twentysix texts, and may have represented weight measures. All but one of
the tablets from Uruk on which the EN system was used were found at a single
location, suggesting that whatever its function, it must have been very restricted
in use.
E Systems
This relatively common group of numerical systems, shown in Table 7.6, denoted
various capacity measures of grain. While its signs are similar to those of the sexagesimal systems, their order and the ratios between successive signs are quite
different. For instance, while the ratio between A and B is 10 in the sexagesimal,
bisexagesimal, and EN systems, it is only 6 in the E systems. The regular system
enumerated capacity measures of barley, the system for germinated barley for
brewing beer, and the * system for barley groats.
U4 System
This rather unusual numerical notation system becomes less so considering that
its function is for recording time and calendrical units. By combining a single
Table 7.5. EN numerals
=10
=2
=2
Numerical Notation
234
Table 7.6. E numerals
System
D =10 C =3
System
=3
System *
E =10 B =6 A =5 J
=10 =6 =5 =5
=10 =6 =5
ideographic sign with numerical signs for 1 and 10, all the major divisions of the
year could be expressed easily.
=10
=12
,
1 month
=3
10 days
=10
.
1 day
Mesopotamian Systems
235
signs resembling later archaic numerals and contain the correct total of tokens,
suggesting that the systems are connected (Nissen, Damerow, and Englund 1993:
127129). Furthermore, there are some similarities between the threedimensional
tokens and the protocuneiform ideograms, suggesting that the tokens developed
into writing through the recognition that, if the total of a transaction is written on
clay, one need not actually use the clay tokens but need only record their values.
SchmandtBesserats conclusions have been received with some skepticism (see
especially Lieberman 1980, Zimansky 1993).5 Firstly, the scope in time and space
of the token system is far greater than that of the protocuneiform numerals;
it is implausible that such a widespread phenomenon represents a uniform system. Moreover, in SchmandtBesserats study of tokens from the UrukJemdet
Nasr period (c. 3000 bc), around twothirds of the tokens come from Susa in
Iran, while only 10 percent come from the very thoroughly excavated site at Uruk
(Lieberman 1980: 353). This suggests that the token system is unlikely to have
given rise to numerals and writing at Uruk. Finally, some of the most common
tokens are correlated with protocuneiform signs for rare objects such as nails and
days of labor, whereas given the accounting function established for the tokens,
we would expect livestock, people, and grain to be the most common tokens, as is
the case in protocuneiform texts (Zimansky 1993: 316).
This discrepancy points to a further problem. The archaic numeral systems
always place a numeralphrase in front of an ideographic sign; 16 + sheep =
16 sheep, and so on. While the protocuneiform numerals partly fuse quantity
and quality, because different systems represent different commodities, they do
not do so completely, because one always needs a further sign to indicate exactly
what is being counted. With the tokens, however, there is no separation of numerals and the objects being counted; to show sixteen sheep, one simply uses sixteen
tokens for sheep. Thus, there is no correspondence between the archaic numeralsigns and the shapes of tokens. The use of tokens sealed within bullae appears to
have been an accounting technology that predated, but then later coexisted with,
the protocuneiform numerals. While some early protocuneiform numerals are
found on clay bullae, this is insufficient evidence that tokens led to numerals.
Conversely, numerical signs resembling the protocuneiform ones have been
found, not on bullae, but on ordinary clay tablets in late preliterate contexts
at Uruk as well as at Jebel Aruda, Susa, and elsewhere (Nissen, Damerow, and
Englund 1993: 127130; see especially Figures 113, 114). These tablets have numerical signs only (no ideograms), and disobey the ordinary rule that once a certain
5
I cannot hope to address her claim that the tokens are ancestral to the protocuneiform
script, and will restrict myself to the similarities and differences between the token system and the protocuneiform numerals.
236
Numerical Notation
number of lowervalued signs have been written, they are replaced with a single
highervalued sign. For instance, one tablet from Jebel Aruda contains twentytwo
B signs, among others (Nissen, Damerow, and Englund 1993: 130). In any of the
later systems, twentytwo signs would have to be replaced by a smaller number of
highervalued signs. Because these inscriptions are found in late preliterate contexts and are similar but not identical to the protocuneiform numerals, they are
immediately ancestral to them and date from a period when the system was still
being developed, weakening the hypothesis that they derived from tokens.
While we do not know the language in which protocuneiform numerals were
read, the Sumerian lexical numeral system is mainly sexagesimal, and furthermore,
10 is a subbase in the lexical numerals just as it is in the protocuneiform numerals (Powell 1971, 1972a, 1972b). On this basis, Powell (1972b: 172) has correctly
discerned that the presence of a sexagesimal system of notation in the archaic
texts from Uruk and Jemdet Nasr constitute [sic] the best indeed irrefutable
evidence that Sumerian is the language of those texts. Yet, as Hyrup (2006:
7982) points out, protocuneiform notation was not meant to record any spoken
language, but was an artificial recording system designed for limited purposes, and
only very gradually became assimilated to Sumerian.6
The multiplicity of protocuneiform numeral systems and bases is not lexical in
origin, but is likely based on Sumerian metrological systems in the Uruk period,
for which we have minimal nontextual evidence. We do, however, have substantial textual evidence for the metrological systems of the Early Dynastic and later
periods. The ratios between various signs in the protocuneiform numeral systems
dealing with measures of capacity, area, and weight are similar to the ratios found
in later Sumerian metrological systems. This supports the contention that the odd
ratios of some of the older systems are due to unattested metrological systems
that continued into betterdocumented periods. The increasing administrative
demands associated with the rise of the Uruk citystate in the late fourth millennium bc created a new need for record keeping, metrology, and accounting, of
which the numerals and the clay bullae are two distinct consequences.
Mesopotamian Systems
237
Mesopotamia directly reflects the numerical abilities of Mesopotamians, and represents a universal stage of concrete numeracy that precedes the modern abstract
number concept. In this respect, his argument is similar to that of Hallpike (1979),
who applies the insights of Piaget, Vygotsky, and others from developmental psychology to draw a parallel between individual cognitive development and the evolution of thought in societies. Damerow claims that the peoples of early Mesopotamia
could not conceive of abstract numbers, but were only capable of concrete counting
(Damerow 1996: 275297). Because multiple protocuneiform numerical notation
systems were used for representing different objects, and a single sign could have
different relative values in different systems, he argues that Mesopotamian scribes
could conceive of 8 sheep or 8 jars of oil but not simply 8. Taken to its logical
conclusion, this would imply that users of the protocuneiform numerals could see
nothing in common between eight sheep and eight jars of oil. I cannot see how this
can be the case; if so, it would be impossible to make any connection between eight
sheep and eight marks on a clay tablet, and numeration would be impossible.
Furthermore, in order for these contextdependent numerals to represent a
stage of archaic arithmetic in the evolution of numeration, as Damerow (1996:
296) suggests, we would expect similar systems to be present in other civilizations.
Yet nothing of the sort can be found in Shang, Predynastic Egyptian, or Zapotec
inscriptions, the other early and independently invented systems. This is not to say
that there are no cognitive consequences of the use of a dozen or more numerical
notation systems, but whatever they are, they will not be universally applicable to
every society.
Finally, we have no idea how many of these systems would have been known to
any individual official, and no evidence from the archaic period as to how numerals were manipulated and used arithmetically. We simply have values and totals,
which do not tell us very much about how people were actually thinking about
number.7 Even if individuals used many systems, this does not prove concreteness of
thought. In contrast to the protocuneiform numerical notation, there was a single
perfectly ordinary set of Sumerian lexical numerals (Powell 1971).8 Someone capable of abstract thought might well use multiple systems of numerical notation to
7
Liveranis (1983) intriguing conclusion that a fragmentary Uruk IVperiod clay tablet
indented with holes may have served as a counting board has not been confirmed and
must remain tentative unless further finds are made.
SchmandtBesserat (1984, 1992) has made much of the parallel between the many protocuneiform numerical notation systems and the use of numeral classifiers in Japanese,
the Mayan languages, and others, where the set of numerals is modified depending on
the class of object being counted. Numeral classifiers are not a feature of Sumerian. If
taken to its logical conclusion, this would imply that the modern Japanese do not have
a concept of abstract number.
238
Numerical Notation
prevent confusion as to the type of thing being counted, or to correspond to metrological systems. There is no qualitative difference between the Uruk systems and
the modern use of Roman numerals to distinguish the foreword of a book from its
main text, or the use of hexadecimal numerals for computing purposes. Ironically,
one of the principles behind the new mathematics movement in North America in
the 1960s was the claim that teaching students to calculate using numerical systems
of different bases would improve their understanding of abstract number concepts. I
believe the Uruk scribes had an abstract number concept, but realized that abstract
written numerals were not the most efficient solution to the problems they were
facing. The theoretical importance attributed to the protocuneiform numerals as
evidence of an evolutionary stage of cognition is entirely unwarranted.
Mesopotamian Systems
239
ProtoElamite
Around 3100 bc, a ideographic writing system developed in southern and western
Iran, the region known as Elam in later Mesopotamian sources. This script, now
known as protoElamite, is attested in over 1,500 texts, mainly from the major
urban center of the region, Susa; most date from the Susa III period around 3000 bc.
A few other protoElamite texts have been found at Tepe Yahya and elsewhere in
modern Iran. It is a linear script, read from right to left and in lines proceeding
from top to bottom. The language it was intended to represent cannot be identified, but protoElamite numerals can be read. While the protoElamite ideograms
are different from those of early Mesopotamia, the protoElamite numerals are
very similar to the protocuneiform systems.
As with the protocuneiform numerals, confusion over the nature and number
of protoElamite numeral systems has delayed their correct decipherment until
recently. Brice (196263) provides a useful summary of several early twentiethcentury efforts to decipher the protoElamite numerals, all of which assume a
single decimal and cumulativeadditive numerical notation system. An adequate
decipherment of the protoElamite numerals has been achieved recently through
the mathematical analysis of the corpus of protoElamite texts by Robert Englund
and Peter Damerow (Damerow and Englund 1989, Englund 1996). Damerow and
Englund realized that, as with the protocuneiform numerals, not only were there
multiple protoElamite numerical notation systems, but the relative values of
individual numeralsigns vary from system to system. There are five major protoElamite systems: three for counting discrete objects, another for capacity measurements, and another for area measurements (Englund 1996: 162).
The protoElamite numerical notation systems for counting discrete objects
are shown in Table 7.8 (Englund 1996: 162; cf. Potts 1999: 78).9 The three systems
are identical for 1 and 10, and the sexagesimal and bisexagesimal systems are further similar for 60. The sexagesimal system, like the protocuneiform sexagesimal
numerals, is not a pure base60 system; instead, each successive number alternates
by factors of 10 and 6; that is to say, it has a subbase of 10. In the bisexagesimal
system, the value 120 comes after 60 (a factor of 2).
9
As with the protocuneiform numerals, I have represented the numerals as they would
be read horizontally (following Assyriological convention; cf. Damerow and Englund
1989) rather than vertically (cf. Englund 1996).
Numerical Notation
240
Sexagesimal
Bisexagesimal
A B
Decimal
A B
Grain
products
Animate
objects
The main signs of the systems for measuring capacity and area are shown in
Table 7.9. Because they are not used for discrete objects, they are represented in
terms of the ratios between values, not as discrete numerical values. These two systems are very similar (though not identical) to the E and GAN2 protocuneiform
systems, so, following Damerow and Englund, I have used those labels.
The striking resemblances between the protocuneiform and protoElamite
numerals make it certain that the latter were modeled on the former (Potts 1999:
7677). In fact, while the respective scripts are entirely dissimilar, it is a matter of personal preference whether we regard the protocuneiform and protoElamite numerals as distinct sets of systems or as two regional variants of a single tradition. Because the first texts from Uruk date to the thirtythird century
bc, while those found at Susa date to the late thirtyfirst century bc, the protoElamite ones cannot have been ancestral to those at Uruk, and must have diffused
from west to east in the context of interregional trade. Given the importance of
the Uruk citystate in the late fourth millennium bc, it is unsurprising that the
numerals would spread to Susa, the other major polity at that time. The main
difference between the two sets of numerical notation systems is the existence of
a decimal system for counting discrete quantities of animals and humans in protoElamite. It may be that the language of the writers of the protoElamite texts had
decimal lexical numerals, whereas we know that Sumerian numerals are primarily
sexagesimal.
X =6 D =10 C =3 E =10
=6
=5
E =10?
=3
=6
J
K
A
=2
Mesopotamian Systems
241
The protoElamite numerals did not spread beyond Susa and a few other sites
in modern Iran. Brices (1963) tentative identification of similarities between the
protoElamite and Linear A (Minoan) numerals cannot be taken seriously as
indicative of a historical connection, given the geographical and temporal distance between the two. The protoElamite numerals ceased to be used around
2900 bc, following the decline of Susa as a major urban polity in the early part
of the third millennium bc and the subsequent rise of the various Mesopotamian
citystates. The numerals in the Old Elamite cuneiform texts, which are roughly
contemporaneous with the Old Akkadian texts in Mesopotamia, are derived from
later Mesopotamian systems rather than from protoElamite (Potts 1999: 79). The
protoElamite numerals are best seen as a brief florescence within a single citystate, rather than as part of a longer tradition.
Sumerian
The only system among the multitude of protocuneiform systems to survive
into the Early Dynastic period (2900 to 2350 bc) was the sexagesimal (or, more
accurately, the decimalsexagesimal) system. While it was originally used only for
counting discrete objects, it began to be used for all numerical functions as the
older metrological systems were abandoned. At the beginning of the Early Dynastic, significant changes were taking place in the script of the region. The older
ideographic and curviform protocuneiform symbol system slowly transformed
into a writing system that used wedgeshaped (cuneiform) signs and expressed
phonetic as well as conceptual information. From this, we can tell that Sumerian
was the language in which the script was read. Yet, despite these alterations to
the script, the numeralsigns remained identical to the archaic sexagesimal ones.
One important change occurred around the twentyseventh century bc, when the
numerals, like the entire script, underwent a ninetydegree rotation, so that they
were written and read horizontally from left to right rather than vertically from
top to bottom. The Sumerian numerals are shown in Table 7.10 (Nissen, Damerow,
and Englund 1993: 28).
These six numeralsigns were combined to make a cumulativeadditive numerical notation system. Normally, groups of four or more signs were arranged in two
rows to facilitate rapid reading. Because this system has signs for both 60 and 3600
(= 602), it has a sexagesimal component. In a purely sexagesimal system, one would
need to repeat each sign up to fiftynine times, which is impractical, but the Sumerian system also has subbase signs for 10, 60 10 (600), and 3600 10 (36,000). The
latter two signs are multiplicative combinations of the small circle for 10 with the
sexagesimal signs for 60 and 3600. This decimal subbase is similar, but not identical, to the use of the subbase of 5 in the Roman numerals. While the figures of the
Numerical Notation
242
10
G
B
Horizontal
A
B
14,254 = EEE\DDD
DD
Vertical
60
\A
3600
36,000
H
I
E
F
C
D E
F
CCCC\BBB\AA
CCC\
AA
19 = BB
600
+ (3 10) + (4 1)
(20 LAL 1)
Roman subbase (V, L, D) could occur only once in any numeralphrase because
5, 50, and 500 are half of 10, 100, and 1000, respectively, the decimal signs in the
archaic Mesopotamian numerals could be repeated up to five times. The sign for
60 is a large version of the sign for 1, just as the sign for 3600 is a large version
of the sign for 10. Because the big 1 is 60 times greater than the regular 1, but
the big 10 (3600) is 360 times greater than its counterpart, I cannot agree with
Liebermans (1980: 343) suggestion that these signs represent the use of sizevalue,
which then evolved into placevalue, over time. This feature simply derives from
the fact that two styli, one twice as large as the other, were used to impress numerical signs on clay tablets (Powell 1972a: 1112).
The Sumerian numerals provide the first evidence for the use of subtractive
notation to express certain numbers, especially those that end in 8 or 9 in the
Western numerals, using a Sumerian ideogram that corresponded with the phonetic value LAL. Thus, instead of writing 19 as one sign for 10 plus nine signs for 1,
it could be written as 20 1, as seen in Table 7.10. A sign or signs placed inside the
LAL sign indicated an amount to be subtracted from the signs preceding it. This
technique was used at Fara (ancient uruppak), perhaps as early as 2650 bc (Jestin
1937: Pl. LXXXIV). There is no evidence of subtractive lexical numerals in Sumerian
comparable to the Latin duodeviginti and undeviginti; the Sumerian words for 18
and 19 are etymologically 10 + 5 + 3 and 10 + 5 + 4, respectively (Powell 1971: 47).
Rather, this innovation had its origin strictly in numerical notation and the desire
to express numbers more concisely.
As in the archaic period, Early Dynastic numerals are found overwhelmingly in
economic or administrative texts, or in scribal exercises related to these functions.
In the archaic period, there was no indication how calculations were being done
(though calculations must have been made). At Fara, however, Sumerian tables
of squares, geometrical and arithmetical exercises, and other arithmetical aids
Mesopotamian Systems
243
10
60
600
3600
36,000
216,000
k o
nm
LAL:
have all been found (Powell 1976, Hyrup 1982). Nevertheless, the use of numerals for representation, especially in administrative contexts, greatly exceeds the
frequency of their use for computation. There is nothing indicating the direct use
of Sumerian numerals for computation (by lining up columns, etc.), as in Greek
and Western arithmetic. Damerow (1996: 236237) laments the fact that, despite
the wealth of Early Dynastic economic records, we have no idea how multiplication was performed; he suggests that it must have been through a nonpermanent
means, such as counting boards, finger reckoning, or mental calculation.
By 2500 bc, the transition from the older Sumerian script to cuneiform signs
had been completed, except for the numerals. Beginning in the Presargonic period
(c. 26002350 bc), the older curviform numerals began to be replaced with a
set cuneiform numeralsigns, while remaining virtually unchanged structurally
(Powell 1972a: 13). This had the advantage of requiring only one stylus for all writing, whether lexical or numerical. While this trend appears to have been initiated
by the Sumerians themselves, it was hastened considerably, starting around 2350
bc, by the rise of Akkadian hegemony over Mesopotamia. These new numeralsigns are shown in Table 7.11 (Powell 1971: 244).
The signs for 1 and 60, which had previously been semicircular and horizontal,
became vertical wedges. The earliest cuneiform numeralsign for 60 was written as
a big 1, just as it had been in the curviform numerals, but because the two signs
were made with the same stylus, the size difference was always minimal, and soon
the two signs became identical (Powell 1972a: 13). This feature does not mean that
the system used the concept of placevalue, although it may have played a role
in the invention of the later sexagesimal positional system (Powell 1972a: 1314).
The old round sign for 10 was replaced by a Winkelhaken or corner wedge, made
by impressing the stylus onto the clay tablet perpendicularly, while the large round
sign for 3600 was represented visually by four (or occasionally five) wedges placed
in a rough circle. In other respects the writing of 600 and 36,000 as 60 10 and
3600 10, and the basic cumulativeadditive structure the cuneiform numerals
were identical to the curviform ones. The phrase used for 216,000, not attested
in the earlier numerals, is a combination of the sign for 3600 and the ideogram
244
Numerical Notation
GAL big, and is quite rare (Powell 1972a: 7). Powell also describes an even more
complex phrase for 12,960,000 (216,000 60), argal unutaga, big everything
which hand cannot touch. For such lexical phrases, we need to ask at what point
a phrase ceases to become part of a numerical notation system. The subtractive
ideogram LAL is used in this system, as in the archaic one, but it is depicted using
two cuneiform wedges.
The replacement of the curviform by cuneiform numerals was a gradual process. While the older system required additional styli to write numerals, and its
numeralsigns generally took up more space, it stood out more clearly in a text
of cuneiform characters, making totaling easier (Powell 1972a: 12). Additionally,
a norm developed by which the two sets of numerals could be used side by side
to indicate different functions. Possibly the older numerals indicated quantities
directly counted, possibly using clay counters, while the cuneiform numerals
enumerated quantities of objects not actually present to be counted (Lieberman
1980: 344345). Alternately, Damerow (1996: 238) notes that some Early Dynastic
economic texts from Girsu use the older numerals for amounts of grain and the
cuneiform numerals for amounts of animals, and hypothesizes that this may have
been done to avoid confusing the two different categories when taking sums. This
use of multiple numerical notation systems is analogous to the modern use of
Roman and Western numerals side by side. The round Sumerian numerals had
been abandoned by the Ur III period (when Sumerian rulers regained control of
Mesopotamia), and are not attested later than 2050 bc (Powell 1972a: 13).
The Akkadian conquest, the most important political event of thirdmillennium
bc Mesopotamia, had a minimal effect on numeration. The Akkadian kings and
officials (c. 23502150 bc) were content to use the cuneiform and even the archaic
numerals for most of the same purposes for which they had been used in the Early
Dynastic period. More change in the numerals is visible in the NeoSumerian
Ur III period (2150 to 2000 bc), during which the archaic numerals disappeared
entirely. One slight modification that was tried in some Akkadian texts was to
write multiples of 60 using units followed by the Akkadian lexical numeral for
60, ui (:), using multiplicative notation (Labat 1952: 244247). Thus,
instead of writing 120 as ff, it would be written as ff:. This is a much more
cumbersome representation, and probably was used in part to distinguish 120 (ff)
from 2 (ff). For the higher decades 70, 80, and 90 the regular Sumerian forms
were always used by the Akkadians (fg, fgg, and fggg, respectively). Regardless,
many Akkadian inscriptions where ui could have been used are written in the
ordinary Sumerian fashion.
The Sumerian cuneiform system is ancestral to all the later systems of Mesopotamia. The Semitic cuneiform decimal systems (Eblaite and AssyroBabylonian)
were directly derived from a Sumerian ancestor. The decimal structure of these
Mesopotamian Systems
245
systems reflected the lexical numerals of the Semitic languages of its users. While
ThureauDangin (1939: 107) believed this tradition to have been developed in the
Old Akkadian period (starting c. 2350 bc), it is now clear from the library at Ebla
that it developed as early as 2500 bc (Pettinato 1981). The sexagesimal cumulativepositional system used in Babylonian mathematics and astronomy was also
modeled on the Sumerian cuneiform system. It may have arisen in the Ur III
period, and was used by the twentieth century bc at the very latest (cf. Powell
1976, Whiting 1984).
The Sumerian cuneiform system continued to be used for most purposes until
the Old Babylonian period (c. 20001595 bc). Around that time, the AssyroBabylonian decimal system began to be used for most administrative, commercial, and
literary functions, while the sexagesimal positional system was used for mathematics and astronomy once again perpetuating the tradition of using multiple numerical notation systems for multiple purposes. Several Old Babylonian
tablets provide translations from the old Sumerian additive numerals to the new
positional system (Nissen, Damerow, and Englund 1993: 146147), indicating
either a need to learn the new positional system or, alternatively, that the older
cuneiform system was already being forgotten. By the fifteenth century bc, it had
disappeared from regular use. However, a peculiar vestige of the base60 Sumerian system survived in certain late inscriptions, particularly those indicating the
sizes of buildings (De Odorico 1995: 4). One such example is the Nameninschrift
of the Assyrian king Sargon II (722705 bc), which describes the dimensions of
the fortress at Khorsabad as 16,283 cubits, the numeral of my name, notated
using Sumerian numerals rather than the AssyroBabylonian system that would
be expected (Fouts 1994: 207). In this case, the notation served literally to indicate
a numeral corresponding to the royal name of Sargon by correlating the signs for
3600, 600, and 60 with the phonetic values ar, nr, and , which could not have
been done using the AssyroBabylonian decimal system. This inscription has been
compared to later Greek, Hebrew, and Arabic gematria using alphabetic numerals
(Chapter 5), but it seems to be an earlier and independent development.
Eblaite
The inhabitants of the citystate of Ebla (in the western part of modern Syria)
spoke a West Semitic language but were strongly influenced by Sumerian culture.
A great library of thousands of Eblaite cuneiform texts dating certainly to the
period prior to 2350 bc (the Akkadian conquest), and possibly as early as 2500
bc, provide us with ample evidence regarding the numerals used by the Eblaites
(Pettinato 1981). This system is almost identical to that used by the Babylonians
some centuries later, and reflects the shared Semitic language and culture of these
Numerical Notation
246
Table 7.12. Eblaite numerals
1
10
60
100
1000
10,000
miat
liim
rba
A
B
C
f
g
f
AA\rba AA liim AAA miat CBBAA
AA\\\\\\\ AAA
100,000
maiat OR
maihu
10,000
1000
100 60 10 10 1 1 = 24,682
two groups in contrast to those of the Sumerians. As indicated in Table 7.12, the
Eblaite numerical notation system consisted of two sets of numeralsigns for numbers below 100, one curviform and the other cuneiform (corresponding with the
Sumerian archaic and cuneiform systems), but only one set of expressions for the
powers above 100 (Pettinato 1981: 183184).
The Eblaite system is cumulativeadditive for values less than 100, and multiplicativeadditive above that point. The signs for 1, 10, and 60 are ideographic
signs identical to those used in the two Sumerian sets of numerals. The two sets of
numeralsigns served quite separate functions: the curviform numerals were used
for basic enumeration and counting discrete objects, while the cuneiform numerals were used only for capacity measures such as the mina and gubar, as well as for
regnal years of kings (Pettinato 1981: 183184). The sign for 60 was used to express
the tens values in numbers between 60 and 99; its presence, a holdover from
Sumerian, is the major irregularity in an otherwise perfectly decimal system. The
signs for numbers above 100 are in fact the Eblaite lexical numerals and were
combined multiplicatively with the unitsigns as necessary.
Because it is decimal and multiplicativeadditive above 100, this system required
only one ideographic sign (the crescent or vertical wedge) for the higher powers;
however, the repetition of intraexponential signs for the units, coupled with the
use of complex two and threesyllable words, meant that numerals were fairly
long and cumbersome. To reduce this length, two features were often used. Firstly,
just as in the Sumerian system, subtractive numerals were sometimes used for
certain numbers to eliminate the need to write seven, eight, or nine unitsigns
by placing the subtrahend after the syllable lal or l. Secondly, the words miat
for 100 and liim for 1000 were often shortened to the single syllables mi and li,
respectively. Thus, in one text, 7879 (expressing a number of gubar measures of
barley) is written in cuneiform numerals as 7 li 8 mi 60 10 10 l1 (7 1000 + 8
100 + 60 + 10 + 10 1) (Pettinato 1981: 134). Such syllabic abbreviations are reminiscent of the Greek acrophonic numeralsigns (Chapter 4).
Mesopotamian Systems
247
AssyroBabylonian Common
Because historians of mathematics are especially interested in the origins of our
base60 units of time and the division of the circle, enormous attention has been
paid to the Babylonian positional numerals the cumulativepositional, base60
system used for astronomy and mathematics. The far more common decimal and
additive numerals (which I call the AssyroBabylonian common system), which
were used for most economic, monumental, and literary purposes throughout
Mesopotamia, are almost forgotten. This system came into common use in the
Old Babylonian period (starting c. 2000 bc), a position it would maintain for
over 1,500 years. The numeralsigns of this system are shown in Table 7.13 (De
Odorico 1995: 4).
The system is cumulativeadditive below 100, multiplicativeadditive above 100,
and is always written from left to right. For the most part, it is purely decimal.
The units were expressed cumulatively, except that 9 could be written using three
overlaid vertical strokes ($) as an alternative to writing it with nine strokes (9)
(De Odorico 1995: 4n). The tens values were usually expressed decimally using
one through nine Winkelhaken corner wedges for 10. The vertical wedge for 60 is
identical to that for 1, but unlike the Sumerian system, f was not normally used to
represent 60 alone, which would have created ambiguity, but only in combination
with signs for 10 and 1 to write numbers from 70 to 99. As in the Akkadian variety
of the Sumerian cuneiform system, the lexical numeral ui (:) was used to
indicate 60 or multiples thereof; however, this phonetic form was not used to write
70, 80, or 90, and even for 60 its use declined over time. Therefore, there were as
many as three different numeralsigns for 60, one decimal () and two sexagesimal (: or f). Above 100, the multiplicative principle was used quite freely
and could be combined in various ways to express very high numbers. The sign for
100 was the syllabic sign ME, an abbreviation of the Babylonian word for 100,
meat, while that for 1000 was a multiplicative combination of the signs for 10 and
100. For instance, a scribe from the period of Sargon II wrote 305,412 as shown in
Numerical Notation
248
10
60
305,412 =
or
100
1000
gi
3i5gi4ia2
((3 100 + 5) 1000) + (4 100) + 10 + 2
Table 7.13 (Ifrah 1998: 139). In theory, this system could be extended as far as one
wished by juxtaposing signs for 100 and 1000 repeatedly, even though there was no
sign for zero. Furthermore, unlike the Sumerian system, in which the signs for 1
and 60 were identical, this system presented no ambiguities to the reader.
While this system was sometimes called Akkadian (ThureauDangin 1939), it
was rarely used during the period of Akkadian control of Mesopotamia and began
to predominate only during the Old Babylonian period. It originated in response
to the increased power of Semitic peoples in Mesopotamia in the latter part of the
third millennium bc: Akkadians, to be sure, but also Eblaites, Babylonians, and
others. Its structure reflects the decimal lexical numerals of the Semitic languages
rather than Sumerian lexical numerals, although the continued use of a special
sign for 60 gives testament to its descent from the Sumerian numerals. All of the
administrative, commercial, literary, and religious texts of the Babylonians and the
Assyrians were written using this set of numerals.
Perhaps the greatest significance of the AssyroBabylonian common system is
the large number of descendant systems it produced. Earliest among these is the
system used at the citystate of Mari around 1800 bc, which blends features of this
system and the Babylonian positional system. In the middle of the second millennium bc, both the Ugaritic and the Hittite cuneiform scripts began using numerals based on the AssyroBabylonian ones, in the context of Mesopotamian trade
with these polities. The Ugaritic texts written between the fifteenth and twelfth
centuries bc at Ugarit on the Mediterranean coast use the cuneiform ideograms
for 1 (f) and 10 (g); however, numerals in Ugaritic were normally written lexically
(Gordon 1965: 42). This was likely borrowed from the AssyroBabylonian system,
but we have no idea whether higher numbers could be expressed through numerical notation. The Old Persian cuneiform numerical notation system, developed in
the sixth century bc (by which time Mesopotamia was under Persian rule), also
derived from the AssyroBabylonian system rather than from any of the numerous
other systems then used in the region. Finally, as I argued in Chapter 3, the earliest
Levantine systems (Phoenician and Aramaic) developed around 800 bc as a blend
of Egyptian hieroglyphic (or perhaps Hittite) and AssyroBabylonian influences.
Mesopotamian Systems
249
p
4
q
7
s
8
40
r t
9 d
Babylonian Positional
The Babylonian positional numeral system is assigned such great importance
by many historians of mathematics that one could easily get the impression
that it was the only noteworthy form of Mesopotamian numeration. Despite
Neugebauers (1957: 17) warning that the positional numerals are a relatively minor
part of the body of Babylonian numerals, these sexagesimal positional numerals,
used for mathematics, have been assigned priority over much more widespread
systems (Sumerian and AssyroBabylonian). In fact, positional numerals were
used in only a limited set of mathematical and astronomical contexts and over a
much shorter period, serving primarily as a means of easing certain computations
(Robson 2008: 756).
The system uses only two basic numeralsigns, the vertical wedge f for 1 and
the cornerwedge or Winkelhaken g for 10, to write any number between 1 and
59. Thus, small numeralphrases were usually identical to those of the Sumerian
cuneiform system. Nevertheless, certain graphic changes (shown in Table 7.14)
were made to the numeralphrases for 4, 7, 8, 9, and 40, so that, instead of grouping signs in at most two rows of up to five signs, three rows of no more than three
signs were used. This shift eliminated any phrases that placed four or five signs side
by side, and may have increased the systems legibility (Powell 1972a: 16).
Unlike the earlier cumulativeadditive Mesopotamian systems, this system was
cumulativepositional, combining the two basic signs in multiple positions to
250
Numerical Notation
express powers of 60. It was thus a base60 system with a subbase of 10. It had an
additive structure within each power, because of the way that 10signs and 1signs
combine together, but a positional structure among different powers. Just as in
the Sumerian and AssyroBabylonian systems, subtractive notation was frequently
used to write numbers such as 9 (10 lal 1) and 19 (20 lal 1) (ThureauDangin 1939:
106). According to the rules of the system, 4,252,914 would be written as a9
d4 b a4 = (19 603) + (44 602) + (20 60) + 14. In addition to expressing
integers, positional numerals could be used to express fractional powers of 60:
1/60, 1/3600, 1/216,000, and so on.
During the Old Babylonian period, the positional numerals did not have any
sign for zero to indicate an empty position within a numeralphrase, nor was there
any way to distinguish an integer from a fraction (i.e., there was no sexagesimal
point). However, many texts list numbers in columns in which the positional
values of all the numbers are lined up with one another, making misinterpretation
less likely (Powell 1976: 421). When numbers are embedded in the middle of a
text or occur alone, the lack of a zero leads to ambiguity; there is no way, except
through contextual information, to determine which positional value expressed
which power, and thus a single numeralphrase could have an infinite number of
readings. The simple phrase a2 3 could mean 723 (12 60 + 3), 43,380 (12
3600 + 3 60), 12.05 (12 + 3/60), and so on, depending on which positional values
we assume are indicated. When the empty position was both preceded and followed by numerals, this difficulty was sometimes obviated by using a large empty
space to indicate the empty position (Neugebauer 1957: 20). Thus, 1 b (80) could
be distinguished from 1
b (3620). Yet this technique was not used universally, and in some texts what looks to be a large space does not bear any numerical significance. Unless numeralsigns were arranged in columns, there was no
way during the Old Babylonian period to distinguish numbers where the empty
position came at the end or beginning of the numeralphrase. Nevertheless, by
organizing numbers in columns, and through commonsense interpretations of
texts, Babylonian mathematicians would not have experienced insurmountable
difficulties in reading numbers despite these ambiguities.
The Babylonian positional notation probably developed, in fact, in the twentyfirst century bc, during the Ur III (NeoSumerian) period. The late Sumerian
system of weight units is purely sexagesimal and notated in a way that could be
ancestral to positional notation (Powell 1972a: 14). Powell (1976: 420) also found
positional numerals on several early texts, which led him to assert that the development of positional numerals occurred in the twentyfirst century bc at the very
latest. Robson (2007: 7879) discusses twentyfirstcentury texts from the cities of
Umma and Girsu that clearly depict sexagesimal placevalue numerals. The development of the notation may have resulted from Ur III administrative reforms,
Mesopotamian Systems
251
which followed from managing much larger amounts of goods than had previously been the case (Powell 1976: 422; Hyrup 1985: 9). Nissen, Damerow, and
Englund (1993: 142), however, remain agnostic regarding Ur III positional numerals, because most texts can be dated only paleographically, and the numerals do
not show much variation throughout time. I am unconvinced by Whitings (1984)
assertion that the positional numerals developed as early as the Old Akkadian
period (i.e., the twentyfourth or twentythird century bc).
Most of the early texts containing the positional numerals are mathematical
texts of the Old Babylonian tradition, and thus date between 2000 and 1600 bc,
with the majority from the latter part of that period (Powell 1976: 419). These
range from simple multiplication tables and arithmetical exercises to complex
problems that can legitimately be called algebra. Education in the positional notation and in pure mathematics was a significant component of scribal education, although of course only a small proportion of the total populace would
have had access to such training. Figure 7.1 is a small square clay tablet on which
a squaring exercise has been written in Babylonian positional numerals; the two
numbers on the left are each 2, 30 (150), and their product (22,500) is depicted
on the right as 6, 15 the final position, whose value would be 0, must be understood from context (NemetNejat 2002: 262263). Many arithmetical exercises
and texts for translating numerals into the new positional system date from the
Old Babylonian period, indicating the existence of a vigorous process for teaching the system to scribes (Nissen, Damerow, and Englund 1993: 142147). Yet,
because nonmathematical texts did not contain positional numerals, scribes who
did not write mathematical texts probably would not have been familiar with the
positional numerals. The converse is not the case, however; mathematical texts
containing positional numerals are often dated in AssyroBabylonian numerals,
showing that mathematicians also knew the common system (Neugebauer 1957:
17). The numerals were rarely attested to have been used to perform arithmetical
calculations directly, as opposed to writing down results. However, it is possible that calculations were made on clay scratch pads that could be moistened
and rewritten to record results, after erasing the preliminary work (Powell 1976:
420421). It is equally probable that some sort of counting board or wax writing
tablet was used, on which the intermediate steps of calculations were not preserved (Hyrup 2002). We do not know much about the precise steps by which
arithmetical computations were performed.
After the end of the Old Babylonian period around 1600 bc, attested texts
containing sexagesimal positional numerals are extremely sparse for over a millennium. The degree to which there was an actual decline in the systems use is
unclear, however. The small number of postOld Babylonian texts, in combination with the relatively small number of placevalue mathematical texts regardless
252
Numerical Notation
Figure 7.1. An Old Babylonian clay tablet (YBC 7294); the numerals at left each signify
150 (2 60 + 3 10) and the product, 22,500, is at right. Courtesy Yale Babylonian
Collection.
of period, may suffice to explain the decline, or it may have actually waned in use.
Robson (2007: 154156) has published a small number of mathematical texts from
late secondmillennium bc Babylonia that denote numbers using this notation.
In NeoAssyrian texts such as lnam iur ankia, an omen text dating to 712 bc,
a technique called downwards upwards, upwards downwards, playfully altered
sexagesimal numerals by swapping positional registers, so that, for instance,
140 (2 60 + 20) became 22 (20 + 2 1) (Robson 2008: 148).
The positional numerals reemerged during the Seleucid period (the beginning
of which is dated from the Alexandrine conquest of 332 bc) in new contexts and
with several structural differences from the Old Babylonian system. First, subtractive expressions such as 20 lal 1 for 19 were no longer used in Seleucid texts
(Neugebauer 1957: 5). Second, while in the Old Babylonian period the positional
system was used exclusively in mathematical texts, by the Seleucid period it was
used in astronomical texts as well (Neugebauer 1957: 14).
The most important change was the introduction, in certain circumstances, of
a sign serving some of the functions of zero, usually written as 0 or %, to fill in an
empty position within a numeralphrase. This sign could be used at the beginning
of a numeralphrase to indicate that the ones place was empty (i.e., to distinguish
a fraction from an integer) or in a medial position to prevent misreading 3620
Mesopotamian Systems
253
as 80, as in the earlier example, but never phrasefinally (Neugebauer 1957: 20).
Neugebauer (1941) emphasized that the primary role of the zerosign was as much
epigraphical as it was mathematical. He demonstrated that in a small number of
texts, a zerosign was inserted, apparently superfluously, in numeralphrases preceded by an amount in tens and followed by an amount in units. This was done in
order to preclude misreading b 7 (20 60 + 7, or 1207) as b7 (27); by writing
the former as b07, the latter interpretation is prohibited. In such numeralphrases, the zerosign does not indicate an empty position, but simply separates two consecutive positions (Neugebauer 1941: 213). In fact, the zerosign was
originally used to separate sentences, which may indicate that it originally merely
indicated separation or space. It is unlikely that the Babylonians conceived of zero
as an abstract number. No cuneiform positional text contains the bare numeralphrase 0; it always occurs in phrases along with other signs. Thus, 0 was not
equivalent to 0 in the same way that b was equivalent to 20. The abstract concept
of zero accompanied by a special sign for that concept developed independently
among the Greeks and Indians, but probably never among the Babylonians.
Despite its use of the positional principle, the system was restricted to an
extremely small group of Babylonian mathematicians and astronomers in the Old
Babylonian and Seleucid periods. It does not appear to have been known by merchants or most administrators, and certainly did not diffuse to other peoples of the
Middle East, such as the Aramaeans, Phoenicians, and Persians. There is no reason
to believe, as was formerly held by some, that the Babylonian positional numerals
survived long enough to be ancestral to the Indian positional numerals (Fvrier
1948: 585; Menninger 1969: 398399). The only direct descendant of the Babylonian positional numerals is the sexagesimal Greek positional system used by classical mathematicians and astronomers to represent fractions. Around the second
century bc, the Greeks combined the Babylonian system with their alphabetic
numerals to produce a positional, base60 numerical notation system (Chapter 5).
Neugebauer (1975: 590) states that the use of a sexagesimal division of the circle
into sixty parts by Eratosthenes (ca. 250 bc) is the earliest evidence of this borrowing, although Eratosthenes did not use the sexagesimal fractions. This Greek system was used only in mathematical and astronomical texts, and only for fractions.
The quasipositional cuneiform system used in a few texts in the citystate of Mari
also appears to derive in part from the Babylonian positional system.
While the Seleucid astronomical texts are important from the perspective of the
history of science, they represent the work of a limited group of scholars whose
knowledge was being surpassed by Greek mathematics and astronomy even in the
fourth century bc. The Greek alphabetic numerals were those used for everyday
purposes as well as for mathematics, and so, slowly, the Babylonian system fell
into decay. The last example of positional cuneiform numerals dates from the first
Numerical Notation
254
Table 7.15. Mari numerals
1
10
100
1000
10,000
century ad (Powell 1972: 6a). That this system, the first positional system ever
and one much lauded by modern scholars, should be replaced, after having nearly
been abandoned once before by its own inventors, suggests that the advantages of
positional systems do not correlate closely with their survival.
Mari
The city of Mari, located on the Euphrates River at the border of modern Syria
and Iraq, was an independent citystate between the twentieth and eighteenth
centuries bc. During this time, it was engaged in extensive trade relations with
Canaan and Babylonia. A large number of cuneiform tablets have been recovered
from Mari, mainly dating to the eighteenth century bc, upon which a very unusual numerical notation system has been found. This system is shown in Table 7.15
(Durand 1987; Ifrah 1998: 142146).
Below 100, the system is purely cumulativeadditive. For the units, it is identical to the AssyroBabylonian system, but in the tens, there is no special sign for
60; rather, the higher decades are written as , , , and , respectively.
For the hundreds, the multiplicative ideogram i (ME) was often used (preceded
by unitsigns for multiples). However, some texts omit the i sign entirely, turning the system into a quasipositional one. Thus, 476 is written in one inscription
as p\\6 (Soubeyran 1984: 34). The sign for 1000 is a syllabic representation of
LIIM, while the sign for 10,000 is created by superposing the signs for 10 and
1000. For powers above 1000, the system is multiplicativeadditive; there are no
instances where higher power signs are omitted. The omission of the multiplicative sign for 100 is very interesting, but it does not represent a fully positional
system. If it did, we would expect 476 to be written as p\7\6 (4 7 6), not as
p\\6 (4 70 6). The Mari system resembles the experiments with positionality in India in the seventh century ad, before the positional principle was fully
adopted (Chapter 6).
Most of the administrative and commercial texts from Mari use the AssyroBabylonian common system, while mathematical texts mostly use the Babylonian positional numerals. One text (M7857) uses both systems side by side (Robson 2008: 130). We thus know that the Mari scribes understood these systems
Mesopotamian Systems
255
10
60
f or : i
7169=
100
1000
10,000
\<
7;\fi:9
(7
1000 +
1 100 + 60
+ 9)
perfectly well. What may have happened, at least in the case of mathematical texts,
is that the aberrant system was used unofficially (perhaps for calculation) and
then retranscribed for official purposes using the sexagesimal positional numerals
(Soubeyran 1984: 34). It is possible that the decimal structure of the Mari numerals was borrowed from the AssyroBabylonian ones, and that the idea of using
positional notation for the hundreds was taken from the Babylonian positional
numerals. This system should be considered as an aberrant and shortlived experiment with positionality, as the conquest of Mari by Hammurabi in 1755 bc ended
its use.
Hittite Cuneiform
In Chapter 2, I discussed the Hittite hieroglyphic system, which was probably
borrowed from the Egyptian hieroglyphic system or the Linear B (Mycenaean)
system. A separate Hittite script, written in cuneiform characters and related to
the various Mesopotamian scripts, was used at the Hittite capital of Hattusha
between the seventeenth and thirteenth centuries bc. The numerals used alongside this script, shown in Table 7.16, are similar to those of the AssyroBabylonian
common system (Rster and Neu 1989: Table 7).
The system is decimal and cumulativeadditive below 100, while the multiplicative
principle is used for higher powers. Numeralphrases are written from left to right.
To write 60 or multiples of 60, the Akkadian loanword ui (:) was employed,
just as it could be in other cuneiform systems. Yet, when writing numeralphrases
between 70 and 99, a simple vertical wedge represented 60 (Rster and Neu 1989:
271). The sign for 100 is simply the ME syllable of AssyroBabylonian numerals borrowed into Hittite, while the complex and apparently lexical expressions for 1000
and 10,000 appear to be indigenous. Thus, 7169 was written as shown in Table 7.16
(Rster and Neu 1989: Table 7). There is no evidence for the use of subtractive notation in the Hittite cuneiform numerical notation system.
Numerical Notation
256
10
100
604=
6wp
Because the royal archives at Hattusha are our main source for Hittite cuneiform inscriptions, we do not yet know a great deal about the range of functions for
which Hittite numerals were used. We can be quite certain that the numerals were
borrowed from the AssyroBabylonian common system, which was widely used at
the time the Hittite numerals are first attested, given the two systems close similarity in structure and numeralsigns. There does not appear to be any connection
between the Hittite cuneiform and the Hittite hieroglyphic numerals, which are
entirely different in structure. With the collapse of Hittite power in the thirteenth
century bc, the cuneiform numerals ceased to be used.
Old Persian
The Old Persian script was invented early in the domination of the Achaemenid
Empire over Mesopotamia, probably near the beginning of the reign of Darius I
(522 to 486 bc) (Testen 1996). It is an alphasyllabary; thus, while its lettersigns
are cuneiform, it represents a distinct break from the older Assyrian and Babylonian scripts. The Old Persian numerals, shown in Table 7.17, closely resemble the
AssyroBabylonian common numerals (Testen 1996: 136).
The system is decimal and cumulativeadditive, with numeralphrases written
from left to right. The sign for 100 combines multiplicatively with preceding unitsigns in the one attested text containing a large numeralphrase. No known Old
Persian texts have numerals of 1000 or higher. Whereas the AssyroBabylonian
units could be grouped in sets of two or three, in up to three rows, Old Persian
units and tens were arranged in at most two rows, with odd units represented at
twice the size of paired ones. There is no trace of sexagesimal notation in Old
Persian; 60 is expressed with six signs for 10 and 600 as 6 100.
Although the Old Persian script has traditionally been regarded as strictly
a displayoriented script used for prestige purposes related to the Achaemenid
monarchy, the recent translation and interpretation of an Old Persian administrative record from the Persepolis Fortification Archive has cast some doubt
on this notion (Stolper and Tavernier 2007). Figure 7.2 shows an Old Persian
Mesopotamian Systems
257
Figure 7.2. Old Persian tablet from the Persepolis Fortification Archive (Fort. 1208101,
obverse), which begins with the number 604, the only evidence for multiplicative notation
for the Old Persian hundreds. Courtesy of Matthew Stolper.
cuneiform text (Fort. 1208101, obverse) originally excavated in 1933 from Persepolis but published only in 2007, which begins with the numeral 604, probably
denoting a quantity of a dry capacity measure (Stolper and Tavernier 2007: 1213).
This numeral is the highest found to date in any Old Persian inscription; it is one
of only two to contain the sign for 100, and the only one to denote higher hundreds, thus attesting the hybrid multiplicative structuring. As an isolated document among many thousands of Elamite and Aramaic economic and other texts
found at Persepolis, Fort. 1208101 does not prove widespread Old Persian literacy
or numeracy in the Achaemenid period, but it does provide tantalizing evidence
that the scripts use was not as limitedpurpose as was once believed.
The invention of the Old Persian numerical notation system occurred in a context of cultural contact between Persians and Assyrians/Babylonians in the late
sixth century. There are a number of Babylonian/Persian bilingual inscriptions,
and the two systems existed side by side at that time (as evidenced by the thousands of cuneiform documents at Persepolis). The absence of any attested trace of
sexagesimal notation is a unique feature that clearly differentiates this system from
its AssyroBabylonian ancestor, however. While Fort. 1208101 shows that Old
258
Numerical Notation
Persian numerals (and indeed, the script) were used administratively in at least
one instance, the Aramaic numerals (Chapter 3) were the system used much more
commonly for international communication and commerce throughout Persia,
and the AssyroBabylonian numerals were used alongside cuneiform inscriptions
in the Elamite language. The Old Persian system did not survive the Alexandrine
conquest of Persia.
Summary
The commonality among the Mesopotamian numerical notation systems is their
use of cumulative notation with signs for 1 and 10. In other respects, there is considerable variation among these systems, whether due to linguistic (Sumerian vs.
Semitic) or functional (administrative vs. mathematical) factors. The survival of
the sexagesimal base over a period of nearly 3,500 years is testament to the Babylonians preservation of Sumerian traditions, but most of the numerals used after
2500 bc were primarily decimal. The mathematical functions of the various systems, while interesting to historians of mathematics, are minimal in comparison
to their administrative and literary functions.
The conquest of Mesopotamia by the Achaemenids and later the Seleucids sounded the death knell for native Mesopotamian traditions, after which
Aramaic and then Greek numerals were used for most purposes. Even the positional numerals, the hallmark of Babylonian arithmetical achievement, quickly
disappeared under conditions of cultural and political domination. The AssyroBabylonian common system was borrowed and modified in regions of the Middle
East where Mesopotamian influence was strong, but the history of Mesopotamian
numerals is largely linear rather than branching, with each system giving rise to
its successor but not giving rise to many systems outside Mesopotamia. Multiple
systems were often employed at the same time within Mesopotamia, the use of
which was determined by their contexts in ways that remain unclear. With the
sole exception of the Levantine systems, which derive from both Mesopotamian
and Hieroglyphic ancestors, the Mesopotamian systems did not give rise to a large
number of descendants either within Mesopotamia or without.
chapter 8
The East Asian numerical notation systems, like the regions scripts, reflect the
pervasive importance of Chinese civilization over the past three millennia. The
classical Chinese system used from the Qin dynasty (221206 bc) to the present
day, and which spread throughout the region, is foremost in duration and significance. Still, the history of East Asian numeration is one neither of total Chinese
hegemony nor of complete stasis, and these systems are more diverse in structure
than any of the other phylogenies I have investigated. Attested historical connections and similarities in their numeralsigns allow us to identify their common
ancestry. Table 8.1 indicates the most common numeralsigns of the East Asian
systems.
260
1 2 3 4 5 6 7 8 9 0 /
! @ # $ % ^ & * ( )
Kitan
Jurchin
Chinese
positional
10,000
1000
Chinese
commercial
100
10
Late rodnumerals
Rodnumerals
Chinese
classical
Shang /
Zhou
261
f g
q <
:
(
~
)
*
>
o
h
>
i
?
j
$
k
%
n
&
262
Numerical Notation
The sign for 1000 is identical to the Shang character for man, probably due to a
homophonic correspondence (Djamouri 1994: 1516; Martzloff 1997: 180181). By
contrast, the use of identical graphs for scorpion and 10,000 may result from
the association of an immensely large number with swarms of colonies of newborn
scorpions (Martzloff 1997: 181). The highest Shang number expressed is 30,000
(Martzloff 1997: 182). Numeralphrases were written in vertical columns read from
top to bottom, with the highest power at the top.
In almost all the oraclebone inscriptions, numeralphrases are not found alone,
but are accompanied by a character for the object being quantified. On this basis,
Djamouri (1994: 33) regards Shang numeralphrases as determinatives of nounphrases, and argues that each sign was read as a single morpheme in the ancient
Chinese language. Each numeralsign corresponds to a single Chinese morpheme,
an atypical correspondence between language and numerals that leads Djamouri
to regard the Shang numerical notation system as a purely linguistic rather than
a graphic phenomenon. This feature, which it shares with the Chinese classical
system, raises the issue of whether we ought to consider such quasilexical formulations to be real numerical notation systems.
Like the Shang script, the Shang numerical notation system was independently
invented. Needhams (1959: 149) tentative suggestion of stimulus diffusion from
Babylonia rests on the dubious notion that the Shang numerals use placevalue.
Moreover, the correspondence of numeralsign and numeralword suggests that the
Shang numerals have a linguistic origin. If the signs originated to represent Old
Chinese morphemes, this further confirms their indigenous development (Djamouri
1994: 1819). Some of the Neolithic marks on pottery and tortoise shells, such as
those dating from 66006200 bc found at Jiahu in Henan province, resemble
numeralsigns, which could extend this systems history back several millennia further in China (Li et al. 2003). Yet there are no numeralphrases only single signs
among these marks, and in any case there is no way to be sure that their meaning
remained constant. The signs probably have a mixed abstract and phonetic origin;
more important than phonetic correspondences may be the fact that most of them
are graphically quite simple as compared to the other Shang characters. The function of the Shang numerals is quite clear, however, in the context of royal divinatory inscriptions.1 There is nothing resembling a Shang accounting text or commercial inventory parallel to those found in Mesopotamia or Egypt.
After the collapse of the Shang Dynasty, large parts of what is now China were
controled by the Zhou Dynasty, first from its western capital at Hao (1027770 bc)
1
Zhang and Liu (198182) go still further and argue that the oracle bones mark the beginning of the longstanding tradition of bagua milfoil divination in the pattern later
exemplified in the Yijing.
263
10
j
r
l
r
Unattested
n
r
p
r
100
1000
10,000
and then, after the failure of the Western Zhou state, by a more decentralized
Eastern Zhou polity centered at Luoyang (770256 bc). The Zhou kingdoms continued to employ the script and numerals of the Shang. In the early Zhou period,
oraclebone inscriptions continued to be written, but from the tenth to the third
centuries bc, Zhou numerals were often stamped on bronze vessels and coins
(Needham 1959: 5). The increasing complexity of Chinese society over this long
period brought the numerals into use for a much wider range of functions than
is documented to have previously been the case. While the Zhou numerals are
structurally identical to the earlier Shang ones, except that the sign for 10 could
also combine multiplicatively with the unitsigns for 2 through 4, the numeralsigns began to exhibit great graphic variability. Pihan (1860: 10) provides a comprehensive chart showing the various numeralsigns used between the sixth and
second centuries bc, containing, for instance, no fewer than thirtyeight different signs for 10,000. Despite this extraordinary paleographic variability, the signs
shown in Table 8.3 were the ones most commonly found on coins and bronzes until
the third century bc (cf. Needham 1959: Tables 22, 23). The powersigns immediately
ancestral to the classical Chinese ones were among the variants used in the late Zhou
period. These signs, shown in Table 8.4, differ greatly from those in Table 8.3.
By the late Zhou, multipliersigns were no longer superposed onto or ligatured
with the powersigns; instead, numerals began to be written more regularly with
unitsigns preceding powersigns from top to bottom. As well, while the older
signs for 20, 30, and 40 were retained, more typically the signs for 2, 3, and 4 were
placed next to the sign for 10, just as with the rest of the tens. The system was still
Table 8.4. Late Zhou powersigns
10
100
1000
10,000
264
Numerical Notation
In the earliest rodnumerals (fourth century bc to third century ad), the use
of zong and heng numerals as appropriate to their position was not strict, so that
horizontal strokes could be used for ones and vertical strokes for tens. However,
the system had stabilized by the end of the Han Dynasty. No zerosign was used at
265
1s
10s
100s
1000s
J
A
J
A
K
B
K
B
L
C
L
C
M
D
M
D
N
E
N
E
O
F
O
F
P
G
P
G
Q
H
Q
H
R
I
R
I
762
7008
905,920
6.49
G
E
K
Q
O
this early date. Most authors presume that the numeralsigns were lined up strictly
by position, leaving blank spaces as appropriate, obviating the need for a zero;
however, Martzloff (1997: 187) notes that there is limited evidence for such spacing
in written rodnumerals (as opposed to physical countingrods).
The rodnumerals constitute a cumulativepositional system with a base of 10
and a subbase of 5. While it is possible to regard each sign such as R for 9 as a
single sign, thus making this system cipheredpositional, the systems true structure
is best reflected by classifying it as intraexponentially cumulative, which allows us to
recognize how the sign is constituted and to note its subbase. While the numerals
6 through 9 are written using compounds of 5 and 1 through 4, the sign for 5 alone
is always five strokes; if a horizontal stroke were used for 5 in a zong position, there
might be more risk of confusion with the horizontal stroke for 1 in the nexthighest
(heng) position. Because the direction of the strokes alternates with each successive
position, the rodnumerals are irregularly positional,2 since a sign takes its meaning
from both its position and its horizontal or vertical orientation. To put the sign G in
the tens position indicates 70, but to put it in the ones or hundreds position would
have violated the systems structure, except during the earliest phase of its history.
The rodnumeral system was infinitely extendable by using these two alternating sets of numeralsigns in successively higher positions. Decimal fractions could
be written by designating one of the places as the units position, with the places to
the right of that one representing 0.1, 0.01, and so on (Volkov 1994: 81). In numeralphrases containing both whole and fractional positions, the ones position could be
2
Martzloff (1997: 205) coins the term dispositional to reflect this irregularity.
266
Numerical Notation
267
that it is inconceivable that the Indians would not have learned of this system from
the Chinese, and, since it is so practical, they obviously would have borrowed it
(Lam 1988: 104). Yet the Indian positional numeralsigns are those of the earlier
Brhm numerals, not of the rodnumerals, and the rodnumerals have no zerosign (whereas the Indian system does). Moreover, the rodnumerals have a quinary
subbase that the Indian numerals lack, and the rodnumerals are intraexponentially cumulative, whereas the Indian positional numerals are ciphered. No Indian
texts of the period mention rodnumerals or any other Chinese numeration.
In the sixth or seventh century ad, the numerals and rodcomputation were
introduced into Japan at a time when Chinese cultural, religious, and political
influence in Japan was enormous. The original rods were long, thin, round, made
of bamboo, and called chikusaku; they were, however, quickly replaced by shorter
square rods known as sangi (Smith and Mikami 1914: 23). There is no evidence
of their use outside China, Japan, and Korea. The last coins to use rodnumerals
are the five chu coins of the Liang Dynasty (502557 ad), but these numerals are
highly irregular (de Lacouperie 1883: 316317). They continued to be written in
Chinese texts and used directly for computation.
In the twelfth and thirteenth centuries (the late Song Dynasty), the rodnumerals as written in texts though not their computingrod counterparts transformed in three significant ways. Although this was a time of considerable political turmoil in China, due to invasions by groups such as the Jurchin and Mongols,
it was also a time of considerable scientific achievement. Table 8.6 indicates the
system as it was used at that time (Needham 1959: Table 22; Libbrecht 1973: 68).
First, new signs for 4, 5, and 9 were introduced, while the original (cumulative)
signs were retained. Because the only signs to change were those in which four
or five cumulative strokes had previously been required, this was probably done
to simplify the signs, though it meant that the written system differed from that
used with physical rods. These changes made the system less cumulative than it
previously had been, so that it approached a cipheredpositional structure. Second,
written numeralphrases sometimes were condensed into single glyphs, compressing
the individual signs together so that they formed a monogram. Needham (1959: 9)
attributes this development to the requirements of the new technology of printing
books. Finally, a circle was introduced as a sign for zero. The first text known to use
a zerosign is the Shu shu jiu zhang (Mathematical Treatise in Nine Sections) of Qin
Jiushao, published in 1247 (Libbrecht 1973: 69).3 Needham (1959: 10) suggests that
the idea of a circle for zero may have been an endogenous development, based on
the philosophical diagrams of twelfthcentury NeoConfucian scholars. I agree with
3
As I will discuss later, this text is also the first to use a circular sign for zero in conjunction with the classical numerals.
Numerical Notation
268
Table 8.6. Late rodnumerals
Zong
(1, 100, ...)
Heng
(10, 1000, ...)
Old Style
5804
E Q
M
S
D
S
N
T
E
U
R
V
I
W
0
0
New Style
Martzloff (1997: 207), however, that this development was more likely related to the
Indian zero, which had passed to China along with the transmission of Buddhism
in the eighth century ad. We may never know, however, whether the exact route of
transmission was through Southeast Asia, Tibet, or India proper.
The rodnumerals were linked directly with arithmetical computation from the
time of their invention. While they began as a system involving the physical manipulation of rods, Chinese mathematicians quickly adopted them for writing results of computations. The earliest strictly mathematical Chinese text, the Jiuzhang
Suanshu (Nine Chapters on the Mathematical Art), which dates no later than the
first century ad and summarizes the learning of earlier centuries, uses them extensively (Lam 1987: 367368; Li and Du 1987: 3337; Volkov 1994: 81). Thereafter the
rodnumerals were a central part of the computation techniques used alongside
most Chinese mathematical and astronomical texts until the sixteenth century.4
Most Chinese characters having to do with computation use the bamboo radical
because of its association with bamboo computing rods (Needham 1959: 72). Tong
(1999) asserts that overreliance on concrete, contextsituated counting rods and
rodnumerals acted as a stumblingblock preventing the development of propositional mathematics in the Song Dynasty. Yet the modifications to the system, including the addition of a zerosign, suggest that the rodnumerals, as an infinitely
extendable notation using the principle of placevalue, could be used as objects
of arithmetical thought independently of their materiality. Dauben (2007: 191)
contends rightly, in my view that the perfectly regular and decimal character
of countingrod arithmetic greatly facilitated the extraction of roots and advanced
work with linear equations.
4
We may never know the true extent of their use, since many printers considered the rodnumerals, with their vertical lines, to be insufficiently literary, and replaced them with
the classical numerals (Needham 1959: 8).
269
Wang Ling (1955: 91) reports that Chinese logarithmic tables were still written with rodnumerals in the early twentieth century, but I cannot substantiate this assertion.
270
Numerical Notation
and the classical Chinese numerals coexisted for nearly 2,000 years, and yet the
former had no noticeable impact on the latter. If there truly existed a unilinear
trend for positional systems to supplant additive ones, we would expect the rodnumerals to replace the multiplicativeadditive classical numerals entirely, or at
least to facilitate their transformation into a cipheredpositional system.
Chinese Classical
The basic numerals associated with the Chinese script are perhaps the most stable
symbol system currently in use; the numeralsigns of the Qin Dynasty (221206 bc)
are practically identical to those used in modern Chinese literature. While there are
structural differences between that system and the way the numerals are normally
used today, ancient numeralphrases are still easy to read. The basic numeralsigns
are shown in Table 8.7a, and a selection of numeralphrases in Table 8.7b.
In traditional writing, numerals, like the script, were arranged in columns from
top to bottom, with the highest powers at the top. In modern writing, numerals are normally written in rows from left to right, although righttoleft writing
is not unknown, in which case righttoleft numeration is employed. The basic
system is multiplicativeadditive; numbers are written by combining the signs for
1 through 9 with the appropriate signs for the powers of 10 to indicate their multiplication, and then taking the sum of these pairs of signs. There is no powersign
for the units; the unitsigns for 1 through 9 stand alone. When writing 11 through
19, the unitsign attached to 10 is always omitted, although in numbers such as
214 the unitsign for the tens is often included. Prior to the Tang Dynasty, it was
optional to put the unitsign 1 in front of 10, 100, 1000, and 10,000 in numeralphrases, but including the unitsign 1 later became standard (Martzloff 1997: 185).
However, in modern Chinese, the powers of 10 alone can be written without the
unitsign. When the multiplier of a power is zero, both the unitsign and the
powersign are always omitted. There is no zerosign in the classical system,
although there is in modern Chinese numerals (to be described later).
In addition to these standard signs, there are three nonstandard signs used for
20, 30, and 40, which have their origins in the Shang/Zhou cumulative signs for
the lower decades (Needham 1959: 13). These signs were most often used in literary
contexts, for paginating certain texts, and when denoting days of the month. In
the fifthcentury mathematical manuscripts found at Dunhuang in Central Asia,
however, they were also used for mathematical purposes (Martzloff 1997: 185).
They are still used occasionally, although the sign for 40 is very rare because it is
not needed to enumerate days of the month. It was always acceptable (and now is
preferred in most contexts) to use the standard multiplicative combinations of the
unitsigns 2 through 4 and the powersign for 10.
271
or
yi
er
san
si
wu
liu
qi
ba
jiu
10
100
1000
10,000
100,000,000
or
or or
shi
bai
qian
wan
yi
20
30
40
nian
xi
sa
or wan wan
Unlike Western numerals, which are grouped in chunks of three digits, Chinese
numerals are grouped in sets of four, using the character wan (10,000, or, if you
will, 1,0000) (Mickel 1981: 83). Any number from 10,000 to 100,000,000 could
be written by placing a multiplicative numeralphrase from 1 to 9999 before the
sign for 10,000. The system did not stop there, however. By the first century ad,
multiples of 100,000,000 could be written by placing a multiplicative numeralphrase in front of two signs for 10,000 or by using a sign for 100,000,000, either
or (Martzloff 1997: 183). Another technique for expressing large powers of 10,
which developed early in the history of Chinese numeration, involved a complex
Table 8.7b. Chinese numeralphrases
15
10
118
RU
100
74,002
100
10
10,000 4
1000
b7d2c4ba9
100
4,703,600,854
1,072,419
10
7 10,000 2 1000 4
100 10
4a73b6ad8b5a4
4
10
10 10,000 8 100
10
Numerical Notation
272
Sign
Phonetic Value
Lower Series
xia deng
Middle Series
zhong deng
Upper Series
shang deng
wan
104
104
104
yi
105
108
108
zhao
106
1016
1016
jing
107
1024
1032
gai
108
1032
1064
zi
109
1040
10128
rang
1010
1048
10256
gou
1011
1056
10512
jian
1012
1064
101024
zheng
1013
1072
102048
zai
1014
1080
104096
system of powersigns that was assigned three different series of values, as shown
in Table 8.8 (Needham 1959: 87; Martzloff 1997: 99).6 These powersigns combine
multiplicatively with the nine basic unit signs, and thus extend the basic system.
Needham (1959: 87) asserts that these signs first appeared in the Shu shu ji yi
(Notes on Traditions of Arithmetic Methods) of Xu Yue, and dates this text to
around 190 ad, but it may be a sixthcentury forgery (Dauben 2007: 300). In any
event, these techniques are well attested from the fifth century ad onward, and
demonstrate a keen interest in extending the range of numeration far beyond that
needed for any practical purpose. While this system may seem hopelessly complex
and ambiguous, this confusion is identical to that resulting from the different values assigned to billion and trillion in American and European usage. In the lower
series, each exponent is one greater than the one before it; in the middle series,
each exponent (excepting wan) is eight greater than the one before it; and in the
upper series, each exponent is double the one before it. The first sign in all three
series is the standard sign for 10,000, and the second sign (yi) is one of the basic
signs for 100 million (thus corresponding to the middle and upper series, but not
to the lower one). Martzloff (1997: 9799) holds that these systems may have been
6
The middle series (zhong deng) values in Martzloff and Needham differ somewhat; I use
Martzloffs values in Table 8.8.
273
borrowed from similar Sanskrit systems transmitted at the time of the introduction of Buddhism into China in the middle of the first millennium ad. None of
these systems was ever in common use.
The Chinese numerals began to take their modern form starting in the third
century bc, developing directly out of the numerals used in the Warring States
period. With the spread of a unified administrative apparatus under the Qin and
Han Dynasties, they spread throughout areas under direct and indirect imperial
control. The unification of China led to many efforts to standardize the forms
of Chinese script and numeralsigns, although this was not accomplished to any
significant extent until late in the Han Dynasty. Figure 8.1 depicts a Han Dynasty
administrative calendar from 94 bc, found at Dunhuang (Gansu province), with
numerals used to enumerate months and days (Chavannes 1913: Plate XV).
Even as the signs of the system were being codified, however, Chinese writers
began to use calligraphic variants and other modifications of the basic system for
specific functions. These variants used different numeralsigns (ranging from mild
paleographic variations to radically different signs), but their structure is identical
to that of the basic system (decimal and multiplicativeadditive). These variants
were strictly functional, not regional.
Perhaps the most important of these are the accountants numerals (da xie shu
mu zi), which developed in the first century bc (Needham 1959: 5, Table 22). Structurally, they are identical to the classical numerals, but while the classical numeralsigns are quite simple, the accountants numerals were intentionally made very
complex; thus they were considered more elegant and less susceptible to falsification. The signs are homophones of the phonetic values of the appropriate Chinese
words, so they bear no graphic resemblance to the basic signs. Hopkinss (1916)
analysis of their origin as phonetic variations of the standard numerals is dated but
quite thorough. Despite their name, they were used not only for accounting but
also, for instance, on thirteenthcentury coins (de Lacouperie 1883: 318319) and
even in a mathematical text, the Tongwen suanzi qianban of 1614 (Martzloff 1997:
184185). Today, they are still used occasionally on checks, banknotes, coins, and
contracts in order to prevent falsification.
Another highly complex variant of the classical numerals are the shang fang da
zhuan, a variant set of numeralsigns that developed in the Han Dynasty (Pihan
1860: 13; Perny 1873: 113). These numerals are highly stylized linear versions of the
standard numeralsigns that were designed to be used on seals, and are still sometimes used for that purpose today. These signs are shown in Table 8.9.
The diffusion of the Chinese classical numerals was associated with the spread of
Chinese political influence throughout East Asia. In the late second century bc, the
Chinese numerals were employed in tributary regions such as the Gansu corridor in
Central Asia, the Vietnamese states, and the colony of Lelang (modern Pyongyang,
274
Numerical Notation
275
10
100
1000
10000
North Korea). The Chinese numerals were borrowed directly (without any transformation) by the Japanese as part of the kanji characters starting in the third century ad.
The Korean hangul script developed in the fifteenth century has no corresponding
numerical notation system, but Koreans often used the Chinese classical system. The
numeralsigns associated with the chu nm script of the state of Annam (in modern
Vietnam) are simply the basic Chinese signs with additional phonetic notation; the
basic Chinese system was also known and used in the region (Pihan 1860: 2021).
The numerals associated with the scripts of nonChinese peoples of China, such
as Tangut (Kychanov 1996) and Miao (Enwall 1994: 86), are also derived from the
basic Chinese system, although sometimes with considerable paleographic modification. None of these systems is structurally distinct from the basic Chinese numerals. Starting in the tenth century, China began several centuries of intensive contact
with its neighbors to the north and west; warfare with these nomadic groups and
the conquest of China in turn by the Kitan and Jurchin led to the development of
Chineseinspired numerical notation systems among these two groups, which are
structurally distinct and described later.
The classical Chinese numerals were nonpositional and used no zerosign for
over a millennium after their development. The positional principle was known
in China, however, through the cumulativepositional rodnumerals that had
been used since 400 bc. Moreover, Chinese mathematicians became aware of
Indian cipheredpositional numerals in the eighth century ad. Qutan Xida,7 an
IndoChinese Buddhist astronomer working at the Tang capital at Changan, first
reported the use of nine unitsigns with a dot for zero in his great astronomical compendium, Kaiyuan zhanjing, written between 718 and 729 ad (Needham
1959: 12; Guitel 1975: 630631). This transmission reflects the enormous scientific
contact that accompanied the introduction of Buddhism into China in the eighth
century ad. Yet the knowledge of cipheredpositional numerals had no attested
impact on traditional Chinese numeration for many centuries.
7
This name is the Sinicization of the authors original name, Gautama Siddharta.
276
Numerical Notation
In the mid thirteenth century, a period of scientific vigor during the late Song
Dynasty, the first zerosigns appeared alongside classical Chinese numerals in
mathematical texts. The first such text was the Shu shu jiu zhang of 1247, the same
document in which the zerosign is first found with rodnumerals (Libbrecht
1973: 69). This modification allowed a circular zerosign to be used whenever one
of the decimal powers in the middle of a numeralphrase was empty. In theory, this
allowed Chinese mathematicians to use only the unitsigns from 1 through 9 in
conjunction with the 0 to express any number thus making the system cipheredpositional. Yet, during the Song Dynasty zero was used only to fill in empty medial
positions, while retaining the powersigns, so that where 12,001 would be written in the classical style as , it is written as in the
Shu shu jiu zhang, a less concise form that provides no other obvious advantage.
Whether this resumption of the use of the zerosign was a result of the continuation of its eighthcentury use, or a reintroduction from India or the Middle
East, is unknown. Starting in the late sixteenth and early seventeenth centuries,
when Chinese mathematicians of the Ming Dynasty were in extensive communication with the West through the intermediary of Jesuit missionaries, this form
of cipheredpositional Chinese notation was employed more regularly (Martzloff
1997: 185). Tables of logarithms appeared at this time, using the nine traditional
unitsigns and a circle for zero in a way identical to the use of the Western signs
0 through 9 (Menninger 1969: 461).
Before the sixteenth century, zero was employed only in mathematical and scientific texts. In the late Ming Dynasty, it began to be used more widely, but rather
than using the circular sign for zero found in the Song texts, a character, ling
() raindrop, which had been used to designate remainders in division, was
assigned the meaning zero. The first text in which it featured prominently is
Cheng Daweis Suan fa tong zong (Systematic Treatise on Arithmetic) of 159293,
which is also the first text to describe the Chinese commercial numerals or ma zi,
and additionally contains the first complete description of the beadabacus or suan
pan (Needham 1959: 16, 7578; Li and Du 1987: 185187). In this and other early
texts, ling was used in exactly the same way as the circlesign had been used previously, with one ling sign for every missing power, so that 30,008 would be written
as (3 10000 0 0 0 8). While the ling sign introduced an element
of positionality into the system, it was not fully positional, since the powersigns
were retained, and ling was used only in medial positions. Chinese writers soon
realized that they could omit all but one ling when multiple consecutive powers
are empty, so that one could write 30,008 simply as (3 10000 0 8). The
classical Chinese system normally uses ling in this manner today.
In modern China, any given number can be expressed in no less than six distinct ways, the choice of which depends greatly on context. Four of these forms are
variants of the classical system. For literary and prestige purposes, the pure classical
277
2d4b6 2d[4b[6
Cipheredpositional
Western
20406
20,406
Chinese numerals (without any sign for zero) are used, thus representing continuity of the system from the Qin Dynasty to the present. In most ordinary prose
writing, some sign for zero is usually introduced in the medial positions, while
retaining the powersigns. The use of ling has even spread to spoken Chinese, so
that the preferred way to say 203 is not simply er bai san but rather er bai ling san.
Where conciseness or clarity is desired, and in most scientific contexts, the nine
unitsigns along with a sign for zero are used positionally, as in the seventeenthcentury logarithm tables. In contexts where there is concern with forgery, the
accountants numerals can be used. The final two options are to use the commercial or Hangzhou numerals, which I will describe later, or Western numerals.
The Chinese classical numerals are ancestral to several numerical notation systems.
The cipheredpositional variant Chinese numerals used in modern mathematics are,
of course, one such descendant, as are, more indirectly, the Hangzhou numerals. The
structurally distinct numerical notation systems used by the nonChinese Kitan and
Jurchin during the period in which they were in contact with (and later dominant
over) the Chinese are also largely derivative of the Chinese classical numerals. Finally,
two more obscure systems, the shochuma numerals used on wooden tallies on the
Ryukyu Islands, and the Pahawh Hmong system developed recently for use among the
Hmong of Laos, may also be derived in part from the Chinese system (Chapter 10).
Western numerals, while known in China from the seventeenth century, were
not widely used until the beginning of the twentieth century; the Shuxue wenda
of 1901, an arithmetical primer for use in elementary schools, was one of the earliest such texts (Martzloff 1997: 3536). In most scientific and technical contexts in
China today, however, Western numerals are preferred. Because the cipheredpositional Chinese numerals with the circle for zero had been used for several centuries
prior, this shift was strictly social and political, unrelated to structural considerations. Mao Zedong was amenable (at least initially) to the replacement of Chinese
numerals by Western numerals, as indicated in a 1956 speech that was later suppressed (DeFrancis 1984: 262263). Nevertheless, the replacement of Chinese with
Western numerals has not been uninterrupted or uncontested. Some institutions
reacted sharply to this trend, and antiWestern sentiment led to the replacement
of Western numerals by the corresponding Chinese numerals in certain academic
publications (DeFrancis 1984: 274275). Western numerals are well known to all
reasonably educated people in China. In Japan and South Korea, the dominance
of Western numerals is considerably greater than it is in China. Nevertheless, the
Chinese numerals continue to be known and taught in these countries, though
278
Numerical Notation
they are acquiring an archaic flavor. In China itself, however, the use of local numerals shows no sign of sharp decline, and there is every reason to believe Chinese
numeration will persist into the foreseeable future.
Chinese Commercial
The Chinese commercial numerals (often known as Hangzhou numerals)8 arose
in the sixteenth century. The numeralsigns of the system are shown in Table 8.11
(Needham 1959: Table 22).
Comparing these signs to those in Table 8.6, we see that all of the unitsigns,
save that for 5, closely resemble the late forms of the rodnumerals used during
the Ming Dynasty, although they have been borrowed haphazardly from the zong
and heng forms. The unitsigns for 1, 2, and 3 use vertical rather than horizontal
lines, showing that they are unrelated to the classical Chinese numerals. Hopkins
(1916: 318) explains the aberrant form of 5 as a form of the character wu, which is a
homophone of the Mandarin numeral word for five. The most common versions of
the powersigns for 10 through 10,000 are obvious variants of the classical systems
powersigns. The circular sign for zero was in use in both the rodnumerals and
the classical system. This evidence strongly suggests that the commercial numerals
originated as a blend of the late rodnumerals and the Chinese classical numerals.
The system is multiplicativeadditive, with the zero used only to fill in empty
medial positions, never at the end of numeralphrases. Unlike the Chinese classical
numerals, commercial numeralphrases place the signs in two rows, with the unit
multipliers of the various powers on the top row and the powersigns, zerosigns,
and the signs for the ones position on the bottom row (Pihan 1860: 6). Numeralphrases were thus read in a zigzag fashion, starting at the top left, proceeding from
top to bottom and then diagonally up and to the right.
This basic system was made more complex by a large number of irregularities.
When the number being expressed was a simple multiple of a power of 10 (e.g.,
50, 800, 2000), the multiplier usually was placed to the left of the powersign (as it
would be in the classical system) rather than above it (Perny 1873: 101). When the
number 10 occurred alone or in numbers such as 610 and 2010, the unitsign 1
was always omitted, and the unitsign could optionally be omitted when the sign
for 10 was combined additively with unitsigns, as in numbers such as 18 and
212. Moreover, the special classical Chinese numeralsigns for 20 (,) and 30 (.)
could be used in the commercial numerals where appropriate (Hopkins 1916: 319).
When there were two consecutive zerosigns in a numeralphrase, they could be
placed one atop the other rather than side by side in the bottom row, as would
8
Other names for this system include ma zi, Suzhou numerals, and hua ma.
279
10
100
1000
or
w y
or
or
10,000 0
40,709
26
162
917
3008
5000
S\\ G
z0x0t
K
vF OR ,F
JF
JF
xaK OR xag
tJ
t
xvG OR xvG
L0
y0H
sy
10,000
2
10
1
6 OR
6
100 10
9
1
100
3
1000
20
2
9
10
0
1000 0
100
OR
100
10
280
Numerical Notation
as 1593. Yet early texts that mention them associate their invention and use with
the great commercial city of Suzhou (in Jiangsu province). As this city came to
prominence only in the sixteenth century, if the attribution of their invention to
Suzhou is correct, a presixteenthcentury origin is unlikely.
As is suggested by their name, the commercial numerals were (and are) used solely
in commercial contexts. They are still used even today in some regions on bills, invoices, and signs in shops and markets (primarily to indicate prices), though their
use is waning in favor of regular Chinese numerals or Western numerals. They are
most common in regions where Cantonese is spoken, including Hong Kong.
Kitan
The Kitan (or Khitan) were an Altaicspeaking people who ruled Manchuria and
other parts of northern China between 907 and 1125 ad, a period now known as
the Liao Dynasty (Kara 2005: 7). While there was no Kitan writing before their
conquest of Manchuria, two scripts were developed shortly thereafter, the large
script and the small script, both based largely on Chinese, and possibly also
under the influence of the Central Asian Uygur script. Neither Kitan writing
system is fully deciphered, because the Kitan language is only poorly known, but
the meanings of the Kitan numeralsigns are understood. The numerals of the
large script are identical to the classical Chinese numerals, but the small script,
purportedly developed by the Kitan scholar Diela during the visit of an Uyghur
delegation to the Kitan court in 924 or 925 ad, had a distinct numerical notation
system. The attested signs of this system are shown in Table 8.13 (Kara 1996: 233).
While the Kitan numeralsigns have a vaguely Siniform appearance, they are
entirely dissimilar to the corresponding Chinese numerals, and must be of indigenous origin. Numeralphrases are multiplicativeadditive and are read vertically
in rows from top to bottom and then right to left across the page, as in traditional
Chinese writing. A slight ciphered element is introduced into the system by the
existence of distinct characters for 20 and 30; this practice is probably derived
from the analogous Chinese signs, and , although the Kitan signs are not
cumulative. It is not known how (if at all) numbers higher than 1000 were written.
Numeralsigns could also serve as phonograms; for instance, the symbol for 5 (tau)
was employed homophonically in the word taoli hare (Kara 2005: 13).
Because Kitan writing is so poorly understood, it is difficult to know the total
scope of contexts in which the numerals were used. Kitan texts are known from
epitaphs on royal tombs, a text on a bronze mirror, some other stone monuments, and inscriptions on seals and vessels (Kara 2005: 9). Most texts were probably historical records of events, in which numerals are used primarily for dating.
The Kitan script and numerals did not outlast the period of Kitan independence,
281
1
10
100
473 =
1
0
/
2
,
3
.
4
/
7
0
3
which ended in 1125 at the hands of the Jurchin. In 1191, the use of the Kitan script
was forbidden by Chinese imperial order, after which time no further Kitan texts
are attested (Kara 1996: 231).
Jurchin
The Jurchin (also Jurchi or Jurchen) were the rulers of what is now known as the Jin
Dynasty in the northern part of China (11151234), and one of the groups constituting the Manchu who conquered China in the seventeenth century. The Jurchin
script, which consists of logograms and syllabograms in addition to a set of numeralsigns, is attested from inscriptions and texts from the twelfth through fifteenth
centuries, but may have developed somewhat earlier (Kiyose 1977). The Jurchin
numerals are shown in Table 8.14 (Grube 1896: 3435; Kiyose 1977: 1323).
Jurchin numerals, like the script, were written in vertical columns read from
top to bottom, with the highestvalued powers at the top. The system is primarily
decimal; although the distinct numeralsigns for 11 through 19 suggest a vigesimal
component, it is a product of the fact that the Jurchin lexical numerals have distinct
words for 11 through 19 that are not connected to those for 1 through 9, but the
irregularity extends no further than the teens (Kiyose 1977: 133). Because the Manchu language, in contrast to Jurchin, had no such words, Jurchin numeralphrases
could also be written using the sign for 10 additively with the signs for 1 through 9.
For writing numbers from 20 to 99, unitsigns from 1 through 9 sometimes were
combined with the powersign for 10 as in the classical Chinese system, so the
Jurchin numerals appear to be multiplicativeadditive. Yet there were also ciphered,
nonmultiplicative Jurchin numeralsigns for 20 through 90. For numbers above
Numerical Notation
282
Table 8.14. Jurchin numerals
1
HPX
XZH
LODQ
&
GXZLQ
XQD
QLQJX
QDGDQ
11
12
13
14
15
16
17
10
DNXQ
X\XQ
XZD
18
19
20
30
40
50
60
VXVDL
JXLQ WHKL
70
80
90
100
QLQMKX
1000
10,000
PLQJDQ
WXPDQ
100, the multiplicative principle was always employed. Thus, the Jurchin system
is structurally closer to hybrid cipheredadditive/multiplicativeadditive systems,
such as the Ethiopic numerals (Chapter 5) and Sinhalese numerals (Chapter 6),
than it is to Chinese. In the SinoJurchin texts from the Ming Dynasty published
by Grube (1896), which date roughly to the period 14501525, only the unitsigns 1
through 9 and the power signs 10, 100, 1000, and 10,000 were used.
A Jurchin large script was introduced in 1120 by Wanyan Xiyin, and was based
on the Kitan script with significant Chinese influences; the script was officially introduced in 1145 by Emperor Xizong, with a number of small script characters
added (Kara 1996: 235). The Jurchin numerals are found on many monuments of
the Jin Dynasty and some manuscript fragments. The writings that survive are
historical and literary in nature, and the numerals in them are mainly dates. Our
best evidence for them comes from the Ming Dynasty (13681644), when Chinese
translators produced a bilingual glossary and translated documents, in which the
numeralsigns just described are found (Kara 1996: 235). The earlier signs differ
paleographically but not structurally from the Ming ones.
Although the Jurchin did not control large regions of China for very long, the
Jurchin script survived for several centuries. It was used on a Ming inscription of
1413, suggesting that it was not simply a historical curiosity, but was being preserved because it was being used (at least by some people). It continued to be used
until at least 1525, at which time Ming translators were still working with Jurchin
documents. The Jurchin were one of the major constituent groups of the Manchu who conquered China in the seventeenth century (in fact, the ethnonyms
Jurchin and Manchu may refer to a single group), but by this time they used
283
Summary
Chinese numerals are central to the history of the East Asian systems. Today, the
classical Chinese numerals (along with positional variants) occupy a role parallel to
the supremacy of the Roman numerals in Europe prior to 1500, despite the increasing use of Western numerals for science, technology, and commerce. This systems
continued strength (at least in China) suggests that it will continue to thrive, especially in nontechnical prose writing. We must also take into account the strong
cultural preference for Chinese symbol systems when analyzing the present state of
the Chinese numerals; functional considerations alone cannot account for it. The
increasing rarity of the Chinese numerals in Japan and Korea represents not only
the functional rejection of an inefficient system, but also resistance to a Chinese
cultural feature in favor of the more international Western numerals.
The Chinese numerical notation system as used today is enormously variable
in structure, and employs a host of representational techniques. On the surface,
this appears hopelessly nonfunctional, and we might question why such a system
would survive. I think that its quasilexical nature the fact that Chinese numerals act as both ideographic scriptsigns and graphic numeralsigns renders this
variability both comprehensible and rational. If Western numerals incorporated
archaisms such as score, or accounted for the fact that 1400 can be one thousand
four hundred but is more commonly fourteen hundred, similar eccentricities might
arise. The Chinese classical numerals are well suited to being read because they account for the irregularities in spoken Mandarin. The basic multiplicativeadditive
structure of the system permits all sorts of structural manipulations, such as the
occasional use of positionality or ciphered signs for the lower decades, without
creating any ambiguity. The systems flexibility and its correspondence with language are thus advantages rather than hindrances.
The comparison of this phylogeny to the ones I have discussed previously is
quite instructive. In Chapters 2 through 7, most systems employed a single common structural principle. By contrast, the East Asian systems display considerable structural variety: cumulativepositional (rodnumerals), cipheredadditive
(Jurchin), cipheredpositional (Chinese positional variant), and multiplicativeadditive (Shang/Zhou, Chinese classical and commercial, Kitan). Yet there can
hardly be any doubt that these systems comprise a cultural phylogeny. The historical connections among systems are well established, and the similarities in the
numeralsigns are quite strong. If we were to rely on structural qualities alone, we
would be at a loss to describe their cultural history.
chapter 9
Mesoamerican Systems
BarandDot
The baranddot numerals were the most commonly used system in lowland
Mesoamerica, both on stone monuments (400 bc910 ad) and the four surviving
Maya barkpaper codices (10001500 ad). This system has long been an object of
1
I treat other, unrelated New World inventions, such as the Inka khipu and the Cherokee
numerals, in Chapter 10.
284
Mesoamerican Systems
285
Baranddot (monumental)
V e
V E
V
T
U
Baranddot (codices)
Aztec
Texcocan lineanddot
20
400
8000
2
1
X v Y x y
V
3\9
0\]
study (Bowditch 1910, Morley 1915). While used in all the lowland Mesoamerican
polities, it is most commonly associated with the Maya. Baranddot numerals are
ubiquitous in most lowland Mesoamerican texts, reflecting both the strong interest in dating and calendrics and the practice of incorporating numerical values
into the names of deities and individuals.
The numbers from 1 to 19 are written by combining a dot sign for 1 and a bar sign
for 5 additively. When the bars are vertical, as is most common on stone inscriptions, they are usually placed to the right of the dots, but they are placed below the
dots when the bars are horizontal, as in the codices and a few monumental inscriptions, particularly early ones (Table 9.2). Short numeralphrases such as these were
normally combined with another glyph indicating the thing being enumerated,
most often time periods. While the primary and original function of the signs was
numerical, some baranddot numerals could also be used syllabically in the Maya
script (though not, as far as we know, in any of the other Mesoamerican scripts). For
instance, four dots could mean 4, but could also indicate near or partial homonyms
of Classic Maya kan four, such as kan sky, height and the first part of kanhan
haughty (Macri and Looper 2003: 262). Mesoamerican hieroglyphic writing on
stone was a very ornate art, and numerals could be altered or ornamented in various
ways that can make reading a numerical value difficult. Ornamental crescents were
often employed in order to fill in a numeral that would otherwise have an empty
space, and these can easily be confused with dots (Thompson 1971: 130). Similarly,
decorative lines were sometimes added to bars for aesthetic purposes, which can
make it difficult to distinguish one from two bars.
In addition, in Maya monumental inscriptions and also (with a different form)
in Maya codices, a sign for 20 was also occasionally used, producing a base20
cumulativeadditive system with a subbase of 5. In addition to the numerical value
of 20, it is also a glyph meaning moon or lunar month (Lounsbury 1978: 764).2
2
Closs (1978: 691) notes that the central dot in the latter of these signs is found only on
inscriptions where the glyph has the numerical value 20, thus distinguishing it from the
more generic moon, where the dot is missing.
Numerical Notation
286
20/moon
(monumental)
20/moon
(codices)
20/moon (?)
(Isthmian)
3\ 9
0\]
0/completion
(monumental)
0/completion
(codices)
The sign for 20 can occur on its own or in conjunction with baranddot numerals from 1 to 19, thus representing numbers as high as 39. However, it is never
repeated in a numeralphrase; that is, one would not write 60 with three 20signs.
The accompanying baranddot numerals could be placed above, below, or to
either side of a 20glyph (Kelley 1976: 23). The 20sign was mainly used to indicate
intervals between dates between twenty and thirtynine days, thus avoiding the use
of combinations of uinals (periods of twenty days) and kins (one day) (Thompson
1971: 139). Very rarely, it was used in expressions for longer time intervals, such
as the irregularly constructed date on Stela 5 at the Maya site of Pixoy, indicating
a quantity of 20 tuns (periods of 360 days) (Closs 1978). In a few instances, to be
discussed, the 20glyph was used for noncalendrical counts as well.
A sign for zero also accompanied the baranddot numerals. There is considerable paleographic variation in the signs used, but a shell sign was commonly
used in the codices, while different signs were used in monumental writing. The
Mesoamerican zerosign is not completely synonymous with its Western counterpart; normally it served as a placeholder within dates, with the rough meaning of
completion of a given cycle of time. While this raises the issue of whether we
should regard this sign as meaning zero at all (Thompson 1971: 137), which I will
discuss later in greater detail, I do not see any reason to deny the Maya their zero.
The Maya zerosign is definitely numerical in function; it is found in the same
contexts as the regular baranddot numerals, but not normally elsewhere, and so
the meaning zero is quite appropriate. While the Maya probably did not have a
concept of zero as a whole number, as is present in Western mathematics, neither
did Seleucid Babylonian astronomers (Chapter 7), for whom the zerosign served
as the marker of an empty medial position.
While baranddot numeration is most closely associated with the Classic period (ca. ad 150900), it developed centuries earlier, in the latter part of the Middle
Mesoamerican Systems
287
Formative period (1000400 bc). While Macri and Looper (2003: 4) insist that
writing on perishable materials must have preceded the tradition of carved writing
on monuments and portable objects, there is no iron law of script development
that requires this to be true. Baranddot numerals are among the first identifiable
signs of Mesoamerican writing systems, and occur in all three of the major Formative script traditions: Isthmian (epiOlmec), Zapotec, and Maya. All three of these
scripts survived into the Classic period, but the Maya inscriptions are by far the
most numerous and significant. Understanding the early history of baranddot
numeration, however, does not give clear priority to any of the three.
The poorly understood Isthmian (or epiOlmec) script, known from a handful
of inscriptions starting in the Late Formative period (400 bc to 50 ad), is associated with the latter stages of the Olmec civilization along Mexicos Gulf Coast.
Recent claims would give Olmec writing a much longer and complex history,
however. The Cascajal block, which appears to date to the early first millennium
bc and to be associated with the Olmecs, does not contain baranddot numerals
(or any other apparent numerals), but as it is a unique artifact, its relevance to
the history of Mesoamerican numeration and writing remains unclear (Rodriguez
Martinez et al. 2006). There is apparently no connection between the Cascajal
block and Isthmian or indeed any other Mesoamerican script. The San Andrs
cylinder seal, found near the Olmec center of La Venta and dating to the seventh
century bc, is asserted by its discoverers to contain the personal name King Ajaw 3
using stylized dots (Pohl, Pope, and Nagy 2002).
Isthmian writing itself is best known from Late Formative texts from the first
century bc to the second century ad, a period when Olmec political fortunes
were on the wane. In some of these texts, baranddot numerals were arranged
for the first time in vertical columns to indicate periods of time and specific dates
in the famous Long Count system, the calendar expressing dates as a series
of numerals indicating five time periods. We can thus assign absolute dates to
these artifacts from the calendrical evidence alone. The earliest of these is Stela
2 (actually a fragment of a wall panel) from Chiapa de Corzo, dating to 36 bc;
similarly, Stela C from Tres Zapotes has a Long Count date corresponding to 31 bc
(Marcus 1976: 4953).3 The longer secondcentury Isthmian inscriptions, such as
the La Mojarra stela 1 (156 ad) and the Tuxtla statuette (162 ad), contain baranddot numeralphrases including Long Count dates (Justeson and Kaufman 1993).
The Long Count date 8.5.16.9.9 (corresponding to July 16, 156 ad) occurs on the
La Mojarra stela, with standard baranddot numerals arranged vertically in a single column. The stela also contains a sign identified by its decipherers as meaning
3
As with all such dates, there is a small possibility that they were inscribed at a date later
than the one given textually.
288
Numerical Notation
20 or moon, parallel to the later Maya practice (Justeson and Kaufman 1993).
However, aside from the numerals, much about the decipherment of Isthmian is
hotly contested, with Justeson and Kaufman (1993) arguing that their decipherment is well under way, but Houston (2004b: 297298) arguing that because of
the contextualized nature of the system, the script may be indecipherable.
The Zapotec civilization of the Valley of Oaxaca in southern Mexico began to
rise to ascendance in the late Middle Formative period, and developed a script
tradition quite distinct from Isthmian. The earliest wellattested Zapotec numeral
is found upon Monument 3 from San Jos Mogote (600500 bc) in the Valley of
Oaxaca, where the dayname 1 Earthquake is written with a stylized dot (Marcus
1976: 4445). If the Isthmian San Andrs cylinder seal is misdated or nonnumerical,
then this inscription is the earliest attested instance of baranddot numeration.
Stela 12 from Monte Albn provides the first example of a combined baranddot
phrase for 8, apparently indicating a day of the Zapotec month (Marcus 1976:
4546). Both of these are Middle Formative and definitively Zapotec rather than
Olmec, and their early date points to Oaxaca as a potential region of origination
for the system. Colville (1985: 796), however, is agnostic as to whether the baranddot system was invented by the Olmec or the Zapotec, since in his analysis,
both used vigesimal lexical numerals with a quinary component, a structure common also to the baranddot numerals.4
While it was once widely held that the Maya tradition was a latecomer in the
history of Mesoamerican writing, there is some evidence that the Maya tradition
may have emerged alongside the Isthmian and Zapotec scripts, perhaps as early
as 400 bce. Monument 1 from El Portn is an extremely eroded stela with one
partially readable column of glyphs that may be ancestral to later Maya glyphs,
including numerals. Although the inscription has no date, archaeological evidence
places it in the late Middle Formative (Harris and Stearns 1997: 122). This raises the
possibility that the three Mesoamerican traditions were essentially contemporaneous, and may have emerged in a context of economic and diplomatic exchange.
Yet there is a distinct scarcity of Maya or Mayaancestral texts from the Middle
and Late Formative. Starting in the second and third centuries ad, the Isthmian
and Zapotec traditions began to wane, and Maya inscriptions predominate in the
Mesoamerican lowlands thereafter. The first certain Maya inscription that uses
baranddot numerals is Stela 29 from Tikal, which dates to 292 ad (Lounsbury
1978: 809); however, Stela 5 from Abaj Takalik, which dates to 126 ad, may also be
an early Maya inscription (Closs 1986: 327).
4
Colville accepts the idea that the Olmecs spoke a MixeZoquean language, whose modern speakers have numerals of this structure; this is not a universally accepted conclusion,
and in any case the numerals almost certainly changed in the intervening millennia!
Mesoamerican Systems
289
Numerical Notation
290
kin
1 day
uinal
20 kins
tun
18 uinals
katun
20 tuns
baktun
20 katuns
2001: 3). The last text on which baranddot numerals occur is one of the books
of Chilam Balam, in which an annotated description of the system is dated 1793
(Thompson 1971: 130). Yet the system essentially had ceased to be used by 1600
and was replaced by Roman or Western numerals.
The terms katun and baktun mean, literally, 20 tuns and 400 tuns. The latter term is
in fact a coinage of Mayanists; there is no evidence that this word was associated with
the glyph in question in ancient times. There are several extremely rare glyphs for longer
periods, again with coined names: pictun (8000 tuns), calabtun (160,000 tuns), and kinchiltun (3,200,000 tuns), which presume a purely vigesimal progression of dates (Closs
1986: 303).
It appears, however, that the Yucatecan and Cakchiquel Maya may have had a purely
vigesimal year of twenty months of twenty days, though their numerical notation does
not reflect this fact (Satterthwaite 1947: 89).
Mesoamerican Systems
291
i%
j#
l!
n$
9@
(14 katuns)
(0 uinals)
exclusively read from left to right and top to bottom. These expressed the
amount of time between the starting point of the Maya calendar (corresponding to the date August 10, 3113 bc, in the widely accepted GoodmanMartinezThompson correlation with the Gregorian calendar) and any other date. In
addition, the amount of time between any two days could be expressed by a
Distance Number, such as 12 tuns, 0 uinals, 4 kins. Modern scholars use a
convention whereby time values are expressed by writing the five numerical
coefficients separated by points; thus, the date shown in Table 9.4 would be
written as 9.14.10.0.12.
For both Long Count dates and Distance Numbers, if the coefficient of a time
period was zero (e.g., 0 uinals in Table 9.4), the Maya included both a zero coefficient and a period glyph for that value, even though it was not logically necessary
to do so in order to interpret the phrase correctly. While it is not known exactly
why the Maya did this, it was probably for aesthetic reasons. Only occasionally
in Distance Numbers (though never in Long Count dates), a period with a coefficient of zero was suppressed (Thompson 1971: 139).
In a few texts, periodglyphs were omitted entirely, and dates were written
simply by placing the five coefficients in a single vertical column. As mentioned
already, the technique was present in the Isthmian and Zapotec inscriptions by
the first century bc, and continued to be used by the Preclassic Maya (Marcus
1976: 4957). Although it was largely abandoned thereafter, Stela 1 at Pestac
contains a date (9.11.12.9.0) written in this format, which corresponds to 665 ad
(Closs 1986: 326327). Most other Maya inscriptions include all the periodglyphs, although sometimes the glyph for the last position (kins) was omitted
(Closs 1986: 308). Our best evidence for the omission of periodglyphs comes,
however, from the Dresden Codex, a Postclassic text that was probably written in the early thirteenth century, though it may be a copy of a much earlier
document (Marcus 1976: 35).7 It is the most astronomically sophisticated of the
surviving Maya texts, and contains more vertical columns of numbers than any
7
One set of five numbers without period glyphs is found on the fifteenthcentury Madrid
Codex that may qualify, but otherwise no other codices have them.
Numerical Notation
292
I
N
J
]
L
9 baktuns
14 katuns
10 tuns
0 uinals
12 kins
other. Table 9.5 shows the Long Count date 9.14.10.0.12 as it would be written
in this manner.
This system requires that all the relevant numerical coefficients be included,
even for periods for which there is a zero coefficient, to ensure that the correct
quantity of time is counted. The bottom value always represents kins, the second
from the bottom uinals, and so on, preventing any misreadings. Because these
units of time are arranged in a mainly vigesimal sequence each higher value
is equal to twenty of the next lower value, except the tun of eighteen uinals
Mayanists today agree that this system of writing dates is a base20 cumulativepositional numerical notation system with a subbase of 5 (Kelley 1976, Marcus
1976, Lounsbury 1978).
If so, then when the Maya wrote number columns such as the one in Table 9.5,
each position must have represented a particular component of a single number. Positional numerical notation systems do this by having each successive position represent the next higher power of a base. Thus, when I write the number 1942, I mean a
single count of some quantity, of which there are 1942, consisting of one thousand,
nine hundreds, four tens, and two ones. In the Maya case, where the lowest unit
expressed is kins, it is quite natural to assume that it counts kins. It is easy to translate
the five periods into counts of days and then to take the sum, as in Table 9.6.
However, if the periodglyphs were meant to be inferred when reading these
columns, then such numerals ought to be read as five separate values, each no
greater than 19, just as they would be if the glyphs were included. How, then, can
we tell whether the interpretation in Table 9.6 is one that the Maya themselves
made, or whether they simply read in the missing periodglyphs? How do we
decide whether the correct interpretation is 1,400,412 kins or 9 baktuns, 14 katuns,
10 tuns, 0 uinals, 12 kins?
If baranddot numerals were used for large quantities of things other than
time, we would have clear instances where the higher positions represent powers
Mesoamerican Systems
293
I
N
J
]
L
9 144,000 days
1,296,000 days
14 7200 days
100,800 days
10 360 days
3600 days
0 20 days
0 days
12 1 day
12 days
= 1,400,412 days
of a base, rather than large calendrical periods. Yet no Mesoamerican texts use
positional baranddot numerals to count noncalendrical amounts. One is
struck, upon comparing Maya inscriptions to those of any other civilization, by
the virtual absence of phrases indicating large quantities of captives taken in war,
goods paid in tribute, wealth owned by individuals, or any other noncalendrical
quantity. In the rare instances where the Maya wrote numbers of other quantities above 19, sometimes they used additive techniques, such as the moonglyph
for 20, which is used in counts of 20 and 21 captives, but this does not allow the
writing of very large numbers (S. Houston, personal communication). In other
cases, it is possible that multiplicative techniques were used. Houston (1997) suggests that on a mural from Bonampak, a bar numeral for 5 was combined with a
glyph, pi, which may have meant unit of 8000 cacao beans, producing a quasinumerical expression for a count of 40,000 cacao beans (Houston 1997). The
Yucatecan, Cholan, and Tzeltalan languages all use numeral classifiers linguistic
particles that obligatorily follow lexical numerals and indicate the thing being
counted (Berlin 1968; Bricker 1992: 7173). Macri (2000) contends on this basis
that the periodglyphs kin, uinal, tun, katun, and baktun, as well as any other
glyphs that follow numbers, should best be interpreted as numeral classifiers.8 If
this interpretation is correct, then by analogy with the calendrical system, the
Maya likely expressed large noncalendrical quantities by combining baranddot
numerals with a sign for a metrological unit that may have been a multiple of
some smaller unit but was not expressed in terms of that unit (just as 1 tun = 360
kins but was not expressed in such terms). These different means of writing larger
8
For Macri, this also explains why the early epiOlmec and Zapotec calendrical inscriptions written in MixeZoquean languages do not use period glyphs but simply series
of baranddot numerals.
294
Numerical Notation
Mesoamerican Systems
295
tuns (and multiples thereof ) occupy a special role in the Maya calendrical system
(Closs 1983). Finally, the glyphs for katun and baktun often show an affiliation
to the basic tun sign. Yet there is no reason to think that the Maya wrote glyphs
for the baktun and katun but then simply ignored them in reading. As an analogy,
the English words decade, century, and millennium etymologically refer to
tens, hundreds, and thousands of years, but 9 millennia, 4 centuries, 3 decades,
6 years is read and understood differently from 9436 years even though both
phrases refer to the same time value. I agree fully with Closs that the kin, uinal,
and tun counts were read separately, but believe that he has not gone far enough,
and regard the Maya Long Count dates as five separate nonpositional counts of
five different time periods.
The assumption that the numerals in the Dresden Codex must have been positional is linked with the belief that positional notation was highly useful for doing
calendrical calculations. Since the Maya did do these calculations, and since these
numbers look like positional notation, it is natural to infer that they were, even
though Classic Maya dates were normally written as five different periods rather
than as a single sum of days. When Mayanists interpret Mayan chronology, they
must translate Maya dates into a single number of days in order to the correlate
Maya and Western calendars (e.g., the GoodmanMartinezThompson correlation
establishes the beginning of the Maya calendar as Julian day number 584,283). Yet,
however the Maya may have read these columns of numbers, there is no evidence
that they ever calculated with them. The Dresden Codex is a repository of calendrical
data, including what appear to be multiplication tables, but there are no calculations
on paper. There is specific ethnohistorical evidence concerning Maya computation,
from Landas Relacin de las cosas de Yucatan, which suggests that the sixteenthcentury Maya did not calculate directly using baranddot numerals:
Their count is by fives up to twenty, and by twenties up to one hundred and by
hundreds up to four hundred, and by four hundreds up to eight thousand; and
they used this method of counting very often in the cacao trading. They have other
very long counts and they extend them in infinitum, counting the number 8000
twenty times, which makes 160,000; then again this 160,000 by twenty, and so on
multiplying by 20, until they reach a number which cannot be counted. They make
their counts on the ground or on something smooth. (Tozzer 1941: 98)
Computation was done on some sort of flat surface, suggesting that some
sort of physical counting board was employed. Some Mayanists have turned
their attention to what sort of physical counters the Maya might have used and
whether the bars and dots used as Maya numerals had physical correlates in rods
and beans, or some other such markers (Thompson 1941: 4243; Tozzer 1941: 98;
296
Numerical Notation
Satterthwaite 1947: 3031; Fulton 1979: 171). Sol Tax, working among the Maya of the
Guatemala highlands at Panajachel in the 1930s, found that they computed using
beans or stones in groups of five and twenty, supporting the idea that the ancient
Maya may have done similarly (Thompson 1941: 42). Counting boards are often
positional in structure, and some use special counters or markers for empty
positions signs that resemble a zero. On this basis, some suggest that numerals
were written positionally in a purely vigesimal fashion for noncalendrical purposes
that is, with the third and fourth positions having the values of 400 and 8000
in emulation of the mode of computation (Marcus 1976: 39; Lounsbury 1978:
764). Yet the host of speculations on the use of baranddot numerals directly in
calculation, without an intermediary computational device, is useless (Sanchez
1961, Bidwell 1967, Anderson 1971, Lambert et al. 1980, Mhlisch 1985). While,
as Anderson (1971: 63) states, it is not unreasonable to suggest that some attempt
to use the numerals directly in computations might have occurred, this pastime
tells us much more about the ingenuity of modern scholars than it does about
the actual practices of Maya mathematics. Just as the Romans and Greeks had a
placevalue abacus but no positional numerical notation system, the presence of a
Maya abacuslike device does not presuppose that they had positional numerals.
The columns of an abacus work just as well if they indicate distinct units of baktuns, katuns, tuns, uinals, and kins as they do if they represent the powervalues
144,000, 7200, 360, 20, and 1. The manipulation of counters is identical, but the
reading of the results is very different.
Unfortunately, the great bulk of Maya codices is now lost to us forever due
to the tragic destruction of manuscripts on Spanish orders in the early colonial
period. It is far too easy to create hypotheses concerning lost positional inscriptions when huge quantities of evidence have literally gone up in smoke. Yet the
surviving evidence does not support the hypothesis that the number columns in the
Dresden Codex should be interpreted as sums of days, and thus as a cumulativepositional numerical notation system. The most parsimonious explanation is that
the omission of periodglyphs was abbreviatory but did not entail a radical rereading of the numerical coefficients.
In his analysis of Maya arithmetic, Fulton noted that it is possible to have a
strictly positional notation, not altogether different from our present one, without any zero whatsoever (1979: 171). Positionality requires some way of avoiding
ambiguity between 749 and 7049, but this may be simply an empty space. Inverting this insight, I believe that the Maya baranddot system had a zero, but did
not use the positional principle. This is not to say that the Maya zero or completionsign was nonfunctional. While it was retained for aesthetic purposes in places
where it was not strictly needed (when periodglyphs were present), the zero
was needed whenever the periodglyphs were omitted and there was an empty
Mesoamerican Systems
297
period. But the purpose of a Maya zero in a number such as 1.0.4 does not appear
to be to indicate that the first number should be multiplied by 360, but rather
simply to indicate that the middle position is empty, and thus the 1 should be read
as 1 tun rather than 1 uinal. While something like positionality is used to distinguish
different units of time, there was no Maya positional numerical notation system.
Once we abandon the notion that the presence of placevalue is an eternal standard
of utility in numeration, we can see that the baranddot system was highly useful for recording dates even though it was not, strictly speaking, positional. The
main baranddot system is cumulativeadditive, and when cumulativeadditive
numeralphrases were combined to express large time periods, what results is a quasipositional calendrical notation, but not a true positional numerical notation.
For an analysis of the specific deities and other symbolism associated with each glyph,
see Thompson (1971): 131137; Macri (1985); Stross (1985).
298
Numerical Notation
sometimes also 13) are expressed additively using 10 (Macri 1985: 75). No other
Mesoamerican numerical notation system uses a decimal subbase. The origin of
this feature probably lies with the lexical numerals of the Mayan language family, which uses decimal structuring to express the numerals from 13 to 19, but has
rather opaque formations for 11 and 12 (in Classic Maya, buluc and lahca), just as
the English eleven and twelve do not show any clear relation to ten (Lounsbury
1978: 762). Additionally, Macri (1985: 48) suggests that it may have been important to have thirteen simple signs to correspond to the thirteen deities used to
name days in the Maya sacred calendar.
The headvariant numerals are relatively common on Maya inscriptions, though
less common than the baranddot numerals. They also appear occasionally in the
Dresden Codex, though not in the other Postclassic codices (Thompson 1971:
131). Macri (1985: 55) hypothesizes that they may have had a Preclassic origin,
but no preMaya inscription uses them. Macri (1985: 48), pointing to phonetic
correspondences between the headvariant signs and the Eastern Maya lexical
Mesoamerican Systems
299
numerals, suggests an Eastern Maya origin for the system, but Stross (1985) points
out that many of the same correspondences exist in the MixeZoquean language
family, to which the Olmec language may have belonged. Yet none of the Isthmian inscriptions contains headvariant numerals, and many centuries lie between
the decline of the Olmec civilization and the appearance of headvariant glyphs.
The head variants are extremely different graphically and structurally from the
baranddot numerals, and cannot have emerged directly from the baranddot
tradition. We had best think of them as a complex set of metaphors by which the
numerical symbolism of deities was used as a code for numerical information, not
as a numerical notation system in their own right.
Given the destruction of so many Maya codices, as well as the imperfect state
of Maya archaeology and hieroglyphic decipherment, it is difficult to say when
the headvariant numerals ceased to be used. Since the Dresden Codex is the only
surviving Postclassic codex to use them, and then only occasionally, it is possible
that they declined in use during the Postclassic period.
Mexican DotNumerals
During the Postclassic period (tenth to sixteenth centuries), many of the peoples
of central Mexico began using a system of dots to represent small integers in their
pictographic manuscript tradition. Since this means of representation lacks a base
and relies only on onetoone correspondence, strictly speaking it does not constitute a numerical notation system, but it deserves some mention here. In their
early history, the Mixtec and Teotihuacni used baranddot numerals, borrowed
from the Maya or the Zapotecs, but after the tenth century ad, the system fell
into disuse (Caso 1965: 955; Langley 1986: 143). While baranddot numerals were
occasionally used in a few later Mixtec codices, apparently for archaic or sacred
reasons, they were largely replaced by a system whereby dots alone were used for
the numbers 1 through 19, representing daynumbers and other objects (Colville
1985: 839841). The peoples of Oaxaca, the Valley of Mexico, and the Gulf Coast
used this system until the time of the Spanish conquest. A numeralphrase was
composed of a series of dots in a single row. To facilitate reading and to save space,
larger numbers were often grouped in segments of three to five units, sometimes
connected by lines, and sometimes changing direction (e.g., horizontal to vertical) in the middle of a numeralphrase. Numbers above 20 were never expressed
in this system.
Given that the central Mexican calendar is part of a Mesoamerican calendrical
tradition, and given the common use of dots for units in both the Maya and dotonly systems, I think it is plausible that between the tenth and twelfth centuries
ad, the use of bars for 5 was gradually abandoned, although the reason behind
300
Numerical Notation
this change is not clear. The influence of Toltec culture, which was becoming
predominant in Mesoamerica at this time, has been cited as the cause of this shift
(Caso 1965: 955). Yet this argument begs the question of why the Toltecs did not
adopt baranddot numerals. Dotonly numerals are not known from anywhere
in Mesoamerica prior to the tenth century ad, so it is unlikely that there was such
a tradition prior to that point. Thus, unless the use of dots for units developed
independently in the two different parts of Mesoamerica, the dotnumerals must
be descended from the baranddot system.
The dotnumerals were ancestral to the later Aztec numerals, a base20 cumulativeadditive system. Because the Aztecs, like the Maya and Mixtecs, used dots for
units, but because, unlike the baranddot numerals, the Aztec system has no quinary component, the dotnumerals are a likely intermediary between the lowland
and highland Mesoamerican systems. Both the dotnumerals and the Aztec numerals use up to nineteen dots for units, the difference being that with the Aztec
numerals, the dots were more regularly grouped in fives, and higher numbers were
written using different signs for the powers of 20. It is generally believed that the
Aztecs inherited their tradition of manuscript writing from the Mixtecs (Colville
1985: 839). Dotnumerals continued to be used in Aztec manuscripts even after
the development of the cumulativeadditive numerals in the fourteenth century.
By the time of the Spanish conquest, the Aztec numerals had supplanted the dotnumerals in some areas outside their tributary area, and were used in many of the
postConquest Mixtec codices (Terraciano 2001).
Aztec
The name Aztec applies most precisely to the Nahuatlspeaking inhabitants of
the region immediately surrounding the ancient city of Tenochtitlan (modern
Mexico City), who controlled a substantial tributary system in central Mexico between the fourteenth and sixteenth centuries. More generally, the term often refers
to the various UtoAztecanspeaking peoples of central Mexico who were under
Nahua rule during this period. The Aztec tributary network, which embraced
numerous small states, produced a large number of manuscripts, using a combination of ideographic and phonographic signs. The considerable debate concerning
whether this Aztec manuscript tradition constituted true writing or simply served
as a mnemonic aid is irrelevant to the study of Aztec numeration. The Aztecs most
definitely possessed a vigesimal numerical notation system, whose signs are shown
in Table 9.8.
The sign for 1 is the dot that was commonly used for units throughout Mesoamerica. The signs for the vigesimal powers are depictions of objects: for 20, a flag
(pantli); for 400, a feather (tzontli, literally hairs); and for 8000, a bag used to
Mesoamerican Systems
301
20
400
8000
X v
Y x :
yyy
xxxxx vvvvv
xxx
vvvvv
vvv
ttttt
tttt
(3 8000) +
(8 400) +
(9 1)
27,469 =
(13 20) +
hold copal incense (xiquipilli) (Harvey 1982: 190). These signs were combined in
a cumulativeadditive fashion, written in horizontal rows with the highest powers
on the left. Although the Aztec numerals, unlike the Maya baranddot numerals,
did not use a sign for 5, groups of more than five identical signs were arranged in
sets of five for easier reading. Groups of five signs were sometimes joined to one
another with a horizontal line underneath the set.
The purely vigesimal structure of the Aztec numerical notation system and the
shapes of its numeralsigns are quite different from those of the lowland Mesoamerican baranddot system. Instead, the Mexican dot numerals are the most
likely ancestor of the Aztec system. It is plausible that the Aztecs originally used
dots alone, but then, as the administrative needs of their tributary system grew,
invented new numeralsigns for 20 and its powers. As far as can be discerned, the
inventors and early users of the Aztec system were not influenced directly by the
lowland Mesoamerican systems of the Maya. The Mexican dot numerals do not
constitute a numerical notation system according to my definition, because they
lack a base, meaning that the Aztec system was invented relatively independently.
The most important function of the Aztec numerals was to record the results of
economic transactions, such as amounts of cacao beans, grain, clothing, and other
goods received from different regions of their tributary system (Payne and Closs
1986: 226230). Numerals were also used in Aztec annals and historical documents, such as the record of the massacre of 20,000 prisoners in the Codex TellerianoRemensis (Boone 2000: 43). Sometimes, when recording amounts of goods,
individual numeralsigns were attached to an equal number of pictographic signs
for goods. Accordingly, one might record 1200 balls of incense not as the numeral
1200 followed by a picture of an incense ball, but rather using three balls of incense, each of which would be placed immediately underneath a sign for 400.
The use of Aztec numerals to record large quantities of tribute and individuals
stands in sharp contrast to the Maya baranddot numerals, which were almost
302
Numerical Notation
wholly calendrical in function. The Aztecs denoted their thirteen months using
series of dots in rows, just as the Mixtecs did, but when they did so, they did not
group dots regularly in groups of five, and thus this represents a continuation of the
dotnumerals in Aztec manuscripts (Boone 2000: 4344). Normally, the Aztecs did
not record dates or other calendrical information using the larger numeralsigns. In
a single text, the Vatican Codex, large periods of time seem to have been expressed
using cumulativeadditive combinations of different signs, the largest of which represents 5206 years with thirteen signs that probably represent 400 (the third sign in
Table 9.8), above which six dots were written (Payne and Closs 1986: 234235).
After the Spanish conquest, the Aztec numerical notation system continued
to be used in various colonial documents. In fact, its use spread well beyond the
Valley of Mexico, as Nahuatl increasingly became a lingua franca used by indigenous highland Mesoamericans. For instance, Aztec numerals are common in the
Mixtec Codex Sierra, a midsixteenthcentury account book that uses Western,
Roman, and Aztec numerals side by side (Terraciano 2001: 4045).10 In a few postconquest manuscripts, fractions could be depicted by segments of 1/4, 1/2, and 3/4
of a dot, low multiples of five by filling in quarters of the pantli flag sign, and 100,
200, and 300 using segments of the tzontli sign for 400 (Vaillant 1950: 202).
A few postconquest Aztec codices use multiplicative rather than strictly additive numerical notation. Guitel (1958; 1975: 177) was the first to point out that
one of the oftenreprinted examples of Aztec numbers depicts a basket of cacao
beans from which four signs for 400 emerge, above which a pantli or flag for 20 is
placed. This numeralphrase represents a total amount of 32,000 cacao beans multiplicatively, as 20 baskets of 1600 beans each, rather than additively, as 4 xiquipilli
of 8000. In a circumstance where cacao beans come in baskets of 1600 beans, however, it is important to denote that there are 20 baskets of 1600 each, not simply
32,000 beans. This does not certify that placing the numeralphrases for 20 and
1600 together means 32,000. However, Guitel was not aware of another text, a
Texcocan document now known as the Codex Kingsborough, where multiplicative notation was used extensively (Paso y Troncoso 1912, Harvey 1982). I will treat
this structurally distinct variant of the standard Aztec system later.
As disease, warfare, and acculturation diminished the strength of Aztec traditions, the old numerals ceased to be used. I do not know of any documents from
later than 1600 that use Aztec numerals. After this point, Roman and especially
Western numerals were employed throughout highland Mexico.
10
Boone (2000: 254) indicates that Oaxacan texts do not contain signs for 400 or 8000; at
least in the case of the tzontli sign for 400, she is incorrect, as this is found in the Codex
Sierra (cf. Terraciano 2001: Figure 2.16). Yet the yeardate, written as 1563 in Western
numerals, is not transliterated in Aztec numerals but rather in Mixtec lexical numerals.
Mesoamerican Systems
303
Texcocan LineandDot
The city of Texcoco in the province of Tepetlaoztoc was one of the most powerful
cities in the Valley of Mexico both before and after the Spanish conquest. While
many sixteenthcentury colonial documents continued to use the Aztec numerals
just described, a handful of Texcocan documents contain a quite different system,
which I will call the Texcocan lineanddot system. These documents have been
studied extensively by Herbert Harvey and Barbara Williams and are identified
collectively as the Tepetlaoztoc Group (Harvey and Williams 1980, 1981, 1986;
Harvey 1982; Williams and Harvey 1988, 1997). The numeralsigns of this system
are shown in Table 9.9 (Harvey and Williams 1980: 500).
This system is cumulativeadditive, with a base of 20 and a subbase of 5. The
sign for 5 consists of five unitstrokes joined together by a curved line, so it is
perhaps just a matter of personal preference whether we see it as a separate numeralsign. A similar technique was occasionally used to group sets of five dots
for 20 into a single unit of 100. Perhaps the most unusual feature of this system is
that, whereas other Mesoamerican numerical notation systems used a dot for the
units, here a vertical stroke denoted the units, while the dot took a value of 20.
Numeralphrases were written in a variety of directions, but were always arranged
in a single line from highest to lowest sign (Harvey 1982: 191).
This form of notation has been found in only three texts, all of which were written
in the vicinity of Texcoco in the 1540s. Two of these, the Cdice Vergara and the
Cdice de Santa Mara Asuncin, were cadastral records written around 1545 to
enumerate individuals and their land holdings. These two are in fact so similar
that they may have been parts of the same manuscript at one point, or at least were
drawn at the same time (Williams and Harvey 1997: 2). The third, the Oztoticpac
Lands Map, was written around 1540, and is also a record of lands, though as a
map rather than a census record (Cline 1966). The primary function of the numerals in all of these cases was to record land measures. Pictographic signs expressing
fractional linear units sometimes accompanied the numerals, but their meanings
are still unclear (Williams and Harvey 1997: 26). Several numeralphrases on the
Oztoticpac map were used to count sums of days, showing that this system was
not restricted to one domain.
The Cdice de Santa Mara Asuncin used a modified form of this system
to express numbers positionally rather than additively. In studying this text,
Harvey and Williams (1980) showed that lineanddot numerals occurred in two
different sections, but served very different functions. In one section, known
as milcocoli, lineanddot numerals were used in the regular manner, written
along the edges of maps of plots of land owned by different individuals to indicate their lengths. In another section, known as tlahuelmatli, the plots of land
Numerical Notation
304
20
100
from the milcocoli section were redrawn as rectangles (regardless of their original shape). This section also contained numerals indicating the areal measurement of each individuals land holdings. Comparing the milcocoli values, which
indicated the lengths of the sides of plots, and the tlahuelmatli values, which
recorded their total area, Harvey and Williams showed that a form of positional
notation was used to record land areas in the tlahuelmatli section using a set of
three distinct registers within a rectangular depiction of a plot of land (Harvey
and Williams 1986: 242). In the top right corner, dotandline numerals indicated values from 1 to 19 in a small protuberance. On the bottom line of the
rectangle, units and groups of five indicated multiples of 20 units. No dots were
ever used in either of these two registers. When dots were found, they occurred
with or without units in the center of the rectangle. Strangely, this third register
also counted multiples of 20 (i.e., lines equal 20 and dots equal 400). No plots
of land show values both on the bottom line and in the center. When the twenties register and the units register were added together, a total area value was
reached. Harvey and Williams found that in 71 percent of the land plots they
examined, the tlahuelmatli value was within 10 percent of the projected area for
that plot based on the milcocoli measures (1980: 501). While this may not seem
remarkably accurate, the plots were often very erratic in shape, so that calculating area was not simply a matter of multiplying length by width. Where there
is no value in the third (central) register, a corn glyph, or cintli, is drawn at the
top of the rectangle (Harvey and Williams 1980: 501). This sign may have been
to indicate that the third register is empty, and thus may have served one of the
functions of a zerosign.
These numbers can be read as a base20 cumulativepositional numerical notation system with a subbase of 5. Unlike Western numerals, in which the positions
are arranged in a straight horizontal line, the Texcocan system uses three registers,
the last two of which have an identical positional multiplier. However, the cintli
glyph is not used to indicate empty positions, but rather provides information as
to where to find the twenties power (on the bottom line, rather than in the center
of the rectangle), and thus is conceptually distinct from the Western zero (and,
indeed, from other zeroes such as the Babylonian zero). While I think that the
correlation established by Harvey and Williams demonstrates that the tlahuelmatli
value represents an area value, I am not fully convinced that it is meant to be read
as a single number; it may instead represent two values, one of which represents a
Mesoamerican Systems
305
Figure 9.1. Numerical phrase from the Codex Kingsborough enumerating the population of Tepetlaoztoc at 27,765 (3 8000 + 9 400 + 8 20 + 5). Source: Paso y Troncoso
1912: 218v.
larger area value that is twenty times another value. I do not know how this issue
could be resolved at present.
A unique Texcocan document from 1555, the Codex Kingsborough, also uses
something like the lineanddot numerals (Paso y Troncoso 1912). It was a record
prepared as part of a legal plea made to the Spanish encomendero of the region,
denoting the massive amount of tribute paid to Spanish officials by the inhabitants of the Tepetlaoztoc region in an effort to convince colonial officials that the
populace was overworked; extensive description in Spanish confirms the numerical values (Harvey 1982: 193). Curiously, this text combines Aztec numerals and
Texcocan lineanddot notation. Lines and chunked groups of five lines indicate
1 and 5, respectively. To write larger numbers, dots organized in lines of five were
placed beside the signs for 20, 400, and 8000. The dots were placed in a single row,
with the signs for 20 and 400 above them and the 8000 sign below them. Thus,
where the regular Aztec numerals use these three signs cumulatively, the Kingsborough numerals are written using just one of each sign, next to which units from 1
to 19 were expressed with dots. Figure 9.1 depicts the numeralphrase 27,765, indicating the population of the district at the time, but replacing the standard Aztec
sign for 8000 with a head above a sack (Paso y Troncoso 1912: 218v).
Whereas the basic lineanddot system is cumulativeadditive, and the tlahuelmatli system is cumulativepositional, this system is multiplicativeadditive.
While the dots look like the 20 dots of the lineanddot system, they each stand
for 1 in this system. The total value of the numeralphrase is taken by multiplying
the dots for units by the values of the power signs and taking the sum. To add
to the complexity of this situation, in some cases the flag glyph for 20 could be
omitted, retaining only the dots (Paso y Troncoso 1912: 261r, 238v, etc.). In these
cases, we have the elements of a cumulativepositional system, since the value of
the twenties power is determined by its position in the numeralphrase through
implied multiplication. Finally, in a couple of numeralphrases, lines are placed to
the left of dots, as where a number is written as II, which might be read from
306
Numerical Notation
right to left as 42 (Paso y Troncoso 1912: 274v). The erratic nature of the system
suggests that whoever wrote it was extremely inventive and was in the process of
experimenting with different means of representation.
The most important question regarding the lineanddot numerals, their positional variant in the tlahuelmatli records, and their multiplicative variant in the
Codex Kingsborough, is whether they existed before the Conquest, or if their
development was stimulated by contact with the Spanish. Neither the Western or
Roman numerals are cumulativepositional or multiplicativeadditive, and neither
uses a base of 20, so the Texcocan systems are structurally distinct from those of
the Europeans. Thus, it would be premature to conclude that Spanish contact
brought about the development of these systems. It would be a mistake to attach
much importance to the use of a vertical stroke for 1 (parallel with both Western
and Roman numerals), given the ubiquity of this notation worldwide. Harvey
and Williams (1980: 503) argue that, while the tlahuelmatli numerals are positional
and have something like a zero, the use of different registers around a rectangle is
quite different from Western positionality, and the zero does not serve the same
functions as the Western zero. On this basis, they regard these systems as a native
invention.
I agree that the Texcocan numerical notation systems are so different from
Western and Roman numerals that the Spanish could not have introduced
them. Nevertheless, these may be instances of stimulus diffusion, which the
Texcocan scribes developed with an awareness of Western and/or Roman numerals but without adopting the form and structure of those systems. That the
Texcocan systems occur in only a handful of documents in a single region in
the generation immediately after the Conquest and cease to be used after only
two decades suggests that this was not a system of great antiquity. I believe that
the multiplicative (Kingsborough) and positional (tlahuelmatli) variants may
well have been stimulated within the rapidly changing social and intellectual
environment of the early colonial period, while the cumulativeadditive lineanddot numerals probably existed in the preConquest period. After 1545, epidemic disease greatly diminished the indigenous population of the region, and
it appears that the Texcocan numerals ceased to be used after the middle of the
sixteenth century.
Other Systems
Because our understanding of Mesoamerican numerals is imperfect, a number
of Mesoamericanists have developed theories regarding other forms of written
numeration. I think it quite likely that more numerical information has been
recorded than we are currently able to read in the Maya, Zapotec, Teotihuacni,
Mesoamerican Systems
307
and Aztec texts. Even if these hypotheses turn out to be incorrect, some elements
of them may be salvaged in the reconstruction of asyet unknown numerical notation systems.
In the 1950s, Howard Leigh postulated that in addition to baranddot numerals, some Zapotec inscriptions contained encoded astronomical data using a different set of glyphs (Urcid Serrano 2001: 4950, 54). In addition to bars and dots,
this supposed system had over twenty unique signs, including elements of base10,
base13, and base20 notation, culminating in a special sign for 1,186,380 (3 3
3 13 13 13 20). I am unconvinced that such a system actually existed in
the form asserted, but the Zapotecs may have encoded numerical information in
some of these glyphs, though not in the way Leigh imagined.
An unusual cumulativeadditive baranddot numerical notation system may
have existed at Teotihuacn, a system in which the bars did not have a fixed value
but could mean 5, 10, or 30, depending on their configuration (Langley 1986: 141).
The nature of the script of Teotihuacn is still controversial, though it is increasingly thought that there was a complex pictographic script of the type used later in
highland Mexico (Taube 2000). However, because Teotihuacn never used phonetic writing, and because, unlike the Aztec situation, there is no body of colonial
documents to explain the numerals, there is no way to confirm the values of any
potential numeralsigns.
Penrose (1984) asserts that in the almanac portions of the Dresden, Madrid, and
Paris codices, the Maya used cryptoquantum numerations to represent an encoded quantity of days separately from the baranddot or headvariant numerals.
He argues that the Maya represented hidden counts of large numbers by assigning
numerical values to special signs indicating the days of the Sacred Round 260day calendar, and then by manipulating them through multiplication. Mayanists
do not appear to be aware of Penroses research, and his conclusions must be
viewed as highly speculative and even pseudoarchaeological. The manipulations
necessary to extract meaningful numerical information from these signs are probably no more than numerological play.
An unusual form of numerical notation is employed on the Codex Mariano
Jimenez, a sixteenthcentury postConquest manuscript from Otlazpan (in the
province of Atotonilco). It is cumulativeadditive and uses dots for units, horizontal lines for twenties, and horizontally oriented tzontli (feather) glyphs for
400, with fractions of 400 depicted by showing partially denuded tzontli signs.
Although treated by Harvey and Williams (1986: 251253) as simply a variation
on the Texcocan system described earlier, the differences between the two systems
suggest that they are quite distinct. If more documents using this sort of notation
are found, we would have yet another postConquest regional variant of the Aztec
numerals.
308
Numerical Notation
Summary
The two features common to all the Mesoamerican numerical notation systems
is that they have a vigesimal base and that they are all cumulative rather than
ciphered. The Maya headvariant glyphs, a sort of ciphered symbolic code that
expresses only units up to 19, constitute a partial exception to this rule. The
presence of a quinary element is quite common, as is the use of dots for units,
but neither of these features is found in all the systems. Like the East Asian
phylogeny (Chapter 8), Mesoamerican numerical notation systems use a variety
of basic principles, and our primary evidence for their commonality is historical
rather than structural.
When the baranddot numerals were the only part of the Maya script to be
deciphered, it must have seemed remarkable to be able to extract calendrical information from such otherwise inscrutable documents. Yet we have a less complete
understanding of the cultural history of the Mesoamerican numerical notation
systems than we do of most Old World families. As our reading of Maya and
Aztec writings becomes more sophisticated, it is to be hoped that we will come to
a clearer understanding of their numerical notation.
chapter 10
Miscellaneous Systems
Around twenty systems do not fit neatly into the phylogenetic classification presented in Chapters 2 through 9. A few, such as the Inka khipu numerals, the Indus
(Harappan) numerals, and the enigmatic Bambara and Naxi numerals, apparently
arose independently of any other system, but gave rise to no descendant systems.
Others are cryptographic or limitedpurpose systems used in the medieval or early
modern manuscript traditions of Europe and the Middle East. The majority of this
chapter, however, deals with systems that emerged in colonial settings under the
influence of the Western or Arabic cipheredpositional numerals, in conjunction
with the development of indigenous scripts. Most of these systems were developed
in subSaharan Africa, but Asian (Pahawh Hmong, Varang Kshiti) and North
American (Cherokee, Iupiaq) indigenous groups have also developed their own
numerical notation systems. Finally, a few systems are probably members of other
phylogenies, but their exact affiliations remain inscrutable enough that no definite
conclusions can be reached.
Inka
The Inka civilization was an enormous state on the Pacific coast of South America that reached its pinnacle between 1438 and 1532. While writing is often (and
mistakenly) seen as a sign of civilization, or at least as a necessity for largescale
bureaucracy, the precolonial Inka state operated in the apparent absence of any
309
310
Numerical Notation
writing system capable of expressing phonetic values. Instead, the primary means
of encoding information was a system of knotted cords of different colors, known
as khipus,1 whose main purpose was to record numerical information to aid in the
administration of the Inka state. About 500 to 600 Inka khipus survive, although
accurate provenances cannot be established for most of them (Urton 1998: 410).
As first established by Locke (1912), khipus encode information using a decimal
positional numerical system of knots, and around twothirds of all attested khipus
encode information in this readily understood fashion. Around onethird, however, do not follow this structure; they remain completely undeciphered, and may
well have encoded nonnumerical information (Urton 1997, Quilter and Urton
2002).
A khipu is a set of colored cotton or wool cords consisting of a main cord (ranging from 10 to 20 cm up to several meters in length) from which multiple cords
are suspended. These numeralbearing cords are subdivided into pendant cords,
which hang directly down from the main cord when it is held horizontally and
stretched taut; top cords, which hang from the main cord but are tied so as to lay
on the opposite side of the pendant cords; and subsidiary cords, which hang from
a pendant cord, top cord, or another subsidiary cord rather than the main cord
(Ascher and Ascher 1980: 1517). The designation that pendant cords hang down
and top cords hang up is an artifice; while they naturally hang on opposite sides
of the main cord, we do not know how they would have faced. In numerical khipus, pendant, top, and subsidiary cords may contain a numeralphrase or, more
rarely, two. The system used to encode information is cumulativepositional with
a base of 10. In each position, the value of that power of 10 is encoded using one to
nine knots or loops. There is no sign for zero; instead, a space was left on the cord
in an empty position. The units position is the one farthest from the main cord
(its loose end), while the highest power is found closest to the main cord. While a
khipu theoretically could express any number (because the system is positional),
in practice, fivedigit numbers are the largest recorded, and these are quite rare
(Ascher and Ascher 1972: 291). Despite this obvious numerical structure, khipus
are often erroneously conflated with unstructured systems that use one knot for
one object (cf. Ifrah 1998: 70). Khipus contain a numerical notation system (a
positional one, in fact) and thus must be compared to written numerals rather
than to simple tallies.
Three different sorts of knots encoded numeralphrases, as seen in Figure 10.1.
To encode a value in the tens, hundreds, or higher powers, the khipu maker would
tie an appropriate number of single knots in a line. For the ones power, however, two
1
The spellings Inka and khipu currently enjoy favor with Andeanists over the older
Inca and quipu.
Miscellaneous Systems
Single knot
10s, 100s, etc.
Long knot
Units: 29
311
Figure8 knot
Units: 1
different types of knot were used. For all the units except 1, the cord was looped
around itself an appropriate number of times for the number being expressed;
the long knot shown in Figure 10.1 represents 4. Because a long knot cannot
be made with fewer than two loops, a value of one in the units position required
the use of a different knot, a figure8. The use of different knots might appear to
take away from the purely positional nature of the system. Yet, because there is
no zerosign, this technique greatly reduced the chance of misreading a cord. If a
cord contained six single knots followed by two single knots, it could not be read
as 62 but only as 620 (or possibly 6200). The use of long or figure8 knots in the
units position makes it much easier to tell which is the units position, and thus to
identify the subsequent positions.
Figure 10.2 depicts an unattested but plausible khipu. The main cord lies horizontally, with the pendant cords (P1 through P4) hanging down and the top cord
(T1) facing up, and with subsidiary cords (S1 through S3) hanging off both pendant and top cords. On this cord, only a single value would have a figure8 knot
(the 1 in the units position on P4); the other units values (3 on P2, 6 on S1, 2 on
P3, 6 on T1, and 6 on S3) would be made with long knots, and all the tens and
hundreds figures with single knots. As is sometimes the case in attested khipus,
the top cord value (776) is equal to the sum of the pendant cords (360 + 23 + 102 +
291), while the value on the top cords subsidiary (S3 = 26) is the sum of the subsidiaries of the pendant cords (20 + 6).
Although we can read the numerical values on khipus, their origin and early
history remain unclear. A set of twelve cotton strings twisted around sticks excavated at the late preceramic pyramid complex of Caral, Peru (c. 26002000 bce)
has been claimed by its excavator to be an early form of khipu; however, this
claim is unsubstantiated, and full data on the artifact remain unpublished (Mann
2005). Bennett (1963: 616) notes that some Mochica vessels (Early Intermediate
period, c. 200600 ad) bear markings that are suggestive of khipus. The first wellsubstantiated evidence for khipu use comes from Middle Horizon sites (c. 600
1000 ad) associated with the Wari civilization in coastal Peru (Conklin 1982).
Numerical Notation
312
T1
776
P1
360
P2
23
S3
26
S1
6
P3
102
P4
291
S2
20
Miscellaneous Systems
313
These khipus cannot be deciphered numerically because of their deteriorated condition (although they may have used nondecimal bases), and they use color in very
different ways than the Inka khipus, but are nonetheless clearly of the same basic
type. Most surviving khipus were collected haphazardly; prior to 2001, only two
archaeological discoveries of khipus had adequate proveniences (Urton 2001: 131).
The khipu system may have developed out of an earlier knotbased system using
simple onetoone correspondence, because knot tallies of this sort are widely distributed in the CircumPacific region (BirketSmith 1966). There is no evidence
of any connection between the khipu notation and any other numerical notation
system, and thus it is definitive that the Andes was home to an independent development of the placevalue principle.
Khipus were a vital part of the Inka recordkeeping system; they were employed
in this capacity for censuses, tributary records, and similar administrative functions. Jacobsen (1964), noting the frequency with which the top cord equals the
sum of the pendant cords, suggests reasonably (but unconfirmably) that such khipus may have been part of a doubleentry accounting system. The decimal base
of the khipu notation system corresponds to the decimal divisions of society by
which the state was administered. Some khipus contained calendrical rather than
administrative information (Ascher and Ascher 1989, Urton 2001). For instance,
khipu UR6 from Laguna de los Cndores contains a series of cords with values of
20 to 22 followed by cords with values of 8 or 9, and the sum total of these cords
is 730 (365 2), strongly suggesting that it may have been a biennial calendar
(Urton 2001: 138143). Most surviving khipus with good provenience have been
recovered from mortuary contexts. The Inka probably placed khipus in the graves
of khipukamayuqs (khipu makers and users). It is unclear whether this implies that
some of them should be read as tomb texts, because at present we are unable to
extract nonnumerical information from them (Urton 2001: 34).
Much ink has been spilled recently about whether khipus constituted something more than a numerical notation system, approximating the functions of a
writing system. Ethnohistorical data suggests that khipus recorded genealogical,
historical, and literary information, which raises the question of what code was
used to do so (Bennett 1963: 618). Gary Urton (1997, 1998, 2001) has argued forcefully that many khipus contain syntactic and semantic information far exceeding
their numerical functions. He contends that purely numerical readings that translate khipu texts as Western numerals inevitably mask, and eliminate from analysis, any values and meanings that may have been attached to these numbers by the
Quechuaspeaking bureaucrats of the Inka empire who recorded the information
(Urton 1997: 2). He argues against the idea that a khipu could have been interpreted
only by its maker or those trained in an idiosyncratic private code (Urton 1998: 412).
The khipus must have recorded some nonnumerical information; a list of pure
314
Numerical Notation
numbers is practically useless. In some way, at least the nature of what was being
counted must have been recorded somehow. The most likely possibility is that this
was done with color; the 1609 Comentarios of Garcilaso de la Vega (15391616)
inform us that colored cords were used to record different commodities (Bennett
1963: 617). Yet many khipus use multiple colors of cord, and there exists no reliable
means of reading the type of items counted.
Recent scholarship has established that at least some khipus encoded toponymic information, associating particular records with the places to which they
refer, which helps us to clarify how information was communicated within the
Inka administrative hierarchy (Urton 2005, Urton and Brezine 2005). A cache of
twentyone khipus excavated from a single urn in the palace of Puruchuco (northeast of Lima, Peru) revealed many whose introductory cords begin with arrangements of three figure8 knots (which normally represent the numerical value 1),
suggesting that this served as an identifier with which any reader could associate
the numerical data. Some of the khipus in this cache encoded identical or nearly
identical information, suggesting that copies might be kept at the site of a khipus
manufacture, with other copies distributed to the capital, Cuzco, or elsewhere.
Urton has also identified threeterm number sets that occur on some of the Puruchuco khipus that do not fit into the numerical structure of the remainder of the
record, and suggests that these are labels, perhaps an ayllu (kinship group) with
which the khipu was associated (Urton 2005: 162163).
The khipus encode at least as much information as the protocuneiform
accounting signs of Mesopotamia (Chapter 7), which identify only items being
counted and the quantity of each item, but which similarly served as a stateoriented bookkeeping system of credits and debits (Urton 2005: 164). Since the
protocuneiform system is regarded as protowriting, it is reasonable to attribute
the same status to the Inka recording system (Salomon 2004). It is possible that
the khipu system, over time, might have developed into a system for representing
speech (though doing so would be more difficult for a knotbased notation than
for a system based on inked or impressed signs). A single khipu cannot be at the
same time both a record of numbers and of things being enumerated and a fully
developed system for recording history and literature. Yet the roughly onethird
of khipus that do not follow an ordinary decimal and positional structure may
well have been nonnumerical. It is equally possible that some numerical khipus
recorded ideas or speech through some sort of code, but without a key, we cannot
definitively conclude that this was the case.
Urtons (1997: 179) speculation that there might have been two precolonial
khipu systems (one for recording quantity and another for recording narrative) is
useful but at present unconfirmed. While postConquest chroniclers state explicitly that the khipus carried only numerical meanings, Urton postulates that the
Miscellaneous Systems
315
316
Numerical Notation
After the Spanish conquest in 1532, khipus continued to be used for the same
administrative functions as they had been previously, and the Spanish, through
Andeans who could read their values, used the data recorded on them (Loza 1998,
Fossa 2000). Brokaw (2002) discusses the cognitive shift required of Quechua
speakers in the sixteenth century with the transition from khipu notation to
European literate conventions, as demonstrated through Guaman Pomas Nueva
cornica. This shift involved changes in how texts were organized, written, and
read, and also forced preexisting Quechua ideas about numbering and counting into conflict with Western textual conventions (e.g., regarding pagination).
Also in the sixteenth century, the mestizo chronicler Blas Valera (15451597), who
advocated for Quechua as a Christian liturgical language in addition to Latin,
developed a system of forty syllabic knots to be used on socalled royal khipus,
which reflected Valeras theories about Quechua as a worthy language and his
conviction that one could not rank societies based on the quality of their writing systems (Hyland 2003: 129135). While Valeras system has occasionally been
regarded as a precolonial invention for which he took credit, thus making the
khipus at least partly a phonetic writing system, it is substantially more likely that
the royal khipus were a colonial invention that applied the notion of phoneticism
to the existing precolonial system.
The widespread use of khipus was curtailed in the 1580s, when they were declared
to be idolatrous and the Spanish colonial administrators decreed that they should be
destroyed. Yet in that same decade Mercedarian friars began using khipus to encode
information about Christian life, using the principles outlined by Valera (Hyland 2003:
136137). While it was once thought that the use of khipus had essentially ceased by the
sixteenth century, it is now evident that their use for secular administration in the colonial period continued. Moreover, in local accounting contexts apart from state control,
khipus have continued to be used by animal herders in parts of Peru and Bolivia for
recording quantities of livestock up to the present day (Bennett 1963: 618619; Ifrah
1998: 6970; Urton 1998: 410; Salomon 2004). Nineteenthcentury khipus found by
the explorer Charles Wiener in Paramonga have systems of knots and bundles quite
different from the precolonial khipu, but confirm that the practice continued in varying forms well after the colonial period (Hyland 2003). These were not simply tallying
systems, however, but were cumulativepositional and decimal, and thus constitute a
survival of the Inka numerical notation system.
A significant part of khipu studies today, then, and of Inka ethnomathematics in
general, relies on ethnographic work with the descendants of the Inka, for example, modern Quechua and Aymara (Quilter and Urton 2002, Salomon 2004).
The modern episteme of numbers rests on different principles than Western
arithmetic, in particular placing great emphasis on even numbers as complete
and odd numbers as incomplete or even dangerous (Urton 1997; Brokaw 2002:
281287). However, between the sixteenth and twentyfirst centuries, significant
Miscellaneous Systems
317
changes almost certainly occurred in these ideas, much as the foundations of sixteenthcentury European mathematics bear only a passing resemblance to modern
practices. Further ethnographic, ethnohistorical, and archaeological data promises
to help resolve some of the remaining mysteries concerning the khipu records,
and the establishment of the Khipu Database Project will help facilitate computer analysis of khipus in museums and collections worldwide (Khipu Database
Project 2004). A complete decipherment of the khipus as they were used in premodern contexts, however, may well be impossible.
Bambara
One of the most peculiar African numerical notation systems was used by the
Bambara of Mali in religious and divinatory contexts (Ganay 1950). Although
details of the systems history are sketchy, we have a fair idea of the numeralsigns and the structure of the system. The Bambara numeralsigns are shown in
Table 10.1.
The Bambara system is structurally irregular; while it is additive, it alternates
between cumulative and ciphered notation, and while it is mainly decimal, it has
vigesimal components. For instance, 1 to 19 are written primarily with vertical
cumulative unitstrokes. The value of a set of vertical strokes is doubled if a horizontal line is crossed through it (effectively dividing the number into two registers,
one above and one below the line). For odd numbers, an additional halfstroke
can be placed at either end of the phrase, sometimes vertically and other times
at an angle. Each of the tens from 20 to 170 has its own sign, which makes the
system ciphered at this point. The signs for 180 and 190 are additive combinations
of 100 + 80 and 100 + 90, respectively. To add a number of units from 1 to 9 to
one of these ciphered signs, an appropriate number of strokes are attached to the
sign for the multiple of 10 (or dots, when adding units to 60, 160, or 170). This
means of representation is decimal each decade has its own sign to which up
to nine unitsigns were attached. Yet, because there are signs for 110, 120, and so
on, it is unlike the Greek cipheredadditive alphabetic numerals (in which 100 is
followed by 200, 300, and so on). Moreover, some of the decadesigns are similar
enough to the ones preceding them (40 vs. 50, 100 vs. 110, 140 vs. 150, 160 vs. 170)
to suggest an additional trace of a vigesimal base. For numbers higher than 200,
the cumulative principle is again employed by repeating the sign for 100 (another
decimal component) as many times as required in a vertical column, with any
needed additional signs placed at the top of the column. Figure 10.3 shows some
higher numeralphrases (as reproduced from Ganay 1950: 300).2
2
Large numeralphrases for 1935 and 4000 are also listed, but are highly irregular, and I
cannot determine what principle has been used to determine their value.
Numerical Notation
318
aa
aaa
aaaa
aaaaa
10
aaaaaa
aaaaaaa
bbbb
bbbbc
bbbbb
11
12
13
14
15
bbbbbc
bbbbbb
bbbbbbc
bbbbbbb
ybbbbbbb
16
17
18
19
20
bbbbbbbb
bbbbbbbbz
bbbbbbbbb
ybbbbbbbb
30
40
50
60
70
80
90
100
110
120
130
140
150
160
170
180
190
The Bambara numerical notation system was used primarily in ritual contexts,
especially those pertaining to divination using numbers (Ganay 1950: 298). Little
is known of its origin, period of use, or decline. It shows no resemblance to any
of the systems that would have been known by Bambara, who had considerable
contact with the Muslim world. While the cipheredadditive Arabic abjad numerals commonly used for divination in the Maghreb are the most likely ancestor,
220
489
230
240
Miscellaneous Systems
319
the Bambara system is quite different in most respects its frequent use of the
cumulative principle, the presence of a vigesimal component, and its numeralsigns. I have no idea whether this system continues to be used, though I suspect
that it does not.
Berber
The Berbers, or Imazighen, live in North Africa and speak a set of closely related
AfroAsiatic languages. For most of their history, the Berbers have been a marginal
people living on the periphery of larger polities (Carthage, Rome, and various
Muslim states), but they have nonetheless retained considerable cultural independence. The Berbers developed a consonantal script on the model of that used
in Punic Carthage possibly as early as the sixth century bc, which was in continuous use until at least the third century ad; the Tifinigh script (still used by the
modern Tuareg for love letters, domestic ornamentation, and games) is descended
from it (OConnor 1996). There is no numerical notation system associated with
either the classical Berber script or its modern descendant. Nonetheless, a distinct
numerical notation system was used by traders in the Berber city of Ghadames (on
the border of Algeria and Libya) in the nineteenth century, and appears to be in use
still (Rohlfs 1872, Vycichl 1952, AghaliZakara 1993). Vycichl (1952: 8182) presents
the system as described by two separate authors, Hanoteau and Si Mohammed
Serif, while Rohlfs presents a third system; I reproduce all three in Table 10.2.
The system is cumulativeadditive and written from right to left, with the decimal powers repeated up to four times and the halved powers only once in any
numeralphrase. Sometimes, groups of signs could be placed in two rows to save
space (Rohlfs 1872). In addition to these signs, Hanoteau claims that a horizontal
line stood for the fraction 1/4, and that this sign could be grouped vertically to
indicate 1/2 and 3/4 (Vycichl 1952: 81). The two sets of numeralsigns are identical,
except for the signs for 500 and 1000. It is possible that both of these systems were
actually used, either in different contexts or at different times. However, it is more
likely that an error of interpretation created the discrepancy, because Hanoteaus
1000sign is essentially identical to Serif s 500sign.
The question of the Berber systems ancestor (if any) is still open. It is possible
that it was an entirely independent development. The similarities between certain
numerical signs and letters of the Berber consonantary (r with 10, f with 500,
and s with 1000) are interesting, but they do not correspond to the Berber lexical
numerals in any obvious way. The Phoenician/Punic numerical notation system is
quite different in its structure, lacking a sign for 5, and employing a special sign for
20 and a hybrid multiplicativeadditive structure above 100. The use of  for 1 and >
for 5 is superficially similar to the Roman system; Ghadames was an important
Numerical Notation
320
Table 10.2. Berber numerals
1
1
1
Rohlfs
1
1/4: E 1/2: J 3/4: O
Hanoteau
Si Mohammed Serif
10
50
100
500
1000
44 = 88
488 = 111
trading post (Cydamus) under imperial Roman control, and there are Roman
numerals on some of the Latin inscriptions found there. However, the systems
are written in different directions and have different signs for the higher values.
Vycichl (1952: 83) suggests that the system derives from the South Arabian numerals. The Berber script may be somehow indebted to the South Arabian (OConnor
1996: 112). If Hanoteaus list of signs is correct, the Berber system, like the South
Arabian, lacks a sign for 500; furthermore, both systems use O for 10. However,
the South Arabian system ceased to be used in the first century bc and was never
used in Africa, so to accept this theory requires that we believe in a twothousandyear unattested history for this system. The system having the most promise as an
ancestor is the ArabicoHispanic variant Roman numerals (Chapter 4) used in a
Spanish Inquisition document of 1576 (Labarta and Barcel 1988: 34). This system
employed , V, and O for 1, 5, and 10, was written from right to left, and was used
in the same general region as the Berber system. Though three centuries is still a
chronological gap that needs to be resolved, it is not nearly so great as the enormous leaps that need to be inferred to hypothesize alternate paths of diffusion.
Ultimately, more data are needed for this system to be assigned unambiguously
to any phylogeny.
The Berber system was used in the nineteenth century for indicating the prices
of trade goods. Rohlfs (1872) learned about this system as a traveler in the Ghadames region, but only ascertained the meanings of the signs through great effort
and negotiation. He thus believed that the system was semicryptographic, restricting the flow of information concerning prices to a limited group of Berber traders
in order to give them an advantage over Arab traders. The system is not especially
difficult to decipher, however, and so I am unconvinced that this purpose was very
important. AghaliZakara (1993: 151153) reports that several numerical notation
systems are still used in the region of Ghadames; one of these is the system just
described; another simply repeats the sign for 10; and a third, inexplicably and
surely incorrectly, is seen as having no signs for the powers of ten but only for the
Miscellaneous Systems
321
10
11
12
13
14
15
16
17
18
19
@ # $ % ^
1938 = 4^* (4 400 + 16 20 + 18)
&
subbases, 5, 50, 500, and 5000. None of these systems is widely used, but they do
appear to be in present use among at least some Tuareg.
Oberi Okaime
In the late 1920s, a syncretic indigenousChristian religious movement known as
Oberi Okaime (or Obri Vkaim) arose among the IbibioEfik, speakers of a set
of related dialects of the NigerCongo language family in southeastern Nigeria.
By 1931, the divinely inspired leaders of this movement had developed an alphabet
(written from left to right) and a set of numeral symbols (Adams 1947; Hau 1961).
The script represented an arcane revealed liturgical language of the sect, but was
not used to write Ibibio. The Oberi Okaime numeralsigns are shown in Table 10.3
(Hau 1961: 295).
The system is cipheredpositional and vigesimal; it is the only known cipheredpositional base20 system with no subbase, with the partial exception of the Maya
headvariant glyphs. The vigesimal structure of the system is based on the similarly
vigesimal Ibibio lexical numerals (Abasiattai 1989: 505506). Numeralphrases are
written from left to right with the highest powers on the left. The inventors of
the Oberi Okaime numerals were educated in Christian missionary schools in the
1920s, where they became literate in English and learned Western numerals. While
none of the numeralsigns resemble the corresponding Western numerals except
for 0, the script and its numerals were strongly influenced by Western traditions
of writing (Dalby 1968: 160161). Haus (1967) highly dubious suggestion that the
Oberi Okaime script derives directly from Minoan Linear A, used thousands of
kilometers away and over three millennia previously, cannot possibly apply to the
numerals. The numerals were used in a relatively small number of liturgical texts
and personal letters among the members of the Oberi Okaime sect. The system
was still used by some individuals when Kathleen Hau corresponded with its leaders in 1961. In 1986, Sunday school classes were begun in order to revive the Oberi
Okaime liturgical language along with the numerals and script, but this appears
Numerical Notation
322
10
100
1000
10000
76 =
or
to have been unsuccessful (Abasiattai 1989: 506). Western and sometimes Arabic
positional numerals are used in the region today.
Bamum
The Bamum live in part of southwestern Cameroon near the border with
Nigeria. In the late nineteenth or early twentieth century,3 Sultan Ibrahim Njoya
(ca. 18751933), a Bamum ruler, took it upon himself to develop a script for his
people. Njoya, aided heavily by an assistant, Nji Mama, developed the script
through several stages, starting with a large logosyllabary and gradually reducing
the number of signs into an eventual syllabary of eighty characters (Tuchscherer
2005: 479). From its inception, Bamum writing made use of numerical notation.
The earliest Bamum numerals are shown in Table 10.4 (Dugast and Jeffreys 1950: 6).
This system is decimal and multiplicativeadditive, with numeralphrases written from left to right. Curiously, the powersign for the units could either precede
or follow the unitsign (Dugast and Jeffreys 1950: 30). The unitsigns for 7, 8, 9,
and 10 were not at this stage fully ideographic, but instead were constructed of
two graphic parts, each of which represented a syllable in the twosyllable Bamum
words corresponding to those numbers (Dugast and Jeffreys 1950: 98). At this
point in the systems history, we could well consider it to be a set of lexical numerals. This is the same problem we encountered with the Shang/Zhou and Chinese
classical systems (Chapter 8), which, not coincidentally, also are multiplicativeadditive and associated with logosyllabic scripts in which some characters (including numeralsigns) are ideograms.
Around 1921, Njoya supervised a transformation of the script into a form
known as mfmf, which altered the numerals from multiplicativeadditive to
3
Dugast and Jeffreys (1950: 4) place its invention in 1895 or 1896, although it may have
been as late as the turn of the century.
Miscellaneous Systems
323
Mende
Around 1917, an Islamic scholar named Mohamed Turay developed a syllabic script
known as Kikakui to represent graphically the Mende language spoken in southern Sierra Leone, almost certainly influenced by the Vai script of Liberia developed in the previous century (Tuchscherer 2005: 478). A few years later, Kisimi
Kamara, Turays grandnephew and student, expanded and revised the script in a
second stage. While Western numerals were always used alongside the Vai script,
the inventors of the Mende script developed a distinct set of numerical signs to
accompany the syllabary, although it is unknown which of Turay or Kamara was
responsible for the innovation. The Mende numeralsigns are shown in Table 10.6
(Tuchscherer 1996: 7175).
The system is decimal and multiplicativeadditive, and numeralphrases are
constructed with the highest powers on the right. Because the system is multiplicativeadditive, no sign for zero is needed or used. Unitsigns are placed above the
corresponding powersigns, and so numeralphrases are read from top to bottom
and from right to left. There are two signs for 10. The first, 10 (+) in Table 10.6,
combines additively with the units for 1 through 9 in order to write 11 through 19,
Numerical Notation
324
10 (+)
10 ()
100
1000
10,000
100,000 1,000,000
m
n
o
P
d
k
\\\\ba
hjm
\F
io
\\e\e\e\e\e\e
e\j\m\n\o\p\
14
128
60,009
5,555,555
while the other, 10 (), is a multiplicative powersign for 10 that combines with
the unitsigns for 2 through 9, or with 10 alone by placing a dot rather than a
sign for 1 above it (Tuchscherer 1996: 7172). The higher powersigns use vertical
strokes to indicate repeated multiplication by 10; the number of strokes represents
the exponent of 10 corresponding to the number. This feature is quite distinct
from the cumulative principle, which always refers to repeated addition of similar
symbols, and is unique to the Mende system. In theory, the system could have
been extended infinitely without using the positional principle, although there are
practical limits to how many verticalstrokes could be read easily.
Some scholars once thought that the numeralsigns for 1 through 10 derived
acrophonically from the Kikakui signs for the first syllables of the numeral words
for 1 through 10 (Tuchscherer 1996: 130132). While the syllabic values and the
numeralsigns correspond, Tuchscherer (1996: 140142) has demonstrated that the
Mende numeralsigns (at least those for 1 through 5) are also similar to certain
signs (and variants) of the Arabic positional numeralsigns. From this, he argues
that the Arabic numerals inspired the signs of the Kikakui syllabary for the first
syllables of number words. While the similarities are not striking enough to prove
the case conclusively, I am reasonably convinced that the Arabic positional numerals influenced the development of the Mende system. Yet the Mende numerals are
multiplicativeadditive, not cipheredpositional (like the Arabic positional system)
or cipheredadditive (like the Arabic abjadbased system). The only other multiplicativeadditive system used in West Africa is the earliest Bamum system, but it
Miscellaneous Systems
325
is a long way from Sierra Leone to Cameroon, and by the time the Mende system
was developed in 1921, the Bamum had switched to cipheredpositional numerals.
Moreover, the use of two different signs for 10 (one additive, one multiplicative)
and the use of repeated strokes to indicate exponents are features that are not
attested elsewhere. Thus, the structure of the Mende system should be regarded as
largely indigenous. Curiously, the modern Mende lexical numerals are not decimal but vigesimal. While this might suggest that the base of the Mende numerical
notation was borrowed from the Arabic numerals, in the nineteenth century the
Mende had decimal lexical numerals (Tuchscherer 1996: 148150). If this system
survived (in even a vestigial form) into the first decades of the twentieth century,
it, rather than a foreign numerical notation system, could have inspired the decimal base of the system.
The Mende numerals were used for a wide variety of functions, and were taught
in schools throughout the 1920s and 1930s. Some individuals used the system for
accounting and record keeping, but it is not clear whether the numerals themselves were used directly for arithmetic (Tuchscherer 1996: 69). Dalby reports that
the syllabary was used by some weavers and carpenters for recording measurements, which would presumably also require numerals (Dalby 1967: 21). Both
the Kikakui syllabary and the numerals continue to be used for some purposes,
including correspondence, record keeping, religious writings, and legal documents
(Tuchscherer 2005: 478).
SubSaharan DecimalPositional
In addition to the African systems just described, which are structurally distinct
from their ancestors, several of the indigenous scripts of subSaharan Africa have
decimal and cipheredpositional numerical notation systems, and are thus structurally identical to their Western or Arabic ancestors. While these systems are of
less interest from a structural point of view, they are noteworthy from a historical
perspective. I list these systems in Table 10.7.
The Bagam syllabary was invented early in the twentieth century in western Cameroon and used briefly by the Eghap (known in scholarly literature as the Bagam) of
that region (Tuchscherer 1999). The only text to preserve Bagam writing and numerals is a recently discovered 1917 description of the system by a British colonial military officer, Captain L. W. G. Malcolm. The numerals probably were borrowed from
the Bamum system rather than from the Western numerals, based on some graphic
resemblances between the Bamum and Bagam sign sets. The Bagam numerals do
not include a sign for zero, but do include a sign for 10. It is thus unclear whether it
was a cipheredpositional system or how (if at all) it expressed larger numbers. In the
early part of the century, the Bamum system was still multiplicativeadditive, which
Numerical Notation
326
Manding
q
L
A
V
r
M
B
W
s
N
C
X
t
O
D
Y
u
P
E
Z
V
Q
F
,
w
R
G
.
x
S
H
/
y
T
I
<
Osmaniya
<
>
Wolof
_ =
Bagam
Bt
Fula (Dita)
Fula (Adama Ba)
Kpelle
10
z
U
J
>
suggests that the Bagam system may also have had this structure. The Bagam script
and numerals are now extinct, and recent ethnographic investigations in the region
have revealed no knowledge of the numerals even among elderly Bagam (Konrad
Tuchscherer, personal communication).
The Bt numerals were invented in late 1957 or early 1958 by Frdric BrulyBouabr, a native Bt from the western part of Ivory Coast, to accompany a
syllabary of over 400 characters that he had invented a year earlier (Monod 1958).
BrulyBouabr, who was fully literate in French, did not use Western models in
developing his scriptsigns, as can be seen from the abstract nature of the numerals.
However, the use of a dot for zero shows at least some influence from the Western
numerals (or perhaps the Arabic numerals, although it is not clear whether BrulyBouabr knew Arabic at all). The unusual sign for 10 may have been used multiplicatively or additively in conjunction with the unitsigns. There is evidence of a
quinary component to the Bt system in the fact that the signs for 6 through 10
are inverted forms of the signs for 1 through 5, with the exception of the extra dot
atop the sign for 5 (Monod 1958: 437). BrulyBouabrs efforts to have this system
accepted among the Bt met with minimal success. I do not know whether it is
still used at present.
Two alphabets invented for the Fula of Mali have accompanying cipheredpositional decimal numerical notation systems. The first of these, known as Dita,
was developed by Oumar Dembl between 1958 and 1966; in keeping with his
being a woodworker, his signs have a linear character (Dalby 1969: 168173).
Dembl attended a Koranic school and spoke French, so the structure of the
system was based on either Western or Arabic numerals. The second system,
invented by Adama Ba, a Fula Muslim literate in French, before 1964, is identical
Miscellaneous Systems
327
in structure, but its signs are more curvilinear and perhaps show some influence
from Western numerals (Dalby 1969: 173174). Neither system was ever used
except by its inventor.
The Kpelle numerals were developed in the 1930s by Gbili, a paramount chief of
the Kpelle in central Liberia, in conjunction with an indigenous syllabary (Stone
1990). Its numeralsigns include a sign for 10 but none for zero, so it is not clear
how, if at all, higher numbers were written. Both Arabic and Western numerals
were known in the region, and the Kpelle script was developed on the basis of
the Vai script, which used Western numerals. Although the Kpelle signs vaguely
resemble both Arabic and Western numerals, no definite historical ancestry can
be assigned to them. The script was used traditionally for tax records as well as for
official communication among chiefs, and was restricted to a small segment of the
populace. Today, most Kpelle use Western numerals, although the indigenous system continues to be used for personal correspondence by a few individuals (Stone
1990: 141; Tuchscherer 2005: 478).
A set of numerals was developed around 1950 by Souleymane Kant, an educated trader who was literate in both French and Arabic, in conjunction with an
alphabet known as Nko (Dalby 1969: 162165). It was designed for use among the
many peoples whose dialects fall under the label Manding, most notably Mandinka, and was intended to provide a means of communication accessible without
the need for formal schooling. The numerals are cipheredpositional and decimal,
and perhaps are related graphically to the Western numerals; however, numeralphrases are written with the highest power on the right. Texts written in this script
apparently included treatises on calculation, suggesting that the numerals may
have been used for arithmetic (Dalby 1969: 163). Nko continues to be used today,
and probably has tens of thousands of users.
Around 1920, an alphabetic nonArabic script known as Osmaniya (also known
as far soomaali and cismaanya) was developed by Ismaan Yuusuf Kenadiid, brother
of the sultan of Obbia, as an alternative to Arabic for writing the Somali language
(Lewis 1958: 140142). The Osmaniya decimal cipheredpositional numerals, like
the script, were written from left to right. The fact that the script fully represented vowel sounds and was written from left to right shows influence from the
Latin alphabet, so it is possible that the numerals were mainly of Western rather
than Arabic origin, but the Osmaniya numeralsigns resemble neither Western
nor Arabic positional numerals. While Osmaniya was declared an official script in
Somalia starting in 1961, a Latinderived orthography was adopted in 1972, after
which Osmaniya was used far less regularly.
Assane Faye developed a Wolof script around 1961 that has a set of cipheredpositional numerals (Dalby 1969: 165168). Faye, who was literate in both French
and Arabic, presumably drew more influence from the Western numerals in
328
Numerical Notation
creating this system, whose signs more closely resemble Western than Arabic
numerals. Numeralphrases were written from left to right. Curiously, Faye also
assigned numerical values to nineteen of the letters of his script (19, 1090, 100)
in imitation of the cipheredadditive Arabic abjad system (Dalby 1969: 167168).
Neither the script nor the numerals survives today; most Wolof use either Arabic
or Western numerals.
Miscellaneous Systems
329
a drawing of a man with four circles (each representing one of the limbs with five
digits) joined with a cross means 22 (Griaule and Dieterlen 1951: 1112; Flam 1976:
37). Another represents a period of sixty years by three rods of decreasing size, each
with the value of 20 (Griaule and Dieterlen 1951: 28). There may not have been a
regular system of correspondences between numbers and signs. In the context of
reckoning and calculation, cowries representing 1, 5, 10, 20, 40, and 80 apparently
were used (CalameGriaule 1986: 232). The exact technique employed is unknown,
however, and this may not have constituted a numerical notation system.
While most systems of tally sticks use only onetoone correspondence (thus
lacking a base), Lagercrantz (1973: 572) reports that among the Ganda and Djaga
of Uganda and Tanzania, tally sticks are also used in which units are marked by
small notches, 10 by a larger notch and 100 by an even larger notch. It is not clear
whether this system recorded cardinal numbers, or whether it is simply a series
of marks equal to the number being counted, of which the tenth is large and the
hundredth larger still. Another tallying system, possibly of more modern origin,
was used on riverboats along the Ogowe River in Gabon in the twentieth century.
When refueling steamships, a stroke on a piece of paper was written for every
ten loads, and a cross for every hundred loads (Lagercrantz 1970: 52). Again, it is
entirely possible that this system was not used to indicate cardinal numbers, but was
simply an ordinal tally. The conceptual distinction between a system used only to
mark items as they are counted, and one used to indicate whole sums after counting should not be underestimated; nevertheless, it is quite plausible that at least
some of these African tallying systems did, eventually, transform into cumulativeadditive numerical notation systems. Regardless, even tallying systems that use
specific abstract signs for powers of a base instead of onetoone correspondence
represent a considerable conceptual advance.
Cypriot Tallies
Buxton (1920: 190) describes an otherwise undocumented numerical notation system used by nonliterate Greek speakers in Cyprus:
The numbers are continually used as follows: a perpendicular stands for a unit, five is
sometimes indicated by a cross and sometimes by a circle, ten either by a circle, by a
theta, or by a cross inside a circle, twenty by a cross inside a circle, where that symbol
has not already been utilized previously; if it has, there seems to be no alternative.
Fifty is written by a loop on top of a perpendicular, and a hundred by two fifties. It
will be seen that two of these symbols are not dissimilar to Arabic numerals, namely,
the circle and the symbol for fifty. The Arabic symbol for five is, however, not circular, and it is possible that the two signs are connected, but the value of the looped
line is in Arabic nine, not fifty. ... At Enkomi a man scores at cards in this way. He
330
Numerical Notation
chalks down units up to four, then he rubs them out and writes a circle, adds units
to ten when he erases them, and draws a line through the circle, draws units up to
fourteen then adds a circle; at twenty he erases the added nine and draws another
line through the theta, which thus becomes a circle with a cross through it.
This evidence indicates that although numeralphrases are constructed sequentially as tallies, rather than being written as a single sum, because intermediate values are erased and replaced with higher values, ultimately the result is a cumulativeadditive numeralphrase with a base of 10, a subbase of 5, and a special sign for 20.
This sort of notation is qualitatively different from simple onetoone correspondence,
or tallying in which intermediate marks are not erased but simply continue on (e.g.,
XXVII vs. IIIIVIIIIXIIIIVIIIIXIIIIVII as two notations of 27). Other than this one
brief description, however, we have no information on the origin, history, or use of the
Cypriot system. While there are parallels between this system and the decimal cumulativeadditive system used in ancient Cyprus (Chapter 2), there is no reason to think
that this system is anything but a locally developed technique, one that is evidently
idiosyncratic given the multiple numeralsigns used for 5, 10, and 20.
Indus
The writing system of the Harappan civilization, centered in the Indus River
valley, is one of the great remaining mysteries in the field of script decipherment. It
was used from around 2500 bc to 1900 bc on several thousand very short inscriptions (averaging five signs per text), and was written primarily from left to right
(Parpola 1996). Unfortunately, there is no reliable basis for deciphering the script,
because the language it represents is unknown (though sometimes asserted to be a
Dravidian language) and there are no bilingual inscriptions. The situation is even
more grave than for scripts such as Linear A, where there are many easily readable
numeralphrases and associated ideograms (see Chapter 2). Many dubious interpretations of Indus numeration have been proposed (e.g., Subbarayappa 1996). We have
barely enough evidence to confirm the existence of a numerical notation system in
the ancient Indus Valley, much less determine its origin, history, or function.
There have been several earnest attempts to decipher the Indus numerals,
mostly relying on the very frequent occurrence of groupings of vertical strokes
on the inscriptions. Table 10.8 shows these numerals as well as the frequency with
which they are encountered in the texts (Fairservis 1992: 62).4
4
Fairservis (1992: 183) provides no count of single and double short strokes because these
are also assigned grammatical functions (as genitive and locative case markers, respectively) in his decipherment.
Miscellaneous Systems
331
Long strokes
10
111
111
1111
\111
1111
1111
11111
\1111
11111
11111
38
70
11
111
1111
111
\11
151
70
38
aa
aaa
aaaa
aaaaa
aaaaaa aaaaaaa
149
365
314
64
22
These signs probably represent low numbers in a cumulative fashion; the short
strokes are grouped into sets of three, four, or five, just as the signs of most other
cumulative systems. The longer ungrouped vertical strokes occur only in the early
Indus inscriptions; during its mature phase, the shorter strokes were used exclusively (Parpola 1994: 82). Because these sets of strokes are paired interchangeably
with nonnumerical graphemes (e.g., the fish sign + is attested in combination
with three, four, six, and seven strokes), we are relatively confident that they were
numeralsigns (Parpola 1994: 81). Yet Ross (1938) long ago pointed out that some
groupings of vertical strokes pair noninterchangeably with other signs, which suggests that they may have had phonetic or grammatical values (Ross 1938; Fairservis
1992: 12). This is parallel to the frequent use of numeralsigns phonetically in
Chinese writing, and resembles abbreviations such as K9 for canine in English.
One enigmatic symbol consisting of three rows of four vertical strokes occurs
frequently, but never in the same contexts as other putative numerals; Fairservis
(1992: 71) argues that it should be read as rain, which may or may not be correct,
but is far more likely than 12. The Indus texts are so short and devoid of contextual information that we must be very careful not to read too much numerical
information into them.
This interpretive framework for the Indus numerals does little to establish
whether this system had a base and used an interexponential principle to write
larger numbers. Fairservis notes that there is a sharp dropoff in frequency after
seven for both the long and short vertical strokes, and that in fact there are no
attested instances of eight or more long strokes. From this, he concludes that the
Indus numerals were probably octal or base8 (Fairservis 1992: 6162). Perplexingly,
however, he then proceeds to assert that there are pictographic signs for 8, 9, 10,
and 11 that were simultaneously numerical and calendrical, indicating the eighth
through eleventh months of the conjectural Harappan calendar, because these four
signs, along with vertical strokes for 1 through 7, are found in association with a
sign that he thinks represents month (Fairservis 1992: 65). This theory has not been
widely adopted by scholars of the Indus script (cf. Pettersson 1999: 103).
332
Numerical Notation
Our best evidence for a legitimate Indus numerical notation system comes from
nine inscribed potsherds and copper and bronze tools found at Mohenjodaro, Canhujodaro, and Kalibangan, inscribed with sets of vertical strokes and crescents/
hooks (and sometimes other scriptsigns). These two signs are sometimes found in
combination on the seal inscriptions, but never in large quantities and never clearly
separated from the rest of the text. Pettersson (1999) adds that, in addition to vertical
strokes and crescents, a distinction needs to be made between vertically and horizontally oriented strokes. One object, a chisel or axe blade (DK7535)5 from Mohenjodaro, contains all three signs, as shown in Figure 10.4 (Parpola 1994: 108).
While I am reasonably convinced that this inscription and similar ones on other
Harappan tools are numerical in function, there is no agreement as to the specific
structure and value of the signs. Fairservis (1992: 6769) has constructed a convoluted argument whereby the vertical strokes (standing for units) can serve either an
additive or multiplicative role in the numeralphrase depending on whether they
follow or precede the crescent sign(s). Pettersson (1999: 102103) points out, however, that there is no case where a crescent sign is both preceded by and followed by
vertical strokes. Fairservis (1992) and Pettersson (1999) argue that because none of
these nine objects contains more than seven of any sign, the Indus numerals must
be octal rather than decimal. Yet Parpola (1994: 82) argues that the crescents probably represent 10 rather than 8. Either of these interpretations of the system would
mean that the Indus numerical notation system was cumulativeadditive.
At present, there is insufficient evidence to decide whether the crescentsign had
a value of 8 or 10. Nine numeralphrases is a very limited corpus from which to
conclude that, since no sign is repeated more than seven times, the numerical base
There is some confusion over the identification of this object, which is assigned different
artifact numbers by Parpola (1994) and Pettersson (1999).
Miscellaneous Systems
333
Naxi
Naxi (also known as Nakhi and Moso) is a TibetoBurman language spoken
by approximately 250,000 people in the northwestern part of Yunnan province
in southwestern China. Naxi is written in three indigenous scripts: dongba (or
tomba), a pictographic notation system, and two syllabaries (geba and malimasa),
in addition to a more recent Latinbased orthography. The dongba script is highly
idiosyncratic, consisting of 1,5002,000 largely pictographic signs with some phonetic components, although there is not a regular correspondence of signs with
334
Numerical Notation
either words or phonemes. It is used primarily as a mnemonic aid or promptbook to assist priests in reciting memorized texts (Bockman 1989: 155). It was
reputedly invented in the twelfth century ad (Coulmas 1996: 353). The earliest datable dongba texts, however, are from the middle of the eighteenth century (Bockman 1989: 153). The dongba script is actively used by some Naxi priests, and has
even been part of modern literacy programs. Nevertheless, because dongba texts
are pictographic and rely on oral and mental knowledge to draw meaning from
them, their interpretation by Western scholars is incomplete and poor.
In at least some dongba texts, numerical notation was used alongside the
script. In the Nichols manuscript first made available to Western scholarship by
F. H. Nichols in 1904, three repeated South Asian swastikalike signs precede
six vertical strokes. These are interpreted by Rock (1937: 236) as representing 100
and 10, respectively, producing a sum of 360, indicating the 360 yuma deities
of the Naxi. Groups of three or more signs are clustered in rows of three signs,
where appropriate. In other dongba manuscripts, a simple cross rather than a swastika represents 100. This is a cumulativeadditive numerical notation system with
a base of 10, leaving open the question of how the number 1 was represented.
Bockman (1989: 1952) suggests that the dongba signs X, +, and were numerical,
but does not assign them specific values. In other dongba manuscripts, however,
vertical strokes or hooked vertical strokes mean 1, and X or + means 10; in one
very clear instance, Hs. Or. Sim. 279 / R. 1912, Blatt 9r 10r, eighteen consecutive
panels depict gods, each enumerated using this system (Janert and Janert 1993:
2753). Like many cumulativeadditive systems, the units 6 through 9 are represented in two or more rows of three to five strokes (3 + 3, 4 + 3, 4 + 4, 3 + 3 + 3); 5
is depicted with a single row of five strokes, but 15 is depicted with X (10) followed
by 5 indicated in two lines of three and two strokes respectively.
The variability in signs suggests that basestructured numerical notation was
used only irregularly or idiosyncratically in the dongba texts, with local or even
individual scribal tradition determining which signs represented which numbers.
Against this position, however, a wide variety of dongba texts contain numeralsigns, and all appear to be cumulativeadditive and decimal (i.e., signs are repeated,
but no sign is repeated more than nine times). The origin of this numerical notation system, and its relation to any other system, remain obscure. It is possible that
it was independently invented, as no other cumulativeadditive systems were ever
used alongside scripts of either East Asian or South Asian origin.
Varang Kshiti
In the twentieth century, several scripts were developed for the various Munda
languages of central and eastern India, of which Sorang Sompeng, Ol