Descriptive Semantics of 7zeltaZ Numeral Classifiers1

BRENT BERLIN AND A. KIMBALL ROMNEY

Stanford University
N MANY languages, counting something is associated with classifying
I or modifying indicators about the thing being counted. I n Tenejapa
TzeltaP such classification or modification is indicated by the presence of
expressions referred to here as numeral c l a ~ s i j e r s . ~ The purpose of this paper is
to explore some semantic implications of this series of numeral classifiers.
The numeral classifiers of Tenejapa Tzeltal are a series of predominantly
monosyllabic, bi-morphemic stems of the canonical shape CV(h) C. They
normally occur as the second member of a numeral phrase where the first
member is a numeral, i.e., in the form:
numeral + numeral classifier k noun.
Numeral phrases are typically followed by a noun (the noun is implied when
not explicitly stated), and some semantic features of the noun are invariably
indicated by the numeral classifiers. (See Appendix A for a sketch of the lin-
guistic features of numeral classifiers in this dialect of Tzeltal.)
Semantically, these forms function very much as adjectivals in English,
specifying certain qualitative features of the referents of nouns. If it were not
for the long precedent in linguistic circles to refer to such forms as numeral
classifiers, a more functionally and descriptively appropriate term would be
nominal qualifiers. Their occurrence in numeral phrases is relatively unim-
portant in considering their broad semantic features.
The major concern here is with denotative meaning as defined by Osgood
(1961: 102), namely, as a conventional, habitual correlation between a non-
linguistic perceptual pattern, ( S ) , and some particular linguistic response,
( R) . Connotative meaning is discussed briefly in the latter part of the paper.
Before turning to the more technical aspects of the subject, it may be use-
ful to attempt to communicate an intuitive understanding of the role of
numeral classifiers in Tenejapa Tzeltal (referred to as TTzel in the remainder
of the paper). Consider the following phrases in English:
three square blocks
three round blocks
The words square and round in these phrases occur after a numeral, they
are followed by a noun, and they indicate something about the noun, i.e., its
shape. I n TTzel, numeral classifiers have these same characteristics as well as
numerous additional ones. For example, in English, it is possible to omit the
modifying words without altering the grammatical appropriateness of the
phrase. However, unlike English adjectivals, TTzel numeral classifiers are
analogous to Navaho verbal categories (Hoijer, 1951 ; Carroll and Casagrande,
1958) in that their occurrence in specific numeral phrases is obligatory. Thus,
79
80 Transcultural Studies i n Cognition
in instances of enumeration of objects, events, or actions, a TTzel speaker
must make a decision as to which of a number of alternative classifiers to
employ in terms of some qualitative feature of the object, action, or event. As
we shall see later, shapes, positions of objects, certain actions performed on
objects, etc., are domains of meaning indicated by classifiers. On the other
hand, colors of objects are never indicated by numeral classifiers.
I n this paper we first present the procedures used for eliciting an extensive
list of classifiers. Then a preliminary taxonomy of classifiers into gross semantic
categories is given, followed by a detailed examination of selected categories of
classifiers covering narrow semantic domains. Finally, the general results are
discussed and placed in the context of broader studies of semantics and
meaning.
Elicitation of the list of numeral classiJers
Knowledge of the highly regular canonical shapes of numeral classifiers
(i.e., CV(h)C), the position of their occurrence, and the segmental phonemic
inventory of TTzel allowed us to elicit most, if not all, of the classifiers in
the dialect. Prior to working with informants, we generated a list of all the
phonemically permissible CV(h)C forms (a finite list of 4,410 forms). We
then devised an eliciting frame that would enable a native speaker to report
which of the generated forms actually functioned as a numeral classifier in
TTzel.
The generated list was presented independently to two informants who
had been trained to read and write phonemically. These informants were
instructed to work through the generated list, asking themselves the following
question :
/ya bal stak kahtahtik te bitik
/
Is it possible for us to count things, one by one,
with the form ?
All generated CV(h)C forms were tested in reduplicated form to indicate con-
tinuous enumeration; e.g., beginning at the head of a list of possible CV(h)C
forms, for example:
/ya bal stak kahtahtik te bitik s-pipip-e/
/ya bal stak kahtahtik te bitik s-pihpipihp-e/
/ya bal stak kahtahtik te bitik s-pepepep-e/
/ya bal stak kahtahtik te bitik s-pehpepehp-e/
/ya bal stak kahtahtik te bitik s-papapap-e/
/ya bal stak kahtahtik te bitik s-pahpapahp-e/
etc.
etc.
Forms with which things could be counted were then further checked by
determining a partial list of what things could be enumerated (i.e., a list of
the nouns occurring with such classifiers). This check was accomplished by
employing the question:
/bitik hehEuk ya stak 7ahtael te
What things can one count with the form
-4
-e?
Descriptive Semantics of Tzeltal Numeral Classi$ers 81
substituting the previously isolated CV(h)C forms in the blank. If the forms
were classifiers, the informants were able to list a series of nominals which in
turn could be placed in the sequence mentioned earlier, i.e., numeral+numeral
classifier & noun.
Thus far, some 557 numeral classifiers have been isolated along with lists
of representative nominals occurring with each classifier. This number is con-
siderably greater than any other inventory thus far published. Without the
systematic eliciting procedures described above, less than a tenth of the classi-
fiers in TTzel would have been discovered, for their actual occurrence in tex-
tual material is infrequent. The list may not be complete, but i t approaches
the full inventory.
Simple taxonomy of total list of numeral classifiers
Preliminary analysis suggested that the classifiers fall into a small number
of gross semantic domains. For example, description of the state of certain
items enumerated by specific numeral classifiers was facilitated for the in-
formant by the manipulation of material objects, i.e., one could demonstrate
the difference in /hlihk laso/ one piece of lazo in its natural state and
/hbehE laso/ one piece of lazo wrapped in single wraps around long, sticklike
objects by pointing and manipulating the lazo into its various states. A whole
series of items could be arranged in the appropriate manner to their respective
numeral classifiers and talked about afterwards.
Other classifiers, however, could be demonstrated only as the action was in
progress; e.g., if enumeration were to occur, i t must be done during the process
of the action.
Still other numeral classifiers seemed to refer to the sounds of a particular
action, while others referred to items which had not received any human ac-
tion, i.e., were natural in the world, as i t were.
The observations led to the formulation of a series of questions that allow
the classification of numeral classifiers into a simple taxonomy. The three main
questions are:
I. /bitik hehtuk te kalal ya kiltik ya stak eahtael ta bavil sbil ta ?alel/
What things have names (classifiers) which are readily known on observing the objects?
11. /bitik hehEuk te ya stak eahtael te kalal ya stak pasele, sok ian ta patil, ya stak San
?ahtael sbil ta me ya B hi1 te bitike/
What things have names (classifiers) which are employed for the enumeration of some
action, as well as the objects which are the result of some action?
111. /bitik hehtuk te ya stak ?ahtael te kalal ya stak pasel, h ha? te ta patil, ma 5 Eikna sbil
ta ?ahtad/
What things have names (classifiers) which are the enumeration of some action, but
after (the action) enumeration is not possible?
Informants were asked these questions about all classifiers. Classifiers fitting
question one were further distinguished in terms of whether or not enumera-
tion referred to objects which were the result of human action. Classifiers
fitting question three were subdivided in terms of whether or not the enumera-
tion referred to the acoustic result of a particular action. This procedure re-
sulted in the following taxonomic classification:
82 Transcultural Studies i n Cognition
I. Numeral classifiers of states (no action during enumeration).
A. Referring to items made by man.
B. Referring to items not made by man.
completion of artion).
A. Referring to actions of living beings.
I). Referring to acoustic results of actions of living beings.
11. Numeral classifiers of states and/or actions (enumeration during process and/or after
111. Numeral classifiers of actions (enumeration imperative during action).
Attempts to further partition this taxonomy from the top have been
unproductive. However, the classifiers do fall into a large number of semantic
domains or categories. These subgroupings are discussed in the following
sections.
Classification of numeral classijers into seinaiilic domains
I n this section we discuss, in some detail, the analysis of small sets of nu-
meral classifiers that are interrelated by virtue of common membership in a
semantic domain or category.
An example of a semantic domain in English is shape. Thus, words such
as round, square, rectangular, etc., may each be thought of as sharing
the feature of saying something about shape. They signal the hearer that
the aspect being talked about is shape. I n addition, each word in the domain
says something different, e.g., round is different than square. Shape is the
gloss for a semantic domain or category. Round, square, etc., are mem-
bers of the category.
As Hymes has observed, All utterances occur contrastively in contexts,
(1962:19). When a TTzel speaker is selecting a numeral classifier to character-
ize the particular shape or state of an object in his physical environment, he
first must select a context which allows for some limited range of meanings,
e.g., he decides to talk in the context of (shape rather than color. This con-
text includes presumably the range of meanings of the specific numeral classi-
fier arid the ranges of meanings of all other semantically appropriate classifiers
which could conceivably be alternatives. For example, a classifier occurring in
a context dealing with the enumeration of objects in terms of some Components
of shape would presumably not be an appropriate alternative for the context
of kinds of actions of eating. One may view classifiers as linguistic units that
indicate certain semantic components of objects in some specified contexts.
The goal of explicating the semantic features of classifiers is thus seen as two-
fold. The first step is to discover the context(s) indicated by each classifier.
Secondly, one must determine the criteria1 attributes of each classifier within
each context. The context and the classifiers occurring in it are defined here as
a semantic domain or category.
We have isolated over one hundred semantic categories on the basis of
similarity judged by native informants. By the application of a simple TTzel
question, /7ay yicoh shoy/ does it (any particular numeral classifier under
consideration) have a close companion? informants were able to sort numeral
classifiers into semantic categories. These domains range in membership from
Descriplive Semantics of Teellal Numeral Classijers 83
two to as many as 20 classifiers per set. Some of the classifiers were not placed
by our informants into any category, i.e., were considered as isolated, and
have not been examined in detail. All numeral classifiers were independently
grouped into semantic categories by two informants. There was agreement
between them in placing classifiers into semantic categories.
Analysis o j itilerizal slruclure of semaialic calegories
I n this section we examine differences in the internal structure of various
categories of numeral classifiers. Two basically distinct types of distributional
and semantic characteristics of these categories have been isolated. I n the first
type, the numeral classifiers are in contrast, both in distribution and in denota-
tive meaning. I n the second type, the numeral classifiers are in complementary
distribution (or free variation) and are denotatively synonymous.
Before presenting TTzel data, a sketch of the method used and an illustra-
tion from English are given.
Assume a set of related modifiers (M) (i.e., modifiers in a single semantic
domain), together with a list of nouns (N) typically occurring as objects of
each of the modifiers. The problem is how to find whether any two modifiers
have the same or different denotative meaning.
Our assumptions are as follows: (1) If two modifiers, MI and Mz, occur
with the same noun, N1, and if the two resulting phrases, MI NI and MzN1,
can be shown to differ in denotative meaning, tlaerc the modifiers, M1 and Mz,
differ in denotative meaning. (2) If two modifiers, MI and MP, never occur
with the same noun, N1, except in free variation), and i j in two phrases, MINI
and M2N2, they have equiavalent denotative meaning, then they are allosemes,
i.e., they are equivalent in denotative meaning.
These assumptions may be illustrated by considering the following cate-
gories in English:
(1) cup of sugar
cube of sugar
and
(2) school of fish
herd of cattle
I n the first case, we have cup (MI ) and cube (M2) occurring with
sugar (N1). They differ in denoting different iamounts and states of the
modified object, sugar. Thus, cup and cube contrast in denotative meaning
in this context. I n the second case, we have school (MI) and herd (Mz)
occurring with different nouns. Both indicate something like a natural group
of some animal. Knowledge of the noun is sufficient to predict which modifier
is appropriate. The two modifiers never occur with the same noun (in this con-
text) with contrastive denotative meaning. We call them allosemes. Examples
of these two types in TTzel are given below.
Type I. Semaiitic categories with numeral classifiers ita contrast. For each
category of numeral classifiers, we constructed a table showing occurrence of
each classifier with a representative list of nouns. The nouns were elicited
84 Transcultural Studies i n Cognition
independently from two informants and rechecked in questions requiring the
joint occurrence of the classifier and noun in the same numeral phrase.
Table I presents the data for a pair of numeral classifiers which form a
semantic category that may be glossed as aggregation of globular objects.
The classifiers /buhs/ and /tol/ are said to be in contrastive distribution,
i.e., they occur in identical linguistic environments with a difference in niean-
ing (meaning is here used identically to Hjelmslevs use of content). To demon-
strate contrast between any two classifiers, we are primarily concerned with
establishing, by informant response, same or different. J ust what the differences
may consist of is a related, but quite distinct, question of the delineation of
criterial attributes.
The denotata of the nouns / Wm/ corn, /kahpe/ coffee, /kaSlan Eenek/
TABLE I. DISTRIBUTION OF NUMERAL CLASSIFIERS IN AGGREGATION OF
GLOBULAR OBJECTS CATEGORY
(+ indicates permissible occurrence, - indicates that occurrence is not permissible.)
Classifiers
Nouns
/buhs/ /tol/
+ +
/?iSm/ corn
+ +
/kahpe/ coffee beans
+ +
/kaglan Eenek/ peanut
+ +
/W/ chili pepper
+ +
/ton/ stone
+ +
/ma#/ corn dough
+ +
/tornut/ egg
etc.
peanut are such that (1) they occur in at least two distinct natural states that
permit their classification with the two classifiers /tol/ and /buhs/ in this
context and/or (2) said denotata can undergo certain actions or manipula-
tions such that they acquire the requisite criterial attributes indicated by the
two classifiers.
One might state the semantic relationship as follows: /tol/ and /buhs/
are labels that indicate separate or distinct bundles of criterial attributes.The
denotata of the nouns of the set are such that when they occur with /tol/
they manifest the criterial attributes indicated by that classifier, as these
contrast with the criterial attributes indicated by the classifier /buhs/.
Conversely, when the denotata of the nouns of the set occur with /buhs/, they
manifest the attributes indicated by that classifier as they contrast with those
indicated by /tol/.
Assuming that contrast has been established between /tol/ and /buhs/,
it yet remains to describe the dimension(s) af contrast between the two forms,
that is, to delineate their distinctive criterial attributes.
Descriptive Semantics of Teeltal Numeral Classifiers 85
The procedure utilized here is to instruct informants to produce and/or
manipulate objects into states appropriate to the contrasting classifiers. Such
observation was made feasible in TTzel due to the linguistic feature (see Ap-
pendix A) that most classifiers are derived from transitive verb or positional
verb roots, the direct objects of which are identical to the nouns occurring
with the respective numeral classifiers. Therefore, informants could be in-
structed to perform the action of the particular verb (from which the classifier
was derived) on a particular object.
I n the set given above, knowledge of the derivational structure of the two
classifiers indicated that they are derived from the positional verb roots /tol/
and /buhs/. By inflecting the derived positional verb root in the imperative
mode and providing a nominal occurring in the list of nouns in the set, /ma#/
corn dough, the following request is constructed:
/tolana ma#/ perform the action indicated by the derived positional verb root /tol/ on the
object designated by the noun /ma#/
The informant then performs the action on the object /ma#/ corn dough
placing i t in a state appropriate for enumeration by the classifier /tol/.
One then proceeds with such a frame until the denotata of all the nouns
occurring with both /tol/ and /buhs/ manifest the criterial attributes, or
are in the appropriate state, of those classifiers. E.g.,
/tolana ?iSim/
/tolana kahpe/
/tolana Eenek/
etc.
/buhsana ?iHim/
/buhsana kahpe/
/buhsana fenek/
etc.
Such a procedure was followed independently with two TTzel informants
providing each of them with a quantity of beans, peanuts, corn dough,
etc., to be manipulated. The results of the actions performed on each of the
objects facilitated the definition of the criterial attributes of the classifiers.
Table I1 gives the definitions for the semantic domain under consideration.
It may be noted that no etic grid is available to descriptive semantic
TABLE 11. DISTINCTIVE MEANING OF CLASSIFIERS IN DOMAIN OF
AGGREGATION OF GLOBULAR OBJECTS
Category
Numeral
Classifiers
Criteria1 Attributes
aggregations of globular-shaped aggregated in a manner such
that maximal horizontal extention of
items is achieved with minimal spac-
ing between items
objects, as corn kernels, coffee beans,
peanuts, chili peppers, stones, pieces
of corn dough, eggs, etc.
/buhs/
aggregated in a manner such
/tol/ that maximal vertical piling is
achieved
86 Transcullural Studies in Cognition
analysis of the kind undertaken here. The resulting emic analysis must per-
force be stated in everyday language until descriptive semantics devises a
metalanguage in which the data may be discussed. However, given the verbal
glosses and some visual aid in the form of drawings or photographs, a non-
native TTzel speaker can manipulate groups of objects in such a way so as to
elicit predictable responses of numeral classifiers from native speakers. As such,
he can fulfill, partially at least, Goodenoughs rather stringent requirements of
behaving . . . in ways which lead to the kind of responses from the com-
munitys members which our theory would lead us to expect (1957: 168).
A variation of the distribution of perfectly contrasting classifiers (as in the
category including /buhs/ and /tol/) is one where the distribution of nouns
with classifiers is defective, that is, where some but not all classifiers occur with
TABLE 111. DISTRIBUTION OF NUMERAL CLASSIFIERS I N QUALITY OF
BREAKAGE CATEGORY
~_ _ _ _ _ _ _ ~_ _ ____ -
-. -
~_ _ _ _ ._ .
Numeral classifiers
b y / /PW/ / P W /kahs/
- Nouns
+
+
-
+
+
+
+
+
+
-
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
/Eahan takin/ wire
/te7/ tree
/te?el k!,im/ stem of corn
/wale? (sugar cane
/lapis/ pencil
/bak/ bone
/si?/ fire wood
/matit/ machete
/kuEiyo/ knife
etc.
lists of nonoverlapping nouns and where some but not all classifiers occur with
lists of identical nouns. This is the most common distribution observed in our
data. An example of such a semantic category is given in Table 111.
Following procedures identical to those given for the preceding semantic
category, the criteria1 attributes have been isolated and defined in Table IV.
Summary observation of the distribution of nouns and classifiers in Table I11
indicates that /huy/, /pub#'/, /pih/, and /kahs/ appear to contrast by
virtue of their occurrence in the same environment with the nouns /Eahan
takin/ wire and /te?/ tree. The denotata of all nouns in the group, however,
do not have this freedom of distribution. /wale?/ sugar cane, for example,
may occur with each classifier in the category other than /huy/. The denotata
of the nouns /lapis/ pencil, /bak/ bone, and /si7/ firewood may fulfill the
criteria1 attributes indicated by the classifiers /pih/ partial breakage and
/kahs/ complete breakage, but not of the classifiers /huy/ bend and return
and /pub$'/ bend and return with indication of bend.
Descriptive Semanlics of Tzellal Numeral Classijers 87
TABLE I V. DISTINCTIVE MEANING OF CLASSIFIERS IN DOMAIN OF
QUALITY OF BREAKAGE.
-- ~ ~.
Category
Numeral
Classifiers
Criterial Attributes
degree of breakage or bending in /huy/ bends in flexible materials such
thin, oblong objects, as pieces of wire, that no actual breakage is ob-
trees, stems of plants, bones, etc. served, that is, the object returns
to its original position after bending
/puhb/ bends in certain materials such
that breakage is observed, but not
through surface of object, such as
the bends in a wax match
/pih/
partial breakage such that the sur-
face is actually broken, as cracked
pencil
/kahs/ complete breakage
Another paradigm exhibiting defective distribution of nouns i n reference
to the classifiers with which they occur is seen in Table V.
The criteria1 attributes of classifiers i n this category were studied by
instructing informants to form /ma#/ corn dough into each shape. Photo-
graphs were taken of forms produced independently by different informants.
Our best description of each of these shapes is as follows:
1) round, disklike forms
a) well-formed, perfect disklike, /sehp/
h) not well-formed, pressed disklike, /len/
c) watery, lacking body, /lehs/
TABLE v. DISTRIBUTION OF NUMERAL CLASSIFIERS I N SHAPE CATEGORY
~-
Numeral Classifiers
tuhl kaht tel sehp len lehs kuh kul tel wol-no1 hen Yo1
Nouns
/ton/stone
/ma#/corn dough
/7a#a~/salt (block)
/7askal/sugar (block)
/ kJa /box
/%?/chili pepper
/?al~d/oranpe
/?on/avocado
/holol/head
/?ah0al/mud
/wit/mountain
/&7/defecation
/i!umante7/tree trunk etc.
88 Transcultural Studies in Cognition
2) round, spherelike forms
a) small sphere, /wol-nol/
b) large sphere, /kol/
c) large sphere, flattened bottom, /hen/
a) thick-bodied, nonpointed head, /kuh/
b) thick-bodied, pointed head, /kul/
c) thin-bodied, nonpointed head, /tel/
4) oblong, nonstanding, wedgelike forms
a) wedgelike, symmetrical ends, /tel/
b) wedgelike, nonsymmetrical ends, /tub$'/
c) cylindrical ends, /kaht/
3) oblong, standing, moundlike forms
Corn dough is an object that may be manipulated in such a manner as to
acquire the attributes of the twelve classifiers in the domain. /ton/ stone,
however, is not malleable. For a stone to take any or all of the classifiers,
it must occur in a natural state with the criteria1 attributes of each classifier
inherently present. Two informants were instructed to search independently
for stones which inherently were of the shapes indicated by the classifiers in
the domain. Stone shapes were compared with the idealized shapes produced
in corn dough. There was a degree of similarity of the forms arrived at inde-
pendently by each informant and in each medium, i.e., corn dough and rocks.
I n reference to the breakage domain and the shape domain, an important
question arises as to whether the defective distribution of nouns with classifiers
affects our earlier statement that each of the classifiers contrast. We would
say it does not on the following grounds:
I n phonemic analysis, for contrast to be established, two distinct phones
must occur in identical phonetic environments with a response from the
informant that the two utterances are different. I n English, for example, the
phonetic segments [n] and [a] are classed as phonemically different by virtue
of their occurrence in two identical phonetic environments with a response of
different from English informants, e.g.,
[ Ibiyn.] bean
[IbiYU.] bing
This contrast is sufficient to establish a phonemic difference for [IJ].
It is not required that the contrast be manifest in all possible environments.
For English, it would be impossible to demonstrate such a contrast as that
given above in the environment #-V . , , #in some such sequence as:
[#.nat.#] /nat/ knot
[#.nat.#l* /gat/* ?
due to the defective distribution of the phonetic type [IJ]: it does not occur
initially after open juncture.
Thus, to return to our classifier examples, contrast in at least one identical
environment is sufficient, and we may state confidently that /huy/, /pub#'/,
/pih/, and /kahs/ contrast, as well as /sehp/, /len/, /lehs/, /tub'#/,
/kaht/, /tel/, /wol/, /kol/, /hen/, /kuh/, /kul/, and /tel/.
What explanation, if any, is available for the defective distribution? I n
linguistics, such a feature is assumed to be an arbitrary characteristic of the
Descriptive Semantics of Tzeltal Numeral Classifiers 89
system. An examination of our tentative criterial attributes, however, in
reference to the denotata of each noun, allows for certain speculations to be
made for numeral classifiers.
We suggest that the denotata of some nouns in each set are such (by some
inherent characteristics) that they would not normally manifest the semantic
features of the classifiers with which they do not occur. Thus, while /bak/
bone can be broken parlially (placing it in a state appropriate to the criterial
attributes of /pih/) or broken completely (placing it in a state appropriate to
the criterial attributes of /kahs/), in no uncortirised context would /bak/
normally be manipulated in a manner appropriate to the so-called bending
classifiers /pub#'/ and /huy/. Similarly, the blade of a /matit/ machete may
be manipulated in a manner appropriate to the bending classifier /huy/ or it
may be snapped in two, placing i t in an appropriate condition for the attributes
indicated by /kahs/. The inherent physical features of machete blades, how-
ever, would not normally allow them to manifest the bending attributes of
/pub$'/ or the broken attributes of /pih/.
A similar argument can be made for the classifiers in the shape domain.
Thus, while /kaSa/ box may exhibit on some occasions the criterial attributes
indicated by the classifier /kaht/, i t would not normally occur with the classi-
fiers /wol-nol/, /kol/, etc., which indicate formal features (i.e., spherical) in-
appropriate to boxes.
Type I I . Semaiatic categories with numeral classijers in complementary
distribution. A second basic distribution of classifiers and nouns and one that
has quite distinct semantic implications is one in which each classifier in a
category may occur with lists of unique, non-overlapping nouns. That is, no
noun occurs with more than one classifier, the classifiers being in complemen-
tary distribution. An example of such a set is seen in Table VI.
Following the procedures for the delineation of criterial attributes, the
TABLE VI. DISTRIBUTION OF NUMERAL CLASSIFIERS IN VERTICAL POSITION CATEGORY
Numeral classifiers
/tuhf/ /tuhE/
Nouns
-
+
+
+
+
+
+
+
+
+
+
-
/pin/ earthen jar
/pis/ small glass
/limete/ liter bottle
/kib/ large pitcher
/turnantel/ tree trunk
/te?/ tree
/wale?/ sugar cane
/te?el 107bal/ stem of banana tree
/te?el ?alZa:/ stem of orange tree
/matit/ machete
etc.
90 Transcultural Siudies in Cognilion
following glosses can be given: /tuhE/ : maintenance of vertical position of
certain cylindrical objects such as jugs, glasses, bottles, pitchers, etc., by
means of their own support. Base of the object must sit directly on surface of
table, ground, etc. /tuhE/: maintenance of vertical position of certain objects
such as tree trunks, tree stems, sugar cane stems, etc., by being inserted in
ground or being held i n hand or other outside support. Base of object must be
submerged.
Another example of a semantic category in which the classifiers are in
complementary distribution is given in Table VI I .
This domain is comprised of classifiers utilized in the enumeration of cer-
tain actions performed on specific objects. I n this instance, the actions enu-
merated all refer to plucking, picking, or digging; the objects are parts of
plants, generally fruit, but including as well leaves and tubers. The set as a
whole may be glossed as the semantic domain of haruesl.
These tentative attributes are offered as criterial:
/tul/ harvesting of small beanlike, leaf-like portions of plants by plucking
/Euy/ harvesting of bananas (by cut of stalk)
/kah/ harvesting of sugar cane and mature corn
/ka?/ harvesting of immature corn
/kohk/ harvesting of predominantly tree fruit
/hohk/ harvesting of underground tubers
The classifiers i n the two domains presented above cannot be said to con-
trast in reference to the nouns with which they occur. We suggest that they be
TABLE VII. DISTRIBUTION OF NUMERAL CLASSIFIERS I N HARVEST C4TEGORY
Numeral Classifiers
tul Euy kah ka7 kohk hohk
Nouns
+
+
+
+
+
+
-
+
+
+
+
-
-
-
+
+
+
/ba#il ?it/ chili pepper
/wamal/ leafy plant
/niEim/ flower
/Eenek/ bean
/bok/ leafy vegetable
/kahpe,l koffee (bean)
/lo?bal/ banana (stalk)
,/wale?/ sugar cane
/?iSim/ corn
/?ahan/ young corn
/7alEag/ orange
/lima/ lemon
/ha?as/ zapote
/tawtaw?iE/ sweet green pepper
/EPin/ potato
/kaSlan Eenek/ peanut
/#aha1 tPin/ sweet potato
etc.
Descriptive Semantics of Tzeltal Numeral ClassiJiers 91
considered as allosemes of some higher order sememe. This conclusion is based
on the evidence that (1) the classifiers occurrences can be predicted by refer-
ence to the nouns in these particular domains (Leal they are in complementary
distribution) and (2) their occurrences in the same semantic category allow for
the inference of their semantic similarity.
The features of complementarity and semantic similarity can be made
clearer by reference again to linguistics. One goal of phonemics is to classify
distinguishable phones into larger-order units, or phonemes. Two principal
criteria are employed: Complementary distribution and phonetic similarity.
(The following discussion has been aided greatly by reference to an analogous
outline used by Roger Brown [1956: 252-541.) Complementary distribution
can be established for two distinct phones when there are no environments in
which they both occur. I n such cases, the . . . class [or phone] can be pre-
dicted from the phonetic context (Brown 1956: 252-5.3).
But complementary distribution in itself will not allow the grouping of two
phones into a single phoneme. The second criterion of phonetic similarity (usu-
ally judged in terms of place and manner of articulation) of the phones must
be satisfied as well. To use a well-worn linguistic example from English, one
notes that the segments [h] and [ g ] are in complementary distribution, but
that their lack of phonetic similarity prohibits their inclusion in a single pho-
neme.
The two domains given above are quite analogous. Each classifier is seen to
be in complementary distribution in that each occurs in environments in which
the others do not occur. This feature of distribution allows for the prediction
of the classifier when given the noun in the domains maintenance of vertical-
ity or harvest. Thus, the classifier /tuhE/ is semantically redundant in the
domain of verticality in that it carries no additional semantic load other than
that already indicated by the nouns with which i t occurs. The same is to be
said, of course, for /tuhC/, and for the classifiers /tul/, /Euy/, /kah/, /ka?/,
/kohk/, and /hohk/ in the harvest set.
The classifier-like lexemes herd, covey, school, in English, are similar. Given
the domain groups of animals each classifier can be predicted from the noun
with which it normally occurs. The classifiers which occur in the frames
- o f cattle, -of quail, and ; o f bass can be
predicted with essentially the same degree of accuracy as can some phones of
English vowels.
Note that such predictability due to environmental features is impossible
in cases of contrast. Given a noun /ma#/ corn dough in the domain of piles
of granular items, i t is impossible to predict the classifier /tol/ or buhs/.
On the contrary, /tol/ and /buhs/ focus on nonredundant semantic features
of /ma$/ in two distinct physical states, as do the shape classifiers in the
shape domain. Likewise, in English, given the domain shaped object and the
noun modeling clay, it would be impossible to predict the occurrence of a host
of shape attributes such as square, round, oblong, etc. I n this context,
the attributives specify some features of the object not indicated by the noun
itself.
92 Transcultural Studies in Cognition
We have yet to focus on the second criteria of determining the allosemic
states of the classifiers in the above domains; that is, their semantic similarity.
From our phonemic example from English, [IJ] and [h], although in comple-
mentary distribution, must be phonemically distinct due to their lack of
phonetic similarity-a similarity that can be determined by reference to lin-
guistics arbitrary phonetic grid arrived at by decades of cross-cultural sophis-
tication and familiarity with numerous languages. No such etic grid com-
prised of some arbitrary grouping of semantic components or critierial
attributes is available for our work with classifiers. Semantic similarity must
be determined by more indirect methods, i.e., informants responses. Our
principal means of determining what classifiers participate in the same seman-
tic domains have been given above, i.e., the judgments of informants when
instructed to classify the randomized list of classifiers into semantically re-
lated categories.
We are suggesting, then, that certain classifiers, while standing in com-
plementary distribution, are nevertheless semantically related, and might
be heuristically designated as allosemes of some more inclusive sememic unit.
The fact that such a unit may lack a single linguistic designator, such as a unit
lexeme, need not hinder such an analysis. The allosemes herd, flock, school,
etc., might be operational in members of the higher level seme groups of
animate beings, although there would indeed be exceptions to such a gloss.
The important feature to note is that such classifiers as herd, flock, etc., are
semantically redundant (in reference to the nouns with which they occur, given
the context of groups of animals) whereas such contrasting attributes as
square, round, etc., in our clay examples, are not. That such a semantic
principle holds identically in other languages as well is of no mean significance.
A third distribution of classifiers plus nouns is what, for lack of a better
term, we label mixed semantic domains. Such categories include classifiers ex-
hibiting contrastive as well as complementary distributional features. One
such domain is given in Table VIII.
This domain, like the harvest domain, is comprised of classifiers utilized
in the enumeration of actions performed on certain objects, The actions all in-
volve cleansing with water; the objects are items that can be washed. Some
criteria1 attributes appear as follows:
/Ha?/ enumeration of times washing of head hair
/sukh/ enumeration of times washing interior of certain objects, e.g., mouth, bowls,
plates, etc.
/pohk/ enumeration of times washing surface of certain objects (not including hair, cloth-
ing, or plants), but not interior
/sahk/ enumeration of times washing clothing
/sahp/ enumeration of times washing beanlike items, leafy vegetables, etc., by immersion
in water, accompanied by vigorous movement of item in water
I n this distribution, /3a?/, /sahk/, and /sahp/, due to their complemen-
tary distribution, are allosemes. What of their relationship to the contrasting
classifiers /sukh/ and /pohk/? The latter two classifiers function as a unit in
reference to the former three and may be called an alloseme indicating
cleansing of all items other than hair, clothing, or small, bean like items and
Descriplive Semantics of Tseltal Numeral Classiers 93
TABLE VIII. DISTRIBUTION OF NUMERAL CLASSIFIERS IN WASHING CATEGQRY
_______
Numeral Classifiers
Nouns
I .
I
I
cleansing domain I
I
I
I
Sa? suhk pohk sahk
I
hair /Sa?/ I
I
clothing /sahk/ I
I
I
vegetables, etc. /sahp/ 1
sahp
/#obi1 holj hair
/?eil/ mouth
/limete/ bottle
/pis/ small glass
/kib/ pitcher
/bohE/ gourd bowl
/pin/ earthen jar
/seC/ plate
/pulata/ plate
/kabal/ hand
/Euhte3/ plank
/sit/ face
/baketal/ body
/mega/ table
/manbana/ apple
/ku?il/ clothes
/Cob/ blanket
/pak/ material
/weSil/ pants
/#ek/ skirt
/bok/ leaf vegetable
/kahpe/ coffee beans
/?iiim/ corn
etc.
94 Transcdlural Studies in Cognition
cedures employed in eliciting 557 distinct forms. Preliminary subdivision of
the whole set into gross semantic taxonomic classes on the basis of systematic
use of questions was presented. Then, a series of semantic categories was de-
fined on the basis of similarities judged by native informants. Examples of
categories selected for discussion in this paper included those of shape, verti-
cal position, washing, harvesting, and others. These categories were sub-
jected to descriptive semantic analysis; and two basically distinct types were
isolated, primarily in terms of the distributional faetures of the classifiers.
Among the many problems left untouched in this paper are those of meta-
phorical extension and a phenomena of regular variation in meaning related to
presence or absence of infixed -h-. Examples of these two problems are pre-
sented below and conclude the paper.
I n regard to metaphorical extension, one would like to inquire into the
semantic extension of the attributes indicated by certain classifiers to objects
which normally would not be expected to exhibit such attributes. An example
would be /koht/, the classifier occurring primarily with animal (i.e., non-
human) beings, but which also may occur, for some speakers, with /mega/
table, /gila/ chair, and /ba#il ?iE/ small red chili pepper. The folk ety-
mology for the use of /koht/ in this metaphorical extension may serve as
clues to understanding the normative semantic range of the classifier. Two
informants indicate that tables and chairs occur with /koht/ due to the fact
that each has four legs, one feature common to numerous animals. The occur-
rence of chili pepper with the classifier is said to be due to the birdlike
resemblances of the chili when dried.
Another phenomena is the semantic relationship between two classifiers
derived from the same verb. One form occurs with the infixed -h- and the
other with no change in phonemic shape, e.g.,
/sut, suht/ < tv /-sut/
/sut/:
/suht/:
/Hut, Huht/<tv /-gut/
/&ut/:
/la#/:
/iuht/:
/lad, lah#/<tv /-la$/
/lah #/ ;
to tie knots in ropelike materials
knots in ropelike materials, normally each knot tied one on top of the
other
knots tied around some item in ropelike materials, but each knot
separated linearly along the material
to pinch with fingernails
pinches with fingernails of person, hands, etc.
pinches with fingernails, taking off small pieces of tortillas, meat, etc.
to stack it in neat fashion
each individual item in a neatly stacked pile, e.g., baskets, earthen jars,
pots, bricks, etc.
each stack of neatly stacked items, baskets, bricks, etc.
APPENDIX A
A linguistic sketch o j Tenejapa Taeltal numeral classijiers
Earlier works dealing with numeral classifiers in other Mayan languages
are few. Tozzer (1921), drawing heavily on lists collected by Zavala and Bel-
tran, cites several for Yucatec Maya. Harris (1947) includes thirteen classifiers
with minimal glosses for Chontal, but Keller (1955) offers a much more
Descriptive Semalzlics of TseZtaE Numeral Classijiers 95
comprehensive list (78 forms) and analysis for the same language. Castillo
(1961) published a list of 33 forms with approximate glosses for modern
Yucatec. For Tzeltal proper, Kaufman (1961), in discussing numerals in
general, lists 48 common classifiers from Tenejapa, Oxchuc, and Aguacatenan-
go. Sarles presents 51 for Oxchuc, and Slocum (1948) presents a much shorter
list for the same dialect.
Kellers (1955) article on Chontal is perhaps the niost relevant as far as
outlining the general semantic features of classifiers which appear, superficially
at least, to function similarly in TTzel.
The numeral classifiers of TTzel are a series of predominantly monosyl-
labic, bi-morphemic stems of the canonical shape CV(h)C. The majority of
classifiers are derived by an infixed -h- morpheme (which occurs between V1
and CI of the stem) from transitive and positional verb roots.4 A small number
of classifiers are derived from noun roots. Some transitive roots function as
classifiers with no change in phonemic shape. I n addition, there are a number
of classifiers whose derivation is uncertain. Infix (-h-) has two allomorphs,
/-h-/ and -p-. /-h-/ is present when Cz is voiceless, but not /?/ or /h/, When
CZ is voiced or is /?/or /h/, -@- occurs. Examples of CV(h)C classifiers derived
from transitive and positional verb roots are:
enumeration of objects in the class of nonhuman animals<positional verb/kot/*
e.g., /kotol/ on all fours
enumeration of near-circular rolls or coils of roplike materials < transitive verb
/ bt / to form into near circular coils
enumeration of actions of stretching elastic materials<transitive verb /be/ to
stretch it
Examples of CVC classifiers derived from transitive and positional verb roots
are :
enumeration of pricks, short stabswith pointed objects, as in bloodletting<transi-
tive verb /hul/ to prick, stab, let blood
enumeration of vertical, oblong mound shaped objects <positional verb /kuh/*
e.g., /kuhan/ to form into vertical, oblong, mound-shaped object
enumeration of pieces of ropelike material tied in a specified manner<transitive
verb /kam/ to tie it in specified manner
Examples of transitive roots which function as classifiers but which exhibit no
change in phonemic shape are:
enumeration of blows with open hand<transitive verb /taH/* e.g., /taitu/ to
strike with open hand
enumeration of snaps of fingers<transitive verb /tos/ to snap fingers
enumeration of action of drawing bow across strings of violin, stringed instru-
ments<transitive verb /kie/ to draw bow across string
/koht/
/Hoht/
/HahE/
/hul/
/kuh/
/kam/
/taH/
/tos/
/ki#/
Examples of monosyllabic classifiers derived from nouns are:
enumeration of potfuls of certain objects <noun /pin/ pot
enumeration of marks, written letters <noun /#lib/ letter
enumeration of pointed objects <noun /E%/ spiney object
/pin/
/fib/
/ E W /
Examples of polysyllabic classifiers are few. Those that have been observed
thus far are given below.
96 Transcultural Sludies in Cognifion
/Eutub/ enumeration of length of measure between index fingertip and tip of thumb in
maximum extended position <noun /Eutub/ index finger
/kaHel/ all at once (occurs only with numeral one) <?
/?ihka$/ enumeration of individual loads < transitive verb stem /?ihka$/ to carry loads
/yalel/ enumeration of momentary occurrences (occurs only with numeral one) < ?
The distribution of numeral classifiers is limited to quantifying expressions6
where they occur as second members of endocentric compounds (specific
nuineralfnumeral classifier). Generally, a noun whose referent is the object
or event being quantified accompanies the quantifying expression. E.g.,
two (persons of the class human) men /Ea?tul winik/
/ho%oht Eahan/
The noun may be absent, but the referent of the numeral expression is
clearly understood, either by contextual features or the highly specific nature
of the classifiers. E.g.,
five (pieces of ropelike material in circular coils) cords
/Eankoht/ four (beings of the class of nonhuman animals)
/hukpehE/ seven (braided pieces of ropelike material)
The compound, numeral+nunieral classifier, may be possessed (after der-
ivation by the nominal suffix -al) and may be considered morphologically
to be a noun. E.g.
/yoHkohtal/ the third animal-like being (y-third-person possessive prefi+?oH numeral
three fkoht numeral classifier with class of non-human animals+-a1 nominal
suffix)
/swaklihkal/ its six, hanging, ropelike objects (s-third person possessive prefixSwak nu-
meral seven+lihk numeral classifier with ropelike objects in natural state+
-a1 nominalsuffix)
As a substitution class, numeral classifiers may be divided into two general
A. Classifiers restricted in occurrence with the numerals (-hun-) one. (The numeral one
B. Classifiers which may occur with any numeral.
Classifiers of subclass A are of two types. One group, of which there are
three examples, is made up of quantifying classifiers which function analo-
gously to what Silverman calls numericals in Gilbertese. Numericals are used
to handle forms which occurred in the position of numeral classifiers but are
not so (1962:SS). They appear to mark off units, or groups of numerals, as
ZOs, 400s, 8,000s. The quantifying classifiers are /tab/ twenties, /bahk/
four hundreds, and /pik/ eight thousands. E.g.,
sub-classes:
has three allomorphs, /h-hu-hun/.
/htab peh/ 20 pesos (a group of 20)
/hbahk peHu/ 400 pesos (a group of 400)
/hpik peg,/ 8,000 pesos (a group of 8,000)
A second group of classifiers of subclass A function syntactically as time
or extent phrases. Three have been isolated and are /?ahk/ momentarily,
/yalel/ all at once, and /teb/ a little. E.g.,
/. . . te kalal ?ay h?ahk #in ya yutyut sbaik , . . /
. . . when momentarily they begin to argue among themselves . . .
Descriplive Semantics of Tzeltal Numeral Classijers
/. . . helaw naH lah bel ta hyalel . . . /
/. . . ?ay hteb . . . /
. . . it happened, or occurred, all at once . . .
. , . there is a little. . .
97
NOTES
f This paper represents the results of research conducted in the field during the months of
J une-December, 1962. Supporting funds have been made available by a National Science Foun-
dation Grant, #G22152, A. K. Romney, Principal Investigator. We are especially indebted to Itoy
DAndrade, Charles Frake, Terrence Kaufman, Paul Kay, Duane Metzger, and Gerald Williams
for their critical comments and readings of earlier drafts. This final version has gained much from
the kind help of Sydney Lamb, who agreed to serve as discussant of the paper at the Merida Con-
ference. Of his numerous suggestions, both written and verbal, we have incorporated many. The
responsibility for the paper in its present form, however, is solely ours.
a Tzeltal is a Mayan language spoken by approximately 45,000 to 70,000 persons in the cen-
tral highland region of the state of Chiapas, Mexico. The dialect of Tzeltal from which the data for
this paper are drawn is that spoken in the mlsnicipio (county or township) of Tenejapa. We offer
our thanks to our two principal informants, /h?alus mentes ton/ of the paraje (dispersed-hamlet-
l i e unit within the municipio) of /kul?aktik/ and /hpetul peres konte/ of the paraje of /PO-
kolum/.
* The term, numeral classifier, was used by Mary Haas (1942) in her discussion of their occur-
rence in Thai. Bloomfield (1933:237) notes the Hindu designation of numeratives (dvigu) for
forms with similar grammatical features. The term auxiliary noun has been used in reference t o
numeral classifiers in Chinese. (For the reference to Chinese, we are indebted to Sydney Lamb,
personal communication.)
4 Positional verb roots-as opposed to transitive verb roots-are recognized formally by two
principal derivational features: (1) they rarely occur as simple CVC inflected stems and (2) upon
derivation with the transitive stem-formative, -an, or the intransitive stem-formative, -ah, they
exhibit an infixed -h-. Transitive verb roots, on the other hand, occur most frequently as simple in-
flected stems and, upon derivation with the above mentioned suffixes, do not exhibit infixed -h-.
The term positional verb root has been used by Slocum (1948) and Berlin (1962, 1963) in agree-
ment with certain obvious semantic similarities of glosses referring broadly to position or shape.
6 There are two parallel series of quantifying expressions in Tenejapa Tzeltal, a general series
and a specific series. I n general numeral expressions, numeral roots are derived by the general
numeral derivational sufh -eb. Numeral classifiers occur only in spec& numeral expressions, and
the numeral roots in these expressions do not take the suffix eb. The general numeral series in
Tenejapa is coming into greater use by bilingual speakers with a knowledge of Spanish, the nu-
meral phrase structure of which is analogous. The exact conditions which predispose a speaker to use
specific numeral expressions in which a numeral classifier is obligatory and which specify some
semantic feature of the thing being counted are not as yet determined. There is indication, how-
ever, that monolingual speakers prefer the specific series over the general.
BIBLIOGRAPHY
BERLIN, BRENT
1962 An outline of Tenejapa Tzeltal morphology. Anthropology Research Papers. Stan-
ford University. Unpublished.
1963 Some semantic features of reduplication in a dialect of Tzeltal. International J ournal
of American Linguistics 29: 211-218.
1933 Language. New York, Henry Holt.
1956 Language and Categories, Appendix to Bruner, J . S., J . Goodnow, and G. A. Austin,
A Study of Thinking. New York, J ohn Wiley.
BLOOMFIELD, LEONARD
BROWN, ROGER
98 Transcultural St di es i n Cognition
CARROLL, J OHN B. and J OSEPH H. CASAGRANDE
1958 The function of language classification in behavior. In Readings in social psychol-
ogy, Maccoby, Eleanor, Theodore M. Newcomb, and Eugene L. Hartley, eds. New
York, Henry Holt.
Morfemas clasificadores del Maya Yucateco. In A William Cameron Townsend.
Mexico, D. F., Instituto Linguistico.
1957 Anthropology and linguistics. Georgetown University Monograph Series on Lan-
guage and Linguistics 9: 167-173.
HAAS, MARY R.
1942 The use of numeral classifiers in Thai. Language 18:201-205.
HARRIS, MARGARET
1947 Nombres clasificadores chontales. Anales del Instituto Nacional de Antropologla e
Historia 2:303-306.
1951 Some Navaho linguistic categories. Language 27: 11 1-120.
1962 The ethnography of speaking. In The Anthropological Society of Washington,
Anthropology and Human Behavior.
1961 Tzeltal numerals. Anthropology Research Papers. Unpublished manuscript. Stan-
ford University.
KELLER, KATHRYN
1955 The Chontal (Mayan) numeral system. International J ournal of American Linguis-
tics 21:258-275.
OSGOOD, CHARLES
1961 Comments on Professor Bousfields paper. I n Verbal Learning and Verbal Behavior,
Coter, ed. Proceedings of a conference sponsored by the Office of Naval Research
and New York University. New York, McGraw-Hill.
SILVERMAN, MARTIN
1962 Numeral classifiers in the Gilbertese language. Anthropology Tomorrow 8:41-58.
SLOCVM, MWANA
1948 Tzeltal (Mayan) noun and verb morphology. International J ournal of American
Linguistics 14: 77-86.
1951 Vocabdario Tsdtac. Mexico, D. F., Instituto Linguistico de Verano.
TOZZER, ALFRED M.
1921 A Maya grammar. Papers of the Peabody Museum of American Archaeology and
Ethnology. Harvard University. Vol. 9.
CASTILLO, MoxsPs ROMERO
1961
GOODENOUGH, WARD
HOIJ ER, HARRY
HYMES, DELL
KAULFMAN, TERRENCE