You are on page 1of 23

Society for American Archaeology

On Tool-Class Use Lives and the Formation of Archaeological Assemblages


Author(s): Michael J. Shott
Source: American Antiquity, Vol. 54, No. 1 (Jan., 1989), pp. 9-30
Published by: Society for American Archaeology
Stable URL: http://www.jstor.org/stable/281329 .
Accessed: 21/06/2014 21:46

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Society for American Archaeology is collaborating with JSTOR to digitize, preserve and extend access to
American Antiquity.

http://www.jstor.org

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
ON TOOL-CLASS USE LIVES AND THE FORMATION OF
ARCHAEOLOGICAL ASSEMBLAGES

Michael J. Shott

The importance of tool-class use lives in the formation of archaeological assemblages is established both in
theory and empirically, and accurate inference from the material record requires that use lives be measured or
estimated with confidence. Unfortunately, no method of measuring use lives directly from archaeological remains
has been developed. However, this important quantity may be related to elementary properties of tool classes such
as size and weight, properties which are themselves directly measurable. Ethnographic data on ceramic vessels,
in which use life is related to such properties, is described and analyzed. Using !Kung San and Ingalik data, a
similar analysis is performed for a variety of nonceramic tools and objects, although few of the stone-tool classes
common to archaeological assemblages are included. There, use life is related most strongly to manufacturing
cost and curation rate, an archaeological measure of which is proposed. Neither is an elementary property, but
they can be estimated accurately for many tool classes. These results are noteworthy but not definitive, and they
underscore the continued value of museum ethnographic collections and ethnoarchaeological research.

Archaeologists have long been aware of the complexities which intervene between prehistoric
cultural behavior and the material record of that behavior. Traditionally, however, their attention
has been engaged by the more urgent problems of time-space systematics-the attempt to place
formal variability in the correct cultural and chronological framework. Decades of research have
brought tremendous strides in this respect, although virtually all archaeologists recognize that further
progress is necessary.
Standing on this foundation, the trends of the past quarter century are familiar to all of us.
Increasingly, our attention has been fixed on more advanced problems of culture process. At the
same time, the discipline has begun to grapple with the highly complex and sometimes daunting
processes by which the operations of prehistoric cultural systems are registered in the material
record. Archaeology has begun to address these complexities in systematic fashion, complexities
which, through pioneering studies (Ammerman and Feldman 1974; Ascher 1961; Binford 1973,
1977; Schiffer 1975, 1976), have come to be known as assemblage-formation processes.
Despite these advances, certain somewhat simplistic practices remain common. For example,
archaeologists frequently assert an identity of form and function for tool classes recovered from
assemblages, and use the relative frequency of various classes as a measure of the relative frequency
of the activities represented by the functions assigned to the classes. In terms used by Ammerman
and Feldman (1974), form: function identities assume particular kinds of mapping relations between
tools and applications. Obviously, such identities are questionable, and many archaeologists have
assailed the practice (e.g., Dibble 1987; Hayden 1977; Odell 1981). But the second aspect of this
practice-equating archaeological with systemic frequency in Schiffer's (1976:26) terms-is the focus
of this paper. It too is questionable, but its effects upon the validity of our inferences has received
far less attention. This is unfortunate, because tool-class use life intervenes between the two domains
in such a manner that archaeological frequencies cannot be considered accurate reflections of systemic
frequencies.

Michael J. Shott, Program for Cultural Resource Assessment and Department of Anthropology, University of
Kentucky, Lexington, KY 40506

American Antiquity, 54(1), 1989, pp. 9-30.


Copyright ? 1989 by the Society for American Archaeology

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
10 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

THE USE-LIFE CONCEPT


"Use life" is a straightforward concept that denotes the length of service of tool classes in systemic
context (Schiffer 1976:60). It can be measured on a variety of scales, including number of strokes
for which a tool is employed (Gallagher 1977:411; Hayden 1979), and the number of specific uses
such as the firing of an arrow (Odell and Cowan 1986). However, use life probably is best expressed
as a function of time. If for no other reason, this treatment insures agreement between different
data sets and tool classes within them. Moreover, measurement as a continuous function of time
is most appropriate, because formation rates of archaeological assemblages can be calibrated only
in this way.
Individual tools and other objects within a class can and probably do have different use lives.
That is, each class is likely to be variable internally. Collectively, however, a single mean value can
be found to express the use life of most tool classes. Thus, the performance of individual specimens
within a tool class can be summarized by the class mean and variance. At this early stage of analysis,
concern is confined to mean values, although the important effect of class variance is acknowledged
(Schiffer 1975:68).
The importance of use life in assemblage formation is matched only by the paucity of relevant
data. Systematic data on class use lives is extremely rare in the archaeological and ethnographic
literature, with a few notable exceptions for ceramic assemblages (David 1972; DeBoer 1985; de la
Torre and Mudar 1982; Longacre 1985). Fewer data still have been compiled for lithics and other
tool classes (Clark and Kurashina 1981; Gallagher 1977; Hayden 1977; Janes 1983; Lee 1979;
Osgood 1940; Tindale 1965; Vincent 1985), making it difficult to find detailed, systematic data
suitable for analysis.

USE LIFE IN ASSEMBLAGE FORMATION: THEORETICAL PRINCIPLES


Ammerman and Feldman (1974) have provided the most comprehensive theoretical discussion
to date on the relation between technological organization and the archaeological assemblages
produced by cultural systems. Their model includes both activities and distinct tool classes with
functional significance. Importantly, it also considers the role of varying tool-class use lives (which
they term drop rates) and the "mapping relations" between tools and activities, by which they mean
the associations between tool classes and activities (Ammerman and Feldman 1974:610). They
also note the important role of activities in the generation of archaeological assemblages, but em-
phasize the equally important role played by the other factors: "There are other major variables
besides activities that enter into the 'making' of an assemblage, and these lead to a more complex
view of how 'function' can be expressed in the composition of an assemblage" (Ammerman and
Feldman 1974:610). They then construct a formal model in which the type and relative frequency
of activities are held constant while tool use lives and mapping relations are allowed to vary. Their
hypothetical exercise results in assemblages that vary in composition without comparable variability
in activities (Ammerman and Feldman 1974:614-616).
The general form of Ammerman and Feldman's model is shown in Table 1, where activities or
task applications are denoted by their relative frequenciesal-an (Shott 1986a:9). Tool classes TI-T,
are listed by row and are followed by their characteristic use lives dl-dn. Mapping relations between
tool classes and activities are given by the m,,-mnm coefficients shown as elements of a matrix
defined by tools T,-T, and activities a,-a,. The resulting abundance of each tool class Tn in
archaeological assemblages is denoted as U,,, and is given by the sum across all activities of the
mapping relations of the class with the activities, as in the following equation (Ammerman and
Feldman 1974:611):

= dj(m,a,+mn2a2 ... +mnan ). (1)

/tn to the sum


The relative frequency of tool class T, is denoted by r and is expressed as the ratio of
. iA. . ,.
of the

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 11

Table 1. A Model of TechnologicalOrganization.

Activities
Tool Tool Discard
Discard
Type Rate a, a2 a3 . a,
Tool: Activity Matrix
T, dl m,l mi2 ml3 min
T2 d2 m21 m22 m23 m2n
T3 d3 m31 m32 m33 . m3n
.. . . . . . . . . . . . . . . . .

...... .. . ... . .... . ..

Tn dn mn mn2 mn3 .. nn
Note: Model from Ammermanand Feldman(1974:Table1).

By this set of relations, a technology and its archaeological representation can be completely
specified. The relative frequencies of tools, r, are expressed as a function not only of activities, but
also of use lives and mapping relations. Even with the same set of activities and tool classes, two
different cultural systems can produce different archaeological assemblages if the tool-class use lives
and the mapping relations differ. The latter, therefore, can be viewed as organizational factors that
condition the form and content of the archaeological record. In this respect, it is irrelevant if one
views them as factors complicating the relation between behavior and material remains (Schiffer
1975) or as integral components of the operation of cultural systems (Binford 1981), nor is the
difference between reconstruction of behavior and inference of long-term cultural process important.
In either case, the importance of tool-class use lives to valid and accurate archaeological inference
is beyond dispute.
Ammerman and Feldman (1974) have shown that the archaeological frequency of a tool class is
the joint product of its use frequency and use life, and that the latter is an independent factor
conditioning the form and content of archaeological assemblages. Tool classes of identical use
frequency can have different archaeological frequencies exclusively as a result of differences in their
characteristic use lives. Tool-class use lives, therefore, figure prominently in assemblage-formation
processes. To accurately interpret the archaeological record, we must be able to control the effects
of this factor, which forces us to devise ways to measure or at least estimate tool-class use lives.
For purposes of this discussion, the integrity of distinct tool classes is assumed. In fact, however,
the identification and definition of tool classes or types is a topic of some debate in archaeology,
and tool-class integrity is compromised by factors such as recycling (Schiffer 1976:38) and curation
(Dibble 1987; Draper 1985). Subsequent discussion will show that the factors closely are related to
one another and to use life. For present purposes, however, their effects are not considered.

USE LIFE IN ASSEMBLAGE FORMATION: AN EXAMPLE


The theoretical importance of tool-class use life is established. Empirical support for its role in
archaeological assemblage formation can be mustered as well, as the following example demonstrates.
The tool assemblages used at any given time in most preindustrial societies are the product of
complex interactions between the activities to which the tools are applied and organizational factors
(Ammerman and Feldman 1974; Shott 1986b). The latter act to complicate the relation between
tool-use behavior and its archaeological signature. They do not eradicate it completely, but they
intervene between behavior and by-products in complex and nonintuitive ways. Thus, simple and
direct relations between the kind and frequencies of activities conducted at a site on the one hand,
and the tools discarded on the other, do not always obtain (Binford 1977:34-36). As a brief illus-
tration of this point, Yellen's (1977:Appendix B) data on the activities conducted at 16 !Kung San
camps and the archaeological record of those camps can be compared (Table 2).

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
12 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

Table 2. OccupationSpan, Activities, and Tool Assemblagesfrom !KungSan Camps.

OccupationSpan -
Man Woman Person ber of ber of NTool
Tool Assemblagesa
Assemblage
Camp Days Days Days Daysb Huntsc Killsd Size Types Collecte Hunte
1 8 13 16 29 6 9 21 5 20 1
2 9 18 18 36 9 2 0 0 0 0
3 11 28 26 54 28 8 15 1 15 0
4 20 48 50 98 - - 13 3 13 0
5 2 4 4 8 2 4 0 0 0 0
6 3 9 6 15 9 2 0 0 1 0
7 10 40 35 75 27 9 5 2 4 1
8 30 180 150 330 - - 3 2 1 2
9 2 6 4 10 5 6 4 1 4 0
10 12 84 72 156 - 2 14 2 12 2
11 3 21 18 39 - 11 2 10 1
12 3 15 15 30 11 3 8 1 8 0
13 5 25 25 50 16 1 31 2 31 0
14 7 56 49 115 - 6 18 4 16 2
15 1 5 5 10 2 2 3 1 3 0
16 6 36 36 72 17 4 7 2 7 0
Note:Data from Yellen (1977:AppendixB).
a
Tool-assemblagefiguresconfinedto tools involved in collectingand hunting,not maintenancetasks. Only
two items were excludedby this treatment.
b Personday figuresare for adults only.
c Indicatesthe man days in which huntingoccurred.
d Indicatesthe numberof animals obtained;it is unclearif all were returnedto camp.
e Indicatesthe numberof tools used in collectingand hunting,respectively.

Occupation span is indicated by woman and man days as well as overall adult person days.
Children were not counted. "Number of Hunts" indicates the number of days in which hunting
trips were mounted from camps. There is no information on the length of the trips nor the possibility
that more than one per day occurred. "Number of Kills" is self-explanatory, although it is unclear
if all animals were returned to camp. Finally, assemblage variables indicate the number of tools
discarded at the camps; they are the part of the camps' archaeological record of interest. Only tools
used in subsistence tasks, not maintenance, were counted, although this treatment caused the omis-
sion of only two items. Total assemblage size is found under "Size," which is the sum of the number
of tools used in plant collection and processing, and hunting, respectively. "Types" is a measure of
assemblage diversity.
Conventional inference of the behavior that produced the camps' archaeological assemblages
would assume that the assemblages simply and faithfully reflect the camps' occupation spans and
the nature and frequency of the tasks performed at them. Specifically, archaeological convention
would expect occupation span to be correlated with assemblage size and perhaps diversity, and the
number of hunting forays or at least the number of animals killed to be correlated with the number
of hunting tools discarded. Because the !Kung practice a relatively strict sexual division of labor,
one may even expect that the number of woman days of occupation would be correlated with the
number of plant collecting and processing tools discarded.
However, when the data are analyzed, none of these expectations is supported. In fact, no straight-
forward relation between the nature and frequency of activities and the number and kind of tools
left in the camps is found. Product-moment correlation coefficients calculated for occupation-span
and tool-assemblage variables failed to attain significant values in every instance. In sum, the
archaeological record of these camps simply or directly does not reflect the activities that transpired
there, and employing standard interpretive conventions would produce spurious conclusions about
the nature and range of the activities that produced the assemblages.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS
USE LIVES 13

Table 3. AggregatedData on Observedand ExpectedTool


Frequencies at !Kung San Camps.

Tool Use Frequencya Use Lifeb Observedc Expectedd


Adze 588 1,825 0 .32
Knife 1,176 1,825 1 .64
Bow 588 1,825 0 .32
Arrow 8,820 365 1 24.16
Quiver 588 1,825 3 .32
Probe 588 1,825 3 .32
Spear 588 3,650 0 .16
Diggingstick 547 180 7 3.04
Mortar 547 1,095 22 .50
Pestle 547 1,095 2 .50
a Calculated
by multiplyingthe numberof persondays of occupationat the
16 camps by the numberof tools of each type per person.
b Use life in days.
c From Yellen (1977:AppendixB).
d Obtainedby dividing use frequencyby use life.

Only when these data are aggregated do meaningful patterns emerge. Table 3 shows the observed
and expected frequencies of a series of tools used daily by the !Kung. It was compiled by combining
Yellen's (1977) data with Lee's (1979:Table 9.10) on the estimated frequencies and use lives of the
various tool types. With several notable exceptions, Table 3 shows substantial agreement between
observed and expected frequencies; major divergences occur in only two cases: arrows and mortars.
In the latter case, the divergence is largely attributable to the fact that 20 of the 22 discarded items
were abandoned in manufacture and not after ate normal period of use (Yellen 1977:220). There is
no obvious explanation for the divergence in the case of arrows. While the overall results indicate
the important role that organizational factors play in assemblage formation, one case of divergence
underscores the equally important contribution of such mundane factors as manufacturing errors
to archaeological assemblages.
Comparison of the tool-class use-life values and site occupation span indicates the reason for the
lack of agreement between tools and activities. Tool-class use lives consistently exceed site occupation
span by a considerable margin, such that the correlation between activities and site assemblages is
weak. In effect, tool discard occurs rarely relative to site occupation span, and very few tools will
be discarded at sites occupied for the periods typical of Yellen's sample. In statistical terms, tool
discard as the chief contributor to archaeological assemblages occurs as a stochastic process with
very low probability at any briefly occupied site. In similar situations, it would be difficult to infer
the kind and frequency of activities at a single site by the study of its archaeological remains. Regular,
predictable relations between the frequency of tools as they are used and their frequency in the
archaeological record only emerge at a regional level, and even there they are not direct.

USE LIFE AND ELEMENTARY PROPERTIES


As this example demonstrates, the importance of tool-class use lives in the formation of archae-
ological assemblages is beyond dispute. This forces us to confront the problem of estimating use
lives directly from the archaeological record, a task of no mean difficulty. Use life is not observable
or measurable directly as are quantities such as length, width, and thickness of tools, so at first
glance the problem may appear insurmountable. Fortunately, recent research has established a
relation between use life and what DeBoer (1985:353) terms "elementary properties" of tools, i.e.,
intrinsic physical properties directly observable on archaeological specimens.
DeBoer (1985:347-348) has noted three ways in which to infer the use life of tool classes. The
first involves observation and measurement in ethnographic context, followed by direct application
to similar archaeological cases. This approach has been very productive in the case of ceramic

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
14 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

Table 4. Age and Size of Fulani CeramicVessel Classes.

MedianAge Median Capacity


Vessel in Years Age x 2 in Liters
Bowl 2.7 5.4 1-2
Small cookingpot 2.7 5.4 3-5
Mediumcookingpot 2.5 5.0 7-10
Cooking/storage 10.2 20.4 15-20
Storage 12.5 25.0 40-60
Note: Data from David (1972:141).

assemblages; stone tools, however, often were abandoned very quickly following contact with in-
dustrial societies, and few ethnographers have observed them in use for any length of time. At
present, virtually no traditional societies commonly employ lithics, preferring metal for most tools
previously fashioned from stone. In sum, few opportunities to observe and measure the use life of
stone tools were available in the past and practically none exist today. Clearly, this avenue of
research holds only limited promise for archaeologists, which is not to deny completely the value
of such efforts. Careful, exhaustive reexamination of ethnographic literature is a possible approach,
and there may be much to be gleaned from these sources. Unfortunately, the effort is likely to be
laborious and time consuming, and the rewards limited in nature. We should by all means consult
ethnographic sources, but we probably will need to supplement them with uniquely archaeological
measures of the factors.
The second way to estimate use lives is through the archaeological record itself, principally from
assemblages formed by the sudden loss or abandonment of tools. This approach is based on the so-
called "Pompeii Premise" (Binford 1981), that treats such assemblages as representative of the
systemic frequency of the various tool classes. From this information, presumably, use life can be
calculated or at least rank ordered, since discrepancies between archaeological and systemic frequency
may be attributed to differences in use life between tool classes. That is, information on systemic
frequency furnished by the unique archaeological assemblages enables the archaeologist to attribute
differences in the relative frequencies of those classes in archaeological assemblages to the role of
use life (Pauketat 1987). As DeBoer (1985:348) notes, this is a problematic undertaking at best.
Leaving aside the fact that such assemblages are comparatively rare and for this reason would have
only limited value, their integrity is questionable. Most archaeological assemblages, even those from
very brief occupations, are the products of several formation processes, not a single one (Binford
1981; Stevenson 1985). It appears that this approach has limited value as well.
DeBoer (1985) has chosen a third approach, which links elementary properties of tool classes to
use life. In his analysis of Amazonian Shipibo-Conibo ethnographic ceramic assemblages, he has
shown that use life bears a significant relation to vessel weight and use frequency. Although the
latter is no more directly observable from archaeological remains than is use life, the use of weight
as a measure of use life is promising. These results demonstrate the potential value of vessel weight
in the analysis of ceramic assemblages. But ceramic specialists have not yet developed methods to
estimate whole-vessel weight from fragments, despite their ability to estimate other properties such
as rim diameter and vessel capacity. The development of such methods holds considerable promise
for the more accurate and informed interpretation of ceramic assemblages. Despite DeBoer's (1985:
355) own misgivings about the value of such elementary properties of tool classes, his results warrant
serious consideration to determine if they can be extended to other assemblages.
Fortunately, it is possible to attempt this evaluation because similar data on other ceramic
assemblages exist. Table 4 presents data compiled by David (1972:141) on the median age and the
capacity in liters of several West African Fulani vessel classes. Assuming that vessel age is distributed
normally, doubling the median value yields the approximate age at abandonment and hence the
use lives of the classes. Vessel capacity can be treated as a measure of size. Vessel capacity is taken
as the midpoint of the range given by David. Thus, measures of use life and size both are available.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 15

25- * storage

cooking/
20 - storage

LU

uL 15 -
Cl)
LU

10-

small
cooking
*pot * medium
-5
bowl bowl cooking
pot

1.0 2.0 3.0 4.0

CAPACITY(in liters)

Figure 1. Attained age and capacity for Fulani vessels (from David 1972).

Analysis shows that use life is associated significantly with the natural logarithm of this measure of
size (r = .87, p = .05, df= 3), a finding similar to DeBoer's. (The natural logarithm of rim diameter
was used, since the untransformed variable is not distributed normally. A similar transformation
was carried out in much of the analysis to follow to improve data conformity to the distribution
requirements of parametric statistics.) Low sample size, however, makes rank-correlation procedures
somewhat more suitable. Using Kendall's r, significant results are not obtained (r = .53, p = .48,
df = 3), suggesting that the correlation between use life and size may be weak. Note, furthermore,
that there is a conspicuous outlier in the relation between use life and size, as shown in Figure 1.
This is the class "medium cooking pot," which fails to conform to the general pattern. This outlier
may account for the difference in results observed between use of r and r as measures of correlation.
Such discrepancies, unfortunately, are an inescapable occurrence; at least some tool classes in most
material-culture inventories will violate the norm. DeBoer (1985:352-353), for example, found a
similar outlier in the Shipibo-Conibo assemblage. At present, there is no way to predict which
particular ones will fail to conform to general patterns. Nevertheless, use life and vessel size ap-
parently are related in the Fulani assemblage.
A second data set is furnished by de la Torre and Mudar (1982) from Becino, Negros, Republic
of the Philippines (Table 5). Again, only attained age rather than completed age or use life is given,
but it can be used as an approximate measure of the latter quantity. Vessel height and rim diameter
are listed (both in cm), and both can be treated as size measures. In this instance, analysis reveals
a correlation between attained age and vessel rim diameter (r = .66, p < .05, df = 11), but not
between age and vessel height (r = - .23, p > .30, df= 11). Figure 2 plots the relation between age
and the natural logarithm of vessel rim diameter in the Becino assemblage.
It is unclear why vessel height is not correlated with use life. Assuming that the multidimensional
variable "size" is the chief determinant of use life, there is no a priori reason why weight, as in the

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
16 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

0 0 0
14-

12-

10- . 0
LU
(0
-J
8 -
C)
CO
LU 0
*>

6-
* -

4-

2-

I I I I I I I I I I I I
1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0

RIMDIAMETER
(incentimeters)

Figure 2. Attained age and rim diameter for Becino vessels (from de la Torre and Mudar 1982).

Shipibo-Conibo case, or rim diameter as in this case, should be correlated more highly with use life
than is height. This especially is puzzling because of the abundant evidence that rim diameter and
height are themselves correlated to a high degree in many ceramic assemblages (Hally 1986; Shapiro
1984; Whallon 1969). However, Shapiro's meticulous analysis of Mississippian ceramic assemblages
from the Southeast demonstrates that different general classes of ceramics-jars and cazuelas in his
case-exhibit different relations between rim diameter and vessel height (Shapiro 1984:705, 708),
so that height does not necessarily correlate strongly with size.

Table 5. Attained Age and Size of Ceramic Vessels from Becino,


Negros, Republic of the Philippines.

Height Rim Diameter Attained Age


Vessel in cm in cm in Years
Water jar 47 6 1
Cooking pot 29 6 10
Water jar 27 43 15
Flower pot 16 21 15
Flower pot 16 19 15
Flower pot 13 15 15
Flower pot 18 21 15
Cooking pot 16 17 10
Rice pot 13 14 10
Cooking pot 11 10 7
Cooking pot 14 12 5
Cooking pot 10 11 4
Cooking pot 12 11 5
Note. Data from de la Torre and Mudar (1982:Table 13).

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 17

These results also parallel those of DeBoer (1985:353), who found a strong correlation between
vessel use life and the natural logarithm of vessel weight, also a size measure. However, the strong
correlation between age and rim diameter in this instance has an even more agreeable implication
for archaeologists, since vessel rim diameter can be estimated without difficulty directly from ar-
chaeological remains in most instances (Fitting and Halsey 1966; Hally 1986; Whallon 1969). It
thus provides a more readily available archaeological measure of vessel size than does weight.
DeBoer (1985:353) argues that weight and use life are correlated for three reasons, which may apply
to the Becino data as well: (1) Size reflects manufacturing cost such that perceived value varies
positively with size, and valu and use life are correlated; (2) size and portability are inversely
correlated, and as portability increases, vessels are more susceptible to breaking, which shortens
use life; and (3) the allometry of ceramic vessels is such that smaller ones are likelier to break.
DeBoer, in short, concludes that the relation between use life and size actually is composed of
several more specific relations. Detailed ethnographic study of additional ceramic assemblages may
establish the relevant relations. Whatever the case, sze as measured by weight, volume, or rim
diameter is an elementary property of ceramic vessels directly recoverable by archaeologists in most
cases, while the other properties listed by DeBoer are not.
To summarize the ceramic analysis, all assemblages exhibit significant associations of vessel size,
variously measured, and use life despite considerable differences in the kinds of data available and
the cultural context of each assemblage. Not all measures of vessel size, however, are correlated
significantly with use life. Nevertheless, these results
are encouraging, since they demonstrate that
archaeologically observable properties of ceramic vessels furnish information on their use lives.
Considering the important role of tool-class use lives in the formation of archaeological assemblages,
the findings can be exploited to improve inferences from archaeological remains. The archaeological
frequency of various vessel classes can be corrected using this information to produce reasonably
accurate estimates of the use or systemic frequency of the classes (e.g., David 1972:142).
It is worth noting that a similar set of relations has been detected in nature. Damuth (1982:426)
shows that lifespan, the biological equivalent of use life, allometrically is related to body weight in
large mammals. He is careful to emphasize that the relation, although strong, holds only for a single
trophic level (Damuth 1982:438). Thus, his analysis illustrates a parallel relation in nature, but
emphasizes that it may not be universal, applying instead to certain classes of objects which are
relatively similar in other respects.

EXTENDING THE ANALYSIS: THE DETERMINANTS OF


USE LIFE IN NONCERAMIC OBJECTS
Analysis of ceramic assemblages suggests that elementary properties of objects systematically are
related to their characteristic use lives. Clearly, a similar analysis of tools, preferably the stone tools
commonly encountered by archaeologists, is warranted. However, the same relation will not nec-
essarily obtain among such classes of objects, since their nature, their mode of manufacture, and
their context of use significantly differ from ceramics.
Stone tools, as well as other artifacts, can enter the archaeological record in a number of ways.
This is a truism hardly in need of elaboration. But it has a number of important implications that
may clarify the relation between use life and elementary properties of nonceramic objects. These
implications include systematic biases inherent in the relative frequency of different tool types in
the archaeological record, the form of the tools comprising those types, and the nature of organization
characterizing different technologies and cultural systems (Shott 1986a). For present purposes, how-
ever, the role of "discard processes" in determining tool-class use lives is the chief concern. Discard
processes (DP) are the processes by which tools enter the archaeological record. The following major
types of discard processes may be identified:
(1) Breakage in production. This involves the accidental fracture of tools before they reach finished
form. Breakage can occur either by accident unrelated to production, such as dropping the tool on
a hard surface or as an unintended result of the production process itself. This is especially likely
to occur in a production process involving numerous stages, such as biface production (Johnson

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
18 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

1981:43-64). This condition, trivially, is indicated by the presence of a break or breaks on a tool.
However, tools can be used after a fracture, often for purposes unrelated to their original function,
a process considered below. Distinguishing breaks that occur during production or use from those
occurring after use can be difficult at times; clearly, the origin of a fracture cannot be identified with
complete certainty. However, at North American archaeological sites many postoccupational frac-
tures probably occurred quite recently relative to the date of occupation, asas consequence of plowing
and related modem activities (Mallouf 1982; Roper 1976). Frequently, such fractures can be dis-
tinguished fairly clearly from old or original breaks, since they expose a fresher, unweathered surface
of the specimen. When a fracture is considered original, it still remains to distinguish unfinished
from finished specimens by such properties as the absence of marginal and basal abrasion, visible
use wear, and resharpening flakes. The steps in the production of most stone-tool classes are known,
and the absence of final production-stage attributes constitutes strong evidence for the identity of
unfinished implements.
(2) Abandonment during or after production. Tools may be abandoned, even if intact, if they
prove unsuitable for their intended purpose. The presence of flaws in the raw material, for example,
or remnant ridges that cannot be removed during thinning can produce this outcome. Such tools
are intact and may appear serviceable. Absence of evidence of use and maintenance, in the form
of edge wear and resharpening flakes, indicates that such items were abandoned without use. Con-
cached
ceivably, functional but unused cached tools may be considered abandoned tools. Caches, however,
usually are unmistakable in appearance; they consist of many items and are found in rather unique
archaeological contexts.
(3) Loss or breakage in use. Following completion and some use, many tools are broken in the
course of normal use. Tools discarded by this process obviously exhibit fracture, but also possess
attributes, such as extensive use wear and marginal resharpening, indicating that they experienced
at least some use and maintenance. This process shares with the first the problem of distinguishing
original from recent fractures. Loss is a straightforward concept, although its frequency can vary
systematically with a number of factors (Ebert 1979). Loss may appear to be a negligible discard
process, and is best viewed as a virtually random factor of little importance. In other instances,
however, it may vary with objective properties of items such as their size. In common with most
discard processes, few ethnographic data exist on the frequency with which loss occurs. Furthermore,
available data do not always agree, although most sources indicate that loss is an uncommon
occurrence. Binford (1977:33) records a loss rate of approximately one item per 186 episodes of
use and curation. Similarly, Yellen (1977:196) reports a single loss of a curated item in what must
have been at least several hundred such episodes. In contrast, Ebert (1979:63) indicates that loss is
a frequent occurrence and is an important contributor to archaeological assemblages. Yellen and
Ebert studied the same forager group at approximately the same time. The reason for the divergence
of their accounts is unclear.
(4) Recycling. This process refers to the use of a tool for different purposes and in a distinct use
context from its original, intended use (Schiffer 1976:38). Recycling is not to be confused with
versatility (Shott 1986b: 19), the use of tools in a variety of task applications. In the latter, a specimen
is maintained and used for purposes at least broadly consistent with its intended functional appli-
cations. If the specimen was designed for hafting, for instance, it remains hafted. Recycling, in
contrast, is the appropriation of a specimen for entirely different and unforeseen purposes. It usually
follows the completion of other processes, such as breakage or depletion, and is designed to modify
the tool, perhaps extending its utility in the process. The use of broken or depleted tools as bipolar
cores-sources of small flakes used expediently and discarded after short periods of use-is a good
example of recycling, as well as a common practice carried out by North American Paleoindian
groups (Goodyear 1982).
(5) Abandonment in use. Tools simply may be lost or abandoned while still serviceable. Aban-
donment occurs when a still-functional item is rejected deliberately and not retained for future use.
Conceptually, abandonment is the opposite of curation, and entails the deliberate discard of tools
possessing utility. This process probably occurs most often upon site abandonment, as size, weight,

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 19

or lack of immediate need cause people to leave serviceable items behind (Gould 1978; Hayden
1977, 1979). Heavy, bulky items, or quickly fashioned expedient tools are the most likely candidates
for abandonment, although it may occur at rates varying with a number of factors. In many cases,
abandonment is unrelated to intrinsic properties of tools. Instead, it may vary with prosaic factors
such as the completion of tasks in which expedient tools are used (Gould et al. 1971). In such
circumstances, abandonment may vary directly with the rate of task performance (Schiffer 1975).
(6) Depletion. Tools that escape production failures, breakage, and loss finally are depleted or
exhausted through numerous use and maintenance cycles. The outcome of this process is a depleted
tool. Such tools may be considered bereft of all their utility, and the outcome may be viewed as the
normal result of tool use, although relatively few tools may be deposited archaeologically by this
process depending on the nature of the technological system in which they are employed. Depletion
is exhibited by properties such as heavy resharpening and steep edge angles, although the latter is
not definitive. At least for unifacial tools, the purpose of resharpening is to maintain a constant
edge angle for whatever task application is desired (Hayden 1979; Nissen and Dittemore 1974),
and even heavily resharpened tools may retain fairly acute edge angles. Tools bearing recognizable
and distinct haft and functional segments should exhibit a high ratio of those segments. It is important
to stress, however, that different specimens of the same tool class, defined by haft-segment attributes
or other criteria, may appear different upon depletion. If their manner of use varied during their
use lives, their depleted form may differ. Some may retain little more than a stub of the original
functional segment, the only remnant left after repeated resharpening. Others, in contrast, may retain
much of one original margin and none of the other. In these cases, length and morphology of the
depleted specimens differ a great deal, but both are depleted nonetheless (Hoffman 1985; Miller
1980: Figure 53). These complications are important to recall in the course of this analysis.
Loss, breakage, and other discard processes may differ systematically between ceramic vessel and
other tool classes, producing in the latter effects upon use life independent of elementary properties.
However, loss and breakage probably systematically are related to size (Ebert 1979:63; Grimes and
Grimes 1985:42; Schiffer 1976:32-33). The use context of ceramic vessels is apt to be relatively
benign, since they chiefly are employed for storage and food processing of various kinds (Braun
1983; Hally 1986; Henrickson and McDonald 1983; Shapiro 1984) and are not transported exten-
sively. Tools, in contrast, are used in a wide variety of applications and numerous contexts, and
are transported a great deal. Finally, vessels are complete objects. Stone tools, at least retouched
ones, almost always are components of larger tools (Oswalt 1976). The distinction between use life
of the complete tool and its lithic component must be maintained carefully, especially in view of
the possibility that other components of tools are valued more highly and are curated heavily (Frison
1978:333; Hearne 1958:55-56, 181; Mason 1981:85; Nissen and Dittemore 1974; Shott 1986a:
199), and that lithic components may be designed for rapid use and abandonment or depletion in
order to increase the use life of the larger tool. Consider VanStone's (1985:12) observation regarding
the Naskapi that "arrow points were often loosely hafted so that the shaft would become detached
and could be recovered.... This was especially important in the barrens where wood could not be
easily obtained for additional shafts." The effects of these various factors upon tool-class use lives
and assemblage-formation processes are difficult to gauge, but they should be taken into account in
a systematic treatment of the issue.
In the view of the number of factors influencing use life of tools and the sometimes complex
relations they exhibit, it is not surprising that use lives apparently vary over a wide range. Stone
and other tool use lives vary from minutes (Gallagher 1977; Hayden 1979:34) to weeks (Tindale
1965) to years (Osgood 1940) to decades (Janes 1983:99). If use lives vary this widely, it is imperative
that we control their effects upon assemblages so that behavioral inferences can be made accurately.
Nevertheless, results of the ceramic analysis clearly justify a similar analysis of other tool classes
to assess the relation between use life and several archaeologically observable properties of tools.
As a first approximation, the following factors may be identified:
(1) Size. As with ceramics, use life may vary directly with object size, which can be expressed
by weight, length, or some product of the object's dimensions.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
20 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

Table 6. Use Life and Related Data on !Kung Tool Classes.

Maintenance
Manufacturing Cost in Total Cost in
Class Cost in Minutes Minutes/Day Minutes/Day Length in cm Use Life in Days
Digging stick 60 .5 3.5 120 180
Bow 900 2.0 2.5 90 1,825
Arrow 900 6.0 8.5 55 365
Poison 300 7.5 9.2 - 180
Spear 600 1.0 1.2 110 3,650
Probe 600 2.0 2.3 400 1,825
Club 300 .3 .6 75 1,095
Adze 600 2.0 2.3 60 1,825
Knife 1,200 4.0 4.6 26 1,825
Fire kit 300 20.0 20.2 - 1,825
Canteena 360 .0 .5 - 730
Canteenb 60 .0 .6 - 180
Kaross 900 .0 2.5 - 1,825
Bag 480 1.3 1.6 - 1,825
Yoke 15 .0 3.0 - 5
Quiver 600 .2 .5 - 1,825
Iron pot 900 1.0 1.5 - 1,825
Tortoise shell 300 .5 .9 - 730
Net 900 .3 .3 - 3,650
Snare 600 2.0 3.6 - 365
Mortar and pestle 600 1.0 1.5 - 1,095
Storage 450 1.0 3.4 - 180
Rack 180 .0 3.0 - 60
Skirt 280 .5 .6 - 1,825
Carriage 120 1.0 1.3 - 365
Sandals 360 .2 .4 - 1,825
Cloth 180 .2 .7 - 365
Chuana 480 .2 .5 - 730
Shirt 180 .2 .7 - 365
Shorts 180 .2 .7 - 365
Hut 900 5.0 15.0 - 90
Bedding 360 2.0 6.0 - 90
Note: Data from Lee (1979:Table 9.10).
a
Egg-shell canteen.
b
Stomach canteen.

(2) Manufacturing cost. Tool-class use lives can be viewed as directly proportional to the amount
of effort expended in production (Schiffer 1985:33). Manufacturing cost can be measured along a
variety of axes, including the rarity of the raw material, measured by distance from point of pro-
duction or use or difficulty in procurement, the degree of elaboration of the finished tool form,
measured perhaps by the number of facets on stone tools, and the number of stages in the manu-
facturing sequence, possibly inferred by replication experiments. It is in this manner, in fact, that
stone-tool replication studies are likely to be most informative.
Unfortunately, these predictions cannot be evaluated directly against ethnographic data since use-
life values for stone tools virtually are nonexistent there. The most direct approach for archaeological
purposes, therefore, must await the collection of suitable data if they exist. Until then, available but
not directly pertinent data must suffice. Fortunately, Lee (1979) and Osgood (1940) provide such
data.
Lee's (1979:Table 9.10) (see Table 6) data have been used previously. As Lee (1979:273) readily
admits, his use-life values are informed guesses and some may be inaccurate. The earlier analysis,
however, strongly suggested that most if not all of his estimates are reasonable. At any rate, the
acquisition of completely accurate use-life values requires directed, long-term research which is

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 21

Table 7. Use Life, Size, and ManufacturingData for Ingalik


Technology.

Manufactur-
Lengthin ing Cost in Use Life Use Fre-
Class Inches Days in Years quencya
Labretawl 4.0 2.0 4.0 1
Bone skin scraper 7.0 1.0 .5 2
Stone skin scraper 8.0 3.0 2.0 2
Bone wood scraper 5.5 1.0 6.0 1
Ulu 4.0 3.0 3.0 2
Bone knife 9.2 2.5 1.5 2
Stone bone cutter 4.8 2.5 1.0 2
Stone axe 14.0 5.0 1.0 2
Stone adze 6.5 3.5 .2 2
Wood wedge - .1 .2 2
Wood maul 16.0 1.0 .2 2
Diggingstick 11.0 .5 3.0 1
Lance 4.8 1.0 4.0 1
Note:Data from Osgood (1940).
a
1, single season; 2, year-round.

impractical under most circumstances. Lee is not an archaeologist, but he has provided an invaluable
service with these data. We can refuse to use them because they are not perfectly accurate, or we
can acknowledge their limitations and use them in a preliminary investigation of the determinants
of tool-class use lives. Unfortunately, the !Kung technology described by Lee includes virtually no
stone tools or parts of tools. Thus, the use lives he estimates for tools with metal components may
differ greatly from the preindustrial lithic equivalents. If the size, manufacturing cost, or maintenance
cost of the modem tools do not vary in corresponding degree, then the relations observed here will
not necessarily apply to stone-tool technologies.
Osgood's data (Table 7) are compiled from a study of the Alaskan Ingalik. Although highly detailed,
they are organized somewhat differently than Lee's data. Manufacturing cost (in days) almost in-
variably is given, but tool-class use lives are reported inconsistently. In some cases, specific values
are given; in others, qualitative information such as "a short time" or "a long time" are given. If
these observations were calibrated to the same scale, they could at least be used as an ordinal
variable. This is not the case, so Osgood's qualitative data cannot be used, reducing the number of
cases having acceptable use-life and ancillary data to 13 (Table 7). However, Osgood also furnishes
information on the use frequency of tool classes. Again, the data are somewhat imprecise, but in
most cases at least the season(s) of use is given or year-round use if reported. With data in this
form, it is impossible to present use frequency data on an ordinal scale corresponding to the number
of seasons per year (Table 7).

Use Life and Size


Size data are available for some !Kung tool classes. Lee (1979) records weight and various di-
mensions of tools, but the most consistently recorded variable is overall tool length. Recognizing
that length is an imperfect measure of size, it nevertheless can be used as an index for that quantity.
Table 6 lists size as length for selected !Kung tool classes. Exhaustive review of Lee or other sources
probably would reveal some additional size data.
Analysis shows that length and use life are not correlated to a significant degree in this assemblage
(T = .08, p = .90, df = 6), leading to the conclusion that tool size is not a suitable measure of use
life. Osgood (1940) furnishes tool-size data similar in nature to Lee's. Again, overall tool length is
given (Table 7). Analysis of these data produces similar results (T = -.14, p = .59, df = 11).
Experimental data also suggest that size and use life are at best weakly correlated among stone tools

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
22 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

(Odell and Cowan 1986:205). Tool size, apparently, is not a valuable predictor of use life in the
ethnographic assemblages studied here.
This finding is unfortunate, since size easily can be measured from archaeological specimens and
is, in effect, the ideal elementary property. Neither is the conclusion surprising, however, both
because the size measure employed is not ideal and because many factors in addition to overall size
have a bearing on tool use life. Unlike ceramics, whose context of use is comparatively benign,
metal and stone tools typically are used in a variety of taxing ways in different contexts. Frequency
of use and nature of the use application (Shott 1986a:9) may differ widely among tool classes of
similar size. It is dismaying but apparent, then, that tool size is not a suitable proxy measure of use
life in the !Kung San and Ingalik material cultures.

Use Life and Manufacturing Cost

Manufacturing cost and use life, however, are correlated significantly in the !Kung assemblage.
Analysis of the complete data set listed in Table 6 produces significant results for the relation
between manufacturing cost and the natural logarithm of use life (r = .49, p < .01, df= 31). Only
looking at portable tools in the data set-most of which roughly are equivalent to archaeological
lithic tool classes (cases 1-10 in Table 6)-manufacturing cost and the natural logarithm of use life
again yield a strong but not as significant correlation (r = .43, .10 > p > .05, df = 9), although
attained significance is lower in this instance.
The role of manufacturing cost in determining use life can be gauged by linear regression analysis.
Since results are significant, they are plotted in Figure 3. Note, however, that despite the significant
finding, the r2 value only attains .24, meaning that three-quarters of the variability in tool-class use
life is not explained by the effect of manufacturing cost. Nevertheless, these results demonstrate the
significant relation between use life and manufacturing cost in !Kung San material culture.
Analysis of Osgood's (1 940) Ingalik data, however, do not yield similar results. There, the observed
association of manufacturing cost and use life is weak as well as negative ( = -.06, p = .86, df =
12), indicating that manufacturing cost and use life are correlated inversely in this technology. The
conclusion is inescapable in this instance: manufacturing cost is not a strong determinant of use life
in the overall Ingalik technology.
Additional information, however, mitigates this finding. Osgood (1940) reports use frequency for
most tool classes, usually as season of use. As Table 7 shows, the classes are used either in a single
season or all year, though not all necessarily are used in the same season. These data can be used
as an ordinal scale variable for the use frequency of tool classes. DeBoer (1985:349) argued that use
frequency and use life should be correlated inversely, since exposure to the discard processes pre-
viously discussed increases with use frequency. His analysis found such a relation (DeBoer 1985:
352). Following his reasoning, the high-use-frequency class in the Ingalik technology should be
characterized by shorter use lives than the low-use-frequency class. Mean use life is 1.58 and 4.25
years, respectively (t = 5.10, p = .00), supporting this expectation.
Use frequency, of course, is not an elementary property of tools, so use life cannot directly be
estimated from it. It also is possible that a different relation between manufacturing cost and use
life exists for the two use-frequency classes. Results of analysis here do not resemble those char-
acterizing the entire assemblage. Both rank-correlation coefficients are positive, and attained sig-
nificance values are higher than those obtained for the complete assemblage (for high-use-frequency
tools, r = .45, p = .19, df= 7; for low-use-frequency tools, r = .50, p = .75, df= 2). For what it is
worth, attained significance is higher for tools used on a year-round basis (i.e., tools with high-use
frequency), suggesting that manufacturing cost and use life are correlated more strongly than they
are among less frequently used classes. Neither correlation is very strong, however.
Ingalik technology is significantly more diverse and complex than its !Kung counterpart. Reasons
for this include pronounced seasonality at high latitudes and the relatively great heterogeneity of
resource distributions there (Oswalt 1976; Shott 1986b; Torrence 1983). In view of this complexity,
it is no surprise that a single factor fails to account primarily for variability in tool-class use life.
This underscores the difficulty of analyzing complex technologies and the fact that relations between

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
TOOL-CLASS USE LIVES 23
Shott]

10,000 -

0 0

o * m

1,000 - 0 0

so 0

(0
-- 0 s 0 0

UJ
LL
l)
*0

100 -
0 0

I I I I I I
200 400 600 800 1,000 1,200
MANUFACTURING
COST (in minutes)

Figure 3. Use life and manufacturing cost for !Kung San material culture (from Lee 1979).

use life and elementary properties probably are strongest within subsets of technologies defined by
shared values of properties such as use frequency and morphology (Shapiro 1984).
Despite its limitations, the relation observed between manufacturing cost and use life in these
technologies is substantial. Determination of manufacturing costs may be regarded as at least a first-
order approximation of use life, although it is clear that additional factors are involved as well.
Although manufacturing cost is not exactly an elementary property of archaeological tool classes,
it can be estimated to some degree of accuracy in a variety of ways, as discussed previously.
This conclusion itself has an important implication. Technical changes that reduce manufacturing
costs, such as the development of blade technologies, may have effects on tool-class use lives. Given
the importance of use life in assemblage formation, such technical changes can have archaeological
effects even if the cultural system itself has not changed or has changed little. It is at least conceivable,
therefore, that some observed changes in archaeological assemblages may register technical changes
in tool production and not broader cultural change.

Use Life and Curation

To this juncture, the analysis has involved ceramics and ethnographically documented material
cultures with few lithic components. Although significant relations linking use life and elementary

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
24 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

properties have been found, the results are not directly relevant to archaeological lithic assemblages.
In fact, it is unlikely that size and use life are related directly among lithics, since the relation within
systemic context will not necessarily obtain in archaeological context. Most stone tools are discarded
at smaller than original size, since stone is a reductive medium. Therefore, the final size of specimens
may not bear a straightforward relation to use life, even if original size did. This makes it critical
to devise methods to reliably estimate original tool size from attributes of the discarded remnant.
Fortunately, some progress already has been made in this endeavor (e.g., Dibble 1987), but additional
research applied to artifact types of different character and cultural affinity is needed.
Fortunately, it may be possible to link completely or discarded size of tools to use life in another
way. Reduction and curation should vary directly, since more heavily curated stone tools should
be more extensively reduced. Curation, then, may offer a way to estimate tool-class use life, but the
concept requires revision to accomplish this.
Curation is a familiar concept to American archaeologists. Defined by Binford (1973), it refers
to the practice of maximizing the utility of tools by carrying them between successive settlements.
Since its introduction, the curation concept has become an organizing principle of many archaeo-
logical studies and has demonstrated its value many times. Nevertheless, it remains incompletely
formulated in theory and unnecessarily limited in practice. Typically, curation is treated as a static
property of particular technologies; that is, either the technology is considered curated or it is not.
In reality, however, and based on a full theoretical formulation, curation can be practiced to varying
degrees, and tool classes can be curated to greater or lesser extent. Curation, in short, is not a static
entity but a continuous variable characterized by case-specific rates.
Theoretically, curation can be defined as the realized utility of a tool, and curation rate can be
measured as the ratio of realized to potential utility. "Utility" is a highly abstract concept and a
thorough definition and discussion is beyond the scope of this paper. However, it has intuitive
meaning as the value obtained through the use of a tool, measured in time, energy, or some other
currency. For archaeological assemblages, a first approximation is the degree of reduction undergone
by items in a tool class, with which it should vary inversely. That is, as the tool is reduced pro-
gressively, its utility should decline in corresponding fashion. Obviously, however, utility may not
decline at the same rate as tool size, so that the latter need not measure remaining utility directly.
This, however, is a complicating factor left aside at present.
Use life may vary directly with curation, such that more heavily curated items are characterized
by higher use lives. Intuitively, it seems reasonable that the more heavily items are curated, the
longer they will last. And if use life cannot be measured directly from elementary properties of tools,
curation may be. The discard processes listed previously describe a trajectory of increasing curation,
and the material correlates of the processes have been considered at some length. They thus provide
a way to measure tool-class curation and, possibly, use life in archaeological assemblages.
Each DP can be viewed as a pathway (Hildebrand 1978) from systemic to archaeological context,
occurring at successive stages in the use lives of tools. At first glance, the DPs might be ordered in
sequence of increasing curation rate, from 1 to 6. However, this will not necessarily be the case.
DPs 1 and 2, "Breakage" and "Abandonment in Production," respectively, may be more likely to
occur among highly curated items. That is, a high frequency of occurrence of these processes may
betoken a tolerance for failure before completion of the manufacturing process if extensive curation
and, by extension, long use life can be anticipated for successfully produced specimens. In this case,
total manufacturing costs are small even with high failure rates, because the costs are averaged
against the great utility of finished specimens. Manufacturing costs, in effect, cannot be evaluated
in absolute terms, but rather against the expected utility of the product. Loss or breakage in use, in
contrast, effectively are random processes although the former probably varies with object size to
some degree. Therefore, they should cut across tool classes, occurring at roughly equal rates in all
regardless of their use life or curation rate. DP 4, "Recycling," may behave similarly. In contrast,
"Abandonment in Use," DP 5, should vary inversely with curation rate. By definition, high curation
involves the extraction of most of a tool's potential utiautility; those deliberately abandoned during
their period of service cannot attain high rates of curation. Thus, rates of abandonment in use

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 25

Table 8. Use Life, Manufacturing Cost, and Discard Among


!Kung Tools.

Discard Process
Use Life Manufacturing Discard Process
Class in Days Cost in Days 1+ 2 3 5 6

Egg-shell canteen 730 360 0 8 0 0


Stomach canteen 180 60 0 0 0 0
Digging stick 180 60 0 2 5 0
Bow 1,825 900 0 0 0 0
Arrow 365 900 0 1 0 0
Quiver 1,825 600 3 0 0 0
Spear 3,650 600 0 0 0 0
Probe 1,825 600 0 2 1 0
Club 1,095 300 0 0 0 0
Adze/axe 1,825 600 0 0 0 0
Knife 1,825 1,200 0 0 1 0
Mortar 1,095 300 20 0 2 0
Pestle 1,095 300 0 1 2 0
Iron pot 1,825 900 0 0 0 0
Spoon 1,095 600 0 0 0 0
Tortoise shell 730 300 0 0 0 0
Fire paddle 365 180 0 0 7 0

should be correlated inversely with curation rate. Finally, DP 6, "Depletion," should be correlated
strongly and positively with curation rate and use life, for obvious reasons.
Therefore, DPs 1, 2, and 6 may be associated with high curation rates, while DP 5 is symptomatic
of low rates. DP 3 is expected to exhibit no correlation with curation. This set of expectations makes
it possible to test the proposition that use life and curation rate are correlated, again using Lee's
(1979) and Yellen's (1977) data. In this case, however, Lee's use-life values are the dependent
variable and the rate or degree of curation as determined by Yellen's account is the independent
variable. Table 8 shows the frequency of the various discard processes by tool class. Recycling is
omitted since it is not reported by Yellen (1977). However, this may be an artifact of the short
observation span relative to most tool-class use lives, such that recycling occurs at rates far too low
to be observed in a brief period. Lee (1979:153) feels that tool recycling actually is a common
practice among the !Kung San, especially for bulky items like nutting stones. Neither class use lives
nor manufacturing cost were found to be rank correlated to a significant degree with any discard
process, a finding consistent with the argument presented above. Apparently, discard processes do
not describe a rank order of increasing curation rate. Therefore, analysis of variance is a more
suitable procedure for determining if use life and curation rate are correlated with data as they are
coded here.
Several assumptions were required in this exercise. When Yellen (1977:Appendix B) records a
tool at an abandoned camp site, it may be lost (DP 3), abandoned in use (DP 5), or depleted (DP
6). It is impossible on present evidence to distinguish in Yellen's data among those processes.
However, it is likely that deliberate abandonment is far more common than at least loss, and Yellen
(1977:196) explicitly reports only a single episode of the latter. Depletion is more problematic to
distinguish from abandonment on present evidence. The solution reached, which admittedly is
questionable, was to code each case of discard as abandonment in use, except for the single explicit
report of loss. Depletion, therefore, was not confirmed for any discarded tool. Several classes,
however, lacked any discarded specimens at any of the camps, although it is assumed that they
were used in the camps. Lee (1979), at any rate, reports that all tool classes he described and listed
usually were part of the standard material inventory of the !Kung. The failure to discard specimens
in those classes, therefore, is attributed here to a high rate of curation. Accordingly, it is assumed
that discard for those classes was characterized primarily by depletion. When a tool is listed as

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
26 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

Table 9. Mean Use Life by DiscardProcess.

DiscardProcess Number Mean Use Life


1 and 2 2 1,460.0
3 3 973.3
5 4 912.0
6 8 1,528.1

broken, it is assumed to be broken in use rather than in production, unless otherwise specified.
Finally, several instances of abandonment before use are reported, but again it is impossible to
distinguish between two processes, 1 and 2 in this case. Therefore, DPs 1-6 are collapsed for analysis
into four classes: I and 2, 3, 5, and 6. Using the foregoing assumptions, each tool class then was
associated with the process through which it most often was discarded.
Table 9 lists mean use-life values by discard process. Inspection of the table provides initial
support for the predictions that DPs 1, 2, and 6 are associated with high-use-life tool classes. Analysis
of variance also yields results supporting expectations. When tool classes identified with DPs 1, 2,
and 6 are pooled against those characterized by DPs 3 and 5, clear but not statistically significant
differences in use life are found (F = 2.09, p = .17, df = 2, 14). The mean use life for the more
highly curated classes is 1,514.5 days, and for the less curated classes 912.1 days. Next, classes
represented by DP 3, loss or breakage in use, were removed, since it was argued previously that the
process should not be strongly associated either positively or negatively with curation rate. Analysis
then was confined to classes characterized by pooled DPs 1, 2, and 6 versus those dominated by DP
5. Mean values here are 1,514.5 and 866.2, respectively. Results were not substantially different
with this treatment (t = 1.23, p = .24, df= 12).
Nevertheless, this brief analysis provides at least modest support for the equation of curation rate
and use life. Clearly, this proposition deserves more thorough study, possible only when detailed
ethnoarchaeological data on the use lives of tool classes and their characteristic discard processes
are collected. Thus, there is an urgent need for such data before additional progress on assemblage
formation processes can be made.

MEASURING CURATION IN THE ARCHAEOLOGICAL RECORD


Earlier, curation was operationalized for archaeological lithic assemblages as the degree of re-
duction undergone by tools. The following case study is a brief example of how curation rate can
be measured in such assemblages. Grimes and Grimes (1985) made a highly detailed study of a
single Paleoindian tool class, the flake shaver, from the Bull Brook I site in Massachusetts. For
present purposes, their most relevant findings concern the size of original and discarded specimens.
The authors define several subclasses of flake shavers; as a matter of convenience, one subclass-
regular, symmetric intact or conjoined tools (Grimes and Grimes 1985:36, 38)-is considered here,
although the treatment could be applied just as easily to others.
To measure curation rate, potential and realized utility must be identified and measured. For-
tunately, Grimes and Grimes (1985:40-41) indicate that the functional section of flake shavers is
confined to their distal margin, and that little lateral attrition occurs during use and reduction. Thus,
once the nonfunctional haft segment is subtracted, overall tool length can be used as a measure of
utility. Furthermore, Grimes and Grimes (1985:43) have identified what they consider is the min-
imum functional length of the tools-approximately 2 cm-which can be viewed as the point at
which potential utility is fully realized. This corresponds to the point of depletion, or the end point
in the use life of specimens in the tool class. They also estimate the mean original length of specimens
as approximately 7.45 cm (Grimes and Grimes 1985:44). Subtracting the minimum from the
maximum length yields 5.45 cm as a measure of potential utility of flake shavers. (Both minimum
and maximum figures are estimated from analysis of regular symmetric and other flake-shaver

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 27

subclasses, but inspection of Grimes and Grimes [1985:Table 3] reveals that they apply to this
subclass separately as well as to the larger aggregated class.)
Curation was defined earlier as the ratio of realized to potential utility. The former quantity is
found in this case by subtracting mean attained length of specimens - 3.57 cm (Grimes and Grimes
1985:Table 3)-from the maximum length of 7.45 cm, giving a result of 3.88 cm. The ratio of this
value to potential utility is 3.88:5.45, or .71. Following the reasoning set forth above, regular
symmetric flake shavers at Bull Brook I were curated at a rate of 71 percent of potential utility.
Similar analyses of complete assemblages could produce a set of curation rates d1-dn. Inserting them
into Ammerman and Feldman's (1974) model and solving for activities a,-a, would yield their
relative frequencies and, therefore, the behavioral correlates of the assemblage.
These calculations and the reasoning on which they were based do not establish that curation
rate is the only factor responsible for characterizing this assemblage. But extraneous sources of
variability in tool metrics are reduced by treatment of a single, relatively homogeneous class whose
degree of reduction can be gauged directly from a single variable, and Grimes and Grimes' meticulous
study, providing the detailed data necessary for the analysis, lends credence to the postulated
measures of potential and realized utility. Accepting the validity of the conclusions, the implications
of this exercise are worth considering. If similar curation rates can e alculated for other tool
cn be
classes, their raw frequencies in archaeological assemblages can be corrected for different curation
rates and use lives. For instance, imagine two tool classes, A and B, that occur in equal numbers
in an assemblage. Assuming equal systemic frequencies, the traditional conclusion would be that
the classes were employed atthe same rate. If, however, A's curation rate was 50 percent and B's
100 percent, then the use rates must be corrected. To occur in equal archaeological frequencies, B
must have been used at twice the rate of A.
This example by no means exhausts the possibilities for inferring tool-class use life from the
archaeological record. Elementary properties are the prred measure of this quantity, but other,
less-direct methods also may prove useful. For instance, caching may be considered evidence for
high curation rates and, by extension, long use life. Cached tool classes are the nearest preindustrial
equivalent of mass-produced i nvolved in caching itself may be added to manu-
facturing costs and has the effect of increasing those costs. As we have seen, manufacturing costs
and use life are correlated positively at least in some cases. In effect, tool classes with high use lives
may be more likely to be cached than are others. In eastern North America, for instance, Early
Woodland Meadowood bifaces often are found in caches. Although detailed and systematic inves-
tigation of their curation rates has not yet been attempted, discarded Meadowood bifaces typically
assume a wide variety of forms, most of which indicate extensive reduction and high curation rates.
Based on the foregoing argument, this tool class may be characterized by high use life. Tool caches
are surprisingly common (Seeman 1984; Wilson 1899), and their potential value to organizational
studies remains largely untapped.

CONCLUSION
Use life is related significantly to elementary properties of material culture, including tool size
and curation rate, but the relation is not constant across all tool classes. In fact, they probably are
best applied only within relatively homogeneous classes (Damuth 1982; Shapiro 1984). For example,
chipped stone and ground-stone tools of similar size may differ considerably in use life, probably
as a function of differences in manufacturing cost. Among chipped-stone tools, however, size and
use life may covary significantly. But even within such broad classes, the form or strength of the
relation may differ among subclasses such as unifaces and bifaces.
Despite these complications, the results of this initial analysis are encouraging. With further testing
and development of appropriate theory relating use life to elementary physical properties of tool
classes, archaeologists will be able to control the effects of a troublesome independent factor-use
life-on the formation of archaeological assemblages. The task is apt to be protracted and difficult,
but the prospects-more accurate and reliable inferences from material remains-are too inviting
to ignore.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
28 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

REFERENCES CITED

Ammerman, A., and M. Feldman


1974 On the 'Making' of an Assemblage of Stone Tools. American Antiquity 39:610-616.
Ascher, R.
1961 Analogy in Archaeological Interpretation. Southwestern Journal of Anthropology 17:317-325.
Binford, L. R.
1973 Interassemblage Variability-The Mousterian and the 'Functional' Argument. In The Explanation of
Culture Change: Models in Prehistory, edited by C. Renfrew, pp. 227-253. Duckworth, London.
1977 Forty-Seven Trips: A Case Study in the Character of Archaeological Formation Processes. In Stone
Tools as Cultural Markers: Change, Evolution and Complexity, edited by R. Wright, pp. 24-36. Humanities
Press, Atlantic Highlands, New Jersey.
1981 Behavioral Archaeology and the 'Pompeii Premise.' Journal of Anthropological Research 37:195-208.
Braun, D. P.
1983 Pots as Tools. In Archaeological Hammers and Theories, edited by J. Moore and A. Keene, pp. 107-
134. Academic, New York.
Clark, J. D., and H. Kurashina
1981 A Study of the Work of a Modern Tanner in Ethiopia and Its Relevance for Archaeological Interpre-
tation. In Modern Material Culture: The Archaeology of Us, edited by R. Gould and M. B. Schiffer, pp.
303-321. Academic Press, New York.
Damuth, J.
1982 Analysis of the Preservation of Community Structure in Assemblages of Fossil Mammals. Paleobiology
8:434-446.
David, N.
1972 On the Life Span of Pottery, Type Frequencies, and Archaeological Inference. American Antiquity 37:
141-142.
DeBoer, W. R.
1985 Pots and Pans Do Not Speak, Nor Do They Lie: The Case for Occasional Reductionism. In Decoding
Prehistoric Ceramics, edited by B. A. Nelson, pp. 347-357. Southern Illinois University Press, Carbondale.
de la Torre, A., and K. Mudar
1982 The Becino Site: An Exercise in Ethnoarchaeology. In Houses Built on Scattered Poles: Prehistory and
Ecology in Negros Oriental, Philippines, edited by K. Hutterer and W. Macdonald, pp. 177-146. San Carlos
University Press, Cebu City, Philippines.
Dibble, H. L.
1987 The Interpretation of Middle Paleolithic Scraper Morphology. American Antiquity 52:109-117.
Draper, N.
1985 Back to the Drawing Board: A Simplified Approach to Assemblage Variability in the Early Palaeolithic.
World Archaeology 17:3-19.
Ebert, J. I.
1979 An Ethnoarchaeological Approach to Reassessing the Meaning of Variability in Stone Tool Assem-
blages. In Ethnoarchaeology: Implications of Ethnography for Archaeology, edited by C. Kramer, pp. 59-
74. Columbia University Press, New York.
Fitting, J., and J. Halsey.
1966 Rim Diameter and Vessel Size in Wayne Ware Vessels. Wisconsin Archaeologist 47:208-211.
Frison, G. C.
1978 Prehistoric Hunters of the High Plains. Academic Press, New York.
Gallagher, J. P.
1977 Contemporary Stone Tools in Ethiopia: Implications for Archaeology. Journal of Field Archaeology
4:407-414.
Goodyear, A. C.
1982 Toolkit Entropy and Bipolar Reduction: A Study of Interassemblage Lithic Variability among Paleo-
Indian Sites in the Northeastern United States. Ms. on file, Institute of Archaeology and Anthropology,
University of South Carolina, Columbia.
Gould, R. A.
1978 The Anthropology of Residues. American Anthropologist 80:815-835.
Gould, R. A., D. Koster, and H. Sontz
1971 The Lithic Assemblage of the Western Desert Aborigines. American Antiquity 36:149-169.
Grimes, J., and B. Grimes
1985 Flakeshavers: Morphometric, Functional and Life-Cycle Analyses of a Paleoindian Unifacial Tool
Class. Archaeology of Eastern North America 13:35-57.
Hally, D. J.
1986 The Identification of Vessel Function: A Case Study from Northwest Georgia. American Antiquity 51:
267-295.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
Shott] TOOL-CLASS USE LIVES 29

Hayden, B.
1977 Stone Tool Functions in the Western Desert. In Stone Tools as Cultural Markers: Change, Evolution
and Complexity, edited by R. Wright, pp. 178-188. Humanities Press, Atlantic Highlands, New Jersey.
1979 Paleolithic Reflections: Lithic Technology and Ethnographic Excavation among Australian Aborigines.
Humanities Press, Atlantic Highlands, New Jersey.
Hearne, S.
1958 A Journey from Prince of Wales Fort in Hudson's Bay to the Northern Ocean in the Years 1769, 1770,
1771 and 1772. Edited by R. Glover. Macmillan, Toronto.
Henrickson, E., and M. McDonald
1983 Ceramic Form and Function: An Ethnographic Search and an Archaeological Application. American
Anthropologist 85:630-643.
Hildebrand, J. A.
1978 Pathways Revisited: A Quantitative Model of Discard. American Antiquity 43:274-279.
Hoffman, C. M.
1985 Projectile Point Maintenance and Typology: Assessment with Factor Analysis and Canonical Corre-
lation. In For Concordance in Archaeological Analysis: Bridging Data Structure, Quantitative Technique,
and Theory, edited by C. Carr, pp. 566-612. Westport Press, Kansas City.
Janes, R. R.
1983 Archaeological Ethnography among Mackenzie Basin Dene, Canada. Technical Paper No. 28. Arctic
Institute of North America, University of Calgary, Alberta.
Johnson, J. K.
1981 Lithic Procurement and Utilization Trajectories: Analysis, Yellow Creek Nuclear Power Plant Site,
Tishomingo County, Mississippi. Archaeological Papers No. 1. Center for Archaeological Research, Uni-
versity of Mississippi, University.
Lee, R. B.
1979 The IKungSan: Men, Women and Work in a Foraging Society. Cambridge University Press, Cambridge,
England.
Longacre, W. A.
1985 Pottery Use-Life among the Kalinga, Northern Luzon, the Philippines. In Decoding Prehistoric Ce-
ramics, edited by B. Nelson, pp. 334-346. Southern Illinois University, Carbondale.
Mallouf, R. J.
1982 An Analysis of Plow-Damaged Chert Artifacts: The Brookeen Creek Cache (41HI86), Hill County,
Texas. Journal of Field Archaeology 9:79-98.
Mason, R. J.
1981 Great Lakes Archaeology. Academic Press, New York.
Miller, P. A.
1980 Archaic Lithics from the Coffey Site. In Archaic Prehistory on the Prairie-Plains Border, edited by A.
E. Johnson, pp. 107-111. Publications in Anthropology No. 12. University of Kansas, Lawrence.
Nissen, K., and M. Dittemore
1974 Ethnographic Data and Wear Pattern Analysis: A Study of Socketed Eskimo Scrapers. Tebiwa 17:
67-88.
Odell, G. H.
1981 The Morphological Express at Function Junction: Searching for Meaning in Lithic Tool Types. Journal
of Anthropological Research 37:319-342.
Odell, G. H., and F. Cowan
1986 Experiments with Spears and Arrows on Animal Targets. Journal of Field Archaeology 13:195-212.
Osgood, C.
1940 Ingalik Material Culture. Publications in Anthropology No. 22. Yale University, New Haven.
Oswalt, W.
1976 An Anthropological Analysis of Food-Getting Technology. Wiley and Sons, New York.
Pauketat, T. R.
1987 A Functional Consideration of a Mississippian Vessel Assemblage. Southeastern Archaeology 6:1-15.
Roper, D. C.
1976 Lateral Displacement of Artifacts Due to Plowing. American Antiquity 41:372-375.
Schiffer, M. B.
1975 The Effects of Occupation Span on Site Content. In The Cache River Project, edited by M. B. Schiffer
and J. House, pp. 265-269. Arkansas Archaeological Survey, Fayetteville.
1976 Behavioral Archaeology. Academic Press, New York.
1985 Is There a 'Pompeii Premise' in Archaeology? Journal of Anthropological Research 41:18-41.
Seeman, M. F.
1984 Craft Specialization and Tool Kit Structure: A Systemic Perspective on the Midcontinental Flint
Knapper. In Lithic Resource Procurement: Proceedings from the Second Conference on Prehistoric Chert
Exploitation, edited by S. Vehik, pp. 7-36. Center for Archaeological Investigations, Carbondale.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions
30 AMERICANANTIQUITY [Vol. 54, No. 1, 1989

Shapiro, G.
1984 Ceramic Vessels, Site Permanence, and Group Size: A Mississippian Example. American Antiquity 49:
696-712.
Shott, M. J.
1986a Settlement Mobility and Technological Organization among Great Lakes Paleo-Indian Foragers. Ph.D.
dissertation, University of Michigan. University Microfilms, Ann Arbor.
1986b Settlement Mobility and Technological Organization: An Ethnographic Examination. Journal ofAn-
thropological Research 42:15-51.
Stevenson, M. G.
1985 The Formation of Artifact Assemblages at Workshop/Habitation Sites: Models from Peace Point in
Northern Alberta. American Antiquity 50:63-81.
Tindale, N. B.
1965 Stone Implement Making among the Nakako, Ngadadjara and Pitjandjara of the Great Western Desert.
Records of the South Australian Museum 15:131-164.
Torrence, R.
1983 Time Budgeting and Hunter-Gatherer Technology. In Hunter-Gatherer Economy in Prehistory. A
European Perspective, edited by G. Bailey, pp. 11-22. Cambridge University Press, Cambridge, England.
VanStone, J. W.
1985 Material Culture ofthe Davis Inlet and Barren Ground Naskapi. The William Duncan Strong Collection.
Fieldiana Anthropology No. 7. Field Museum of Natural History, Chicago.
Vincent, A. S.
1985 Plant Foods in Savanna Environments: A Preliminary Report of Tubers Eaten by the Hadza of
Northern Tanzania. World Archaeology 17:131-148.
Whallon, R.
1969 Rim Diameter, Vessel Volume, and Economic Prehistory. Michigan Academician 2:89-98.
Wilson, T.
1899 Arrowpoints, Spearheads, and Knives of Prehistoric Times. In Report of the United States National
Museum for the Year Ending June 30, 1897, pp. 811-988. Washington, D.C.
Yellen, J.
1977 Archaeological Approaches to the Present. Models for Reconstructing the Past. Academic Press, New
York.

This content downloaded from 193.104.110.144 on Sat, 21 Jun 2014 21:46:59 PM


All use subject to JSTOR Terms and Conditions

You might also like