Sie sind auf Seite 1von 18

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 157

EVALUATING THE DISTRIBUTIONAL APPROACH TO INFERRING MARKETPLACE EXCHANGE: A TEST CASE FROM THE MEXICAN GULF LOWLANDS
Christopher P. Garraty

Over a decade ago Kenneth Hirth (1998, 2000) developed a distributional approach for archaeologically inferring the existence of marketplace exchange based on analyses of domestic artifact collections. Domestic collections, he reasoned, will be relatively homogeneous in areas where most or all households rely on marketplace exchange to acquire domestic provisions. The present study evaluates Hirths distributional approach using a statistical measure of diversity (heterogeneity) to quantify variability among domestic collection units over a large area. The data for this study come from the Middle Postclassic lower Blanco region of Veracruz (A.D. 1200A.D. 1350), an unknown context of marketplace exchange. A comparison of diversity scores calculated on surface sherd collections from the lower Blanco region with scores from Late Postclassic Teotihuacan (A.D. 1350A.D. 1520)a known context of marketplace exchangesuggests the existence of a marketplace exchange system in the lower Blanco region, likely centered at the town of El Sauce. In addition, changes in intercollection diversity (sherds) and obsidian concentrations with increasing distance from the center suggest El Sauces market service area encompassed a radius of approximately six to nine kilometers. Casi hace una dcada Kenneth Hirth (1998, 2000) desarroll su mtodo distribucional para deducir arqueolgicamente la existencia del intercambio de mercado basada en anlisis de las ensambladuras domsticas de artefactos. Las ensambladuras domsticas son relativamentes homogneos en lugares donde la mayora de las unidades domsticas dependen del intercambio de mercado para obtener las provisiones domsticas. Este estudio evala el mtodo distribucional usando una tcnica estadstica de diversidad para cuanticar variabilidad entre las ensambladuras domsticas a travs de un rea grande. Especcamente, se trata de variabilidad entre las ensambladuras de la region del bajo Ro Blanco en Veracruz, un contexto desconocido del intercambio de mercado, durante el perodo posclsico medio (12001350 d. C.). Una comparacin de las estadsticas de diversidad calculadas en las mismas ensambladuras cermicas con las ensambladuras cermicas de Teotihuacn durante el perodo posclsico tarde (13501520 d. C.), un contexto conocido del intercambio de mercado, sugiere la existencia de intercambio de mercado en la regin del bajo Ro Blanco, centrada probablemente en el centro de El Sauce. Adems, patrones de cambiar en las estadsticas de diversidad y en las concentraciones de obsidiana con el aumento de la distancia del centro sugieren que la rea de servicio del mercado abarcara un radio de seis a nueve kilmetros.

n this article, I evaluate Kenneth Hirths (1998, 2000:182194) distributional approach to archaeologically inferring marketplace exchange and address some of the problems with its application. 1 Hirth identies marketplace exchange in archaeological contexts based on artifact compositions from domestic collections. Marketplace exchange, he argues, generates relatively uniform domestic assemblages, because all or most households within a given market service area provision themselves from a central source. An important signature of marketplace exchange is that households provision themselves indepen-

dently of one another and without regard to broader social and political relationships. ... The result is an increase in the homogeneity of material culture assemblages between households of different social rank (Hirth 1998:456, emphasis added). All households serviced by a single marketplace or market system have equal access to the same assortment of commodities (barring the existence of exchange barriers, such as sumptuary laws). Hirth devised the distributional approach to infer the presence of a marketplace exchange system at the Epiclassic center of Xochicalco in highland central Mexico (ca. A.D. 700A.D. 900) based on

Christopher P. Garraty School of Human Evolution and Social Change, Arizona State University, Tempe, AZ 852872402 (cpgarraty@asu.edu) Latin American Antiquity 20(1), 2009, pp. 157174 Copyright 2009 by the Society for American Archaeology
157

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 158

158

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

analyses of variability among domestic artifact collections. Hirths distributional approach holds a great deal of promise for better understanding the evolution of premodern economies and market systems, but a number of commentators have been critical of his approach. Smith (1999) argues that archaeologists cannot gauge what sort of domestic signature marketplace exchange will leave unless we compare unknown against known cases of market provisioning. A more appropriate test, according to Smith, is to compare Hirths observations from Xochicalco against cases where marketplace exchange was known to be prevalent (among the Aztecs or Romans) and where it appears to have been absent or scarcely developed (Andean states). Hirths approach underscores the need to better understand how other, noncommercial means of domestic provisioningsuch as redistribution, gift giving, and small-scale barteraffect the composition of domestic assemblages (Blanton 1998; Hicks 1998). Also crucial are the different scales at which marketplace and nonmarketplace transactions occur. Hirth (1998:456, 2000:188189) concentrates on uniformity in material culture assemblages among households of different social rank. I suggest that an equally important indicator of marketplace exchange concerns the uniformity of material culture assemblages over a large area. Localized, nonmarketplace- provisioning networks for example, kin-based networks of exchange (Abbott 2000; Sahlins 1972)may generate uniform household compositions among a group of households but at a much smaller scale than in market contexts. Households in nonmarketplace contexts are more apt to procure domestic wares through a variety of exchange networks and/or manufacture their own wares. In the Tonto Basin of eastern Arizona (from A.D. 1320 to A.D. 1450), for example, Rice and colleagues (1998:129) found that individual settlements independently and separately established external trade contacts, resulting in domestic assemblages that are uniform at a local scale (within individual settlements) but are heterogeneous at a regional scale. In market contexts, conversely, household inventories are likely to be relatively uniform over a large area (i.e., the area serviced by a single market center), albeit with some degree of variability resulting from differ-

ences in consumer preferences and purchasing power (see Smith 1999). According to Hirth (1998:455), other centralized forms of provisioning, such as redistribution, also might generate modest interhousehold uniformity but not to the same extent or in the same way as marketplace exchange. Redistribution transactions adhere to political networks and, therefore, product distributions will exhibit sociopolitical linkages, rather than a common provisioning source available to all households. Furthermore, it is unlikely that redistribution networks will incorporate everyday domestic goods, such as utilitarian wares that are subject to frequent use-related stress and need to be frequently replaced (e.g., domestic cooking pots). Few, if any, premodern governments possessed the resources or organizational efciency required to redistribute domestic wares over a large area. With respect to common domestic goods, therefore, redistribution systems likely generate less uniform domestic inventories over a large area than marketplace exchange systems. Another critique of the distributional approach concerns Hirths tendency to treat marketplace exchange as a presenceabsence phenomenon (Hassig 1998; Wilk 1998). The size and organizational structures of market systems varied considerably in different times and places in the ancient and modern world (Smith 1974, 1976). Hirth does not address how diverse market organizations differently affect the composition of domestic artifact assemblages. Directly relevant to Hirths approach is Bohannan and Daltons (1962; see also Kurtz 1974) study of markets in Africa, which demonstrates variability in the extent to which households in different areas rely on marketplace exchange to procure basic provisions. In peripheral market contexts, households do not rely on the marketplace for most of their basic provisions. Rather, peripheral markets are patronized chiey by target marketers, who trade in the market when they happen to have a surplus to exchange or need some item they do not produce (Hicks 1987:91). Differential use of marketplaces may occur among elite and nonelite households and among urban and rural populations within the broader market system. In fully integrated market contexts, conversely, the majority of households rely on marketplace exchange to procure all or most everyday provisions. One challenge for archaeologists is to infer

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 159

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

159

G ul fofM exci co

TEO TI H UACAN
Paci f i c O cean

LO W ER BLA N C O R EG I ON

N ort h 0 300 Ki l om et ers 600

Pi co de O ri zaba

Figure 1. Map of Mesoamerica showing approximate locations of the lower Blanco region, Teotihuacan, and the Pico de Orizaba obsidian source.

subregional variation in market integration among households. The method developed here provides a tool for evaluating market integration among households of different rank or at varying spatial scales. To my knowledge, since Hirth published his article in 1998, no follow-up studies have been devised to address these concerns and to empirically test or rene the distributional approach. My goal here is to expand on Hirths approach and address some of his critics. I introduce an alternative method of identifying marketplace exchange using a diversity measure (Mangurran 1988, 2004; Pielou 1975) that quanties variability among household artifact collections over a large area. The data used for this study come from Middle Postclassic period (ca. A.D. 1200A.D. 1350) sherd and obsidian collections from the lower Blanco region in the Mexican Gulf lowlands (Figure 1). These materials are from surface collections of the Proyecto Arqueolgico La Mixtequilla (PALM) in southcentral Veracruz (Stark 1991, 2006). To infer whether marketplace exchange was prevalent in the lower Blanco region, I rst compare diversity scores from a known context of marketplace exchangeLate Postclassic Teotihuacan (Garraty 1998, 2000, 2006a)against the Middle Postclassic lower Blanco region, an unknown con-

text, which addresses Smiths critique of Hirths approach. The comparative study suggests a probable marketplace exchange system in the lower Blanco region during the Middle Postclassic period, likely focused at El Sauce, the only known Middle Postclassic center in the region. I also analyze intraregional distributional changes in diversity statistics and in obsidian prismatic blade concentrations to infer the approximate radius of El Sauces market service area, which provide an empirical means of assessing market integration as a function of distance from the market center. This second analysis deals with the spatial dimensions of market integration in the lower Blanco region and helps to address Hassigs (1998) and Wilks (1998) point that marketplace exchange cannot be reduced to a simple presenceabsence phenomenon. I want to be clear that in this article, I am more concerned with methodology than with substantive interpretation or culture history. My goal is not to develop denitive conclusions about marketplace development and exchange in the lower Blanco region. Rather, my purpose is to demonstrate the efcacy of the approach used here to operationalize the distributional approach in hopes that the method developed here will be critically evaluated and adapted for similar studies in other archaeological contexts.

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 160

160

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

Measuring Diversity Diversity measures, which are most frequently used for studies of ecological biodiversity, provide a simple way of quantifying and comparing differences among a set of analytical units using a multivariate dataset (Pielou 1975, 1977; for discussions of diversity measures in archaeology, see Kintigh [1984], Lanata [1996], and the papers compiled in Leonard and Jones [1989]). The variables might refer to, for example, animal species or, in this case, distinct pottery types; analytical units may be sampling loci (e.g., surface collections) or distinct ecosystems. Diversity measures can be loosely grouped into three types: richness, evenness, and heterogeneity.2 Richness is simply the number of variables present in an analytical unit. For example, an ecosystem sustaining 20 plant species is richer than one sustaining 15 species. Evenness measures deviation between observed variable proportions for a given case and a hypothetical even distribution of variables. For example, in a dataset with four variables, perfect evenness is dened as 25 percent per variable; evenness statistics thus measure the cumulative differences between observed percentages and perfect evenness across all variables. The diversity measure used for the study broadly referred to as heterogeneityfollows the same basic principle as evenness. Rather than calculating differences from perfect evenness, however, heterogeneity is the differences in the variable composition between each analytical unit and a hypothetical standard or representative variable composition, which forms the expected proportions against which observed proportions are compared. In this study, the hypothetical collection-unit composition is calculated as the variable (column) percentages over all cases combined (Boone 1987:340). In other words, the expected reference assemblage is simply the row percentages calculated from the sums of counts for each variable or category. In some cases, it may be more appropriate to use an external dataset that represents an idealized set of proportions. The heterogeneity measure (H score) employed here is loosely based on Boones (1987) method of calculating interassemblage diversity in medieval North Africa. Following Boones approach, I dene perfect heterogeneity as the global or overall

proportions of each category (j) among all cases within a site or region (i). In this study, the cases are the individual surface collection, which presumably represent debris generated by one or several nearby households; the variables include various pottery type categories. The global type proportions that is, the reference dataset compose the expected values (Pij) against which the individual collection proportions (pij) are evaluated. The H score is the sum of the squared deviations between the observed per-unit percentages and the expected percentages across all type categories ([Pij pij]2). This calculation yields a single, positive score for each collection. Lower scores indicate that the observed percentages closely resemble the expected percentages, suggesting greater uniformity. Higher scores suggest that one or more categories dominate the collection, indicating greater heterogeneity. Boones calculation includes a complex weighting procedure in which each percentage is standardized as a ratio of one of the categorys proportions (which one is not important), which is meant to control for variability in discard rates of different artifact types (Boone 1987:340). For this study, controlling for the differential discard intensities is beside the point. Kintigh (2002) offers a simpler way of calculating distances between individual collections units and overall type percentages by using Brainerd-Robinson coefcients of similarity (BR) (Brainerd 1951; Robinson 1951). The BR coefcient is the sum of the differences between these percentages across all variables subtracted from 200, thereby generating an integer between 0 and 200 in which a score of 200 represents maximum homogeneity. For ease of interpretation, the BR coefcients can be scaled as proportional values so that lower scores indicate greater intercollection uniformity and higher scores indicate heterogeneity (see Kintigh 2002).3 Heterogeneity measures offer a useful way of evaluating the existence of marketplace exchange, as dened by Hirths distributional approach. As long as the cases (collections) correspond to domestic consumption units (one or a few households), and the categories compose relatively distinct commodities or units of purchase (visually distinct decorated pottery types or suites of similar types), then the mean or median heterogeneity scores among a set of domestic artifact collections can potentially

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 161

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

161

reveal the relative degree to which households over a large area obtained goods from a common provisioning source, presumably a central marketplace (see above for rationale). Low H scores thus potentially suggest marketplace exchange, and high H scores suggest multiple, presumably nonmarket provisioning sources. When interpreting H scores, one must be mindful of the many other possible bases for variation among a set of collections, including natural and cultural formation processes (see Plunkett 1998). For these reasons, artifact compositions inevitably vary among collections to some extent, given sampling vagaries and interhousehold differences in domestic activities, consumer preferences, and purchasing power. It is therefore unreasonable to expect very uniform domestic artifact collections (i.e., H scores at or near zero). Even in contexts in which a single marketplace provisioned all households over a large area, a certain degree of tolerance of intercollection variability is appropriate. However, it is reasonable to predict that households serviced by a central marketplace or market system over a large area, on average, will have more uniform artifact collections (lower H scores) than households in contexts where marketplace exchange was absent or poorly developed. For the present analysis, I expect that the data from Late Postclassic Teotihuacan will generate lower H scores than the data from the lower Blanco region, if marketplace exchange was not prevalent in the latter area. Using Heterogeneity to Identify Marketplace Exchange Heterogeneity is only useful as a relative, not an absolute, measure of diversity among the collections (see Kintigh 1989:26). It is therefore only suited for comparing collections, for example, within or among sites and regions. One pertinent use of this method is to calculate mean or median H scores for pooled collection units over a region or site (or locality within a site) so that different sites or subregions can be compared. The H statistic works particularly well where one can compare the scores from a known case of marketplace exchange against an unknown case (see below). Ideally, the sites or regions chosen for analysis roughly correspond to the area serviced by a single marketplace or market system, preferably a

large area subjected to full-coverage, systematic survey. (Excavation projects seldom encompass a large enough area for robust interhousehold analyses over such a large area.) Interpreting intercollection heterogeneity is fraught with complications, however. First, the H statistic, like most statistical measures, is susceptible to sampling vagaries: small sample sizes may articially inate H scores, which are calculated from percentages (Boone 1987:341; Kintigh 1989:29). In some cases, it may be necessary to combine individual collections into larger units of analysis to achieve more robust sample sizes. One also must avoid including one or two unusually large collections among a set of cases, or else those very large cases will overwhelm in effect, denethe overall percentages (based on the column sums across all collections) from which H scores are derived. A potentially more troubling problem is inherent in dening the scale of analysis. As explained above, for example, uniformity among a group of domestic artifact assemblages may occur on a small scale, even in nonmarket, localized exchange settings. Marketplace exchange likely generate more homogeneous domestic assemblages over a much larger area. However, at larger scales, archaeologists may unknowingly study a region that incorporated two or more separate market systems in ancient times, producing a boundary problem (see Johnson 1977:498) that skews the regional homogeneity results. Blantons (1996:59) crosscultural study of early market systems is helpful in determining the expected size of a single marketplaces service area. A standard market area, according to Blanton (1996:59), ranges from about 50 to 200 km2 in area with radii ranging from 4 to 9 km. Blantons observation provides a useful baseline for interpreting whether a study area incorporates one market zone area or several adjacent zones. As explained below, the two survey areas compared for this study roughly correspond to the expected service area for a single market center. Despite these problems, heterogeneity measures have the potential to illuminate the long-term evolution of premodern economies and market systems in contexts where written records are unavailable (but see Carrasco [1998] for a contrary perspective). First, comparing H scores from different periods may shed light on the nascent development of mar-

LAQ20(1) Garraty:Layout 1

3/3/09

3:41 PM

Page 162

162

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

ketplace exchange, the tempo of market integration over time, and the expansion and contraction of market zones. Second, because market participation in peasant-market societies is typically uneven (Plattner 1989:188190; Smith 1972:215; Wanmali 1981), comparing H scores in different subareas of a region may elucidate differences in market participation among urban and rural communities or between elite and nonelite populations. Third, comparisons of H scores among artifact categories (e.g., ne vs. domestic goods) may provide insights into differences in the scales and mechanisms of exchange for different classes of commodities. In the remainder of this article, I use the H statistic to analyze archaeological surface collections from the lower Blanco region. As stated above, my principal objective is not to derive conclusive evidence of marketplace exchange. In some cases, sample sizes are simply too small to permit robust analyses. The case study is meant to demonstrate the efcacy of using H scores to operationalize the distributional approach and provide a touchstone for similar studies. Inferring Marketplace Exchange from a Known to an Unknown Context In this section, I compare surface sherd collections from Late Postclassic (Aztec) Teotihuacan (ca. A.D. 1350A.D. 1520), a known locus of marketplace exchange, with collections from the Middle Postclassic lower Blanco region (ca. A.D. 1200A.D. 1400), an unknown locus for marketplace exchange. The data for the comparative study come from surface collections taken during two full-coverage systematic survey projects. Before presenting the results, I rst introduce the survey projects and the sherd collections used for the study. The Survey Projects The surface collections from the lower Blanco region come from the Proyecto Arqueologa La Mixtequilla (PALM), directed by Barbara Stark (1991, 2006) in the 1980s, 1990s, and early 2000s (Figure 2). Although the PALM project covered a total area of 99 km2 in several noncontiguous survey blocks, the present study focuses exclusively on the Central Block (49.2 km2) and six smaller blocks (surveyed by Speaker [2001] in 1989) sur-

rounding it (22.4 km2), which together encompass 61.8 km2. The other PALM survey blocks are excluded from this study because they possessed very few or no surface sherd collections assigned to the Middle Postclassic period. The Late Postclassic Teotihuacan data come from the surface collections of the Teotihuacan Mapping Project (TMP), directed by Ren Millon (1973; Millon et al. 1973) in the 1960s (Figure 3). The TMP survey encompassed a 36-km2 contiguous survey block and includes one Late Postclassic center, San Juan Teotihuacan (Garraty 2000, 2006a). Of importance, even though Teotihuacan is typically conceived as a site, and the PALM survey area covered part of a large region, the sizes of the two coverage areas are not drastically different. The portion of the PALM study area considered for this study is roughly 25 km2 larger than the TMP survey area. Even so, both projects encompass survey areas that fall within the range that might have been serviced by a single marketplace, based on Blantons estimates (see above). Both projects were essentially nonsite surveys predicated on recording and collecting individual surface features, such as house mounds or discrete artifact concentrations, over a large and continuous area. In both projects, each surface collection likely represents artifact accumulations relating to one or perhaps a few nearby households. I therefore consider the two survey areas to be suitable for comparison.4 Both surveys areas also likely include a single center and surrounding settlements. The 36-km2 TMP survey area includes the known market center of San Juan Teotihuacan and its immediate surrounding territory. San Juan very likely housed a central marketplace that served the local community in the Teotihuacan area during the Late Postclassic period (Garraty 2006a). The Central Block and surrounding blocks in the PALM survey area includes El Sauce, the only known Middle Postclassic center in the lower Blanco region, and surrounding residences (Curet et al. 1994; Garraty and Stark 2002). If a central marketplace was present in the region, it likely would have been located at El Sauce. Both regions also included pottery and obsidian producers that may have vended their wares in marketplaces. Late Postclassic (Aztec) pottery was produced at a number of locations in the Basin of

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 163

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

163

Figure 2. The PALM Central Block (dark gray) and nearby blocks surveyed by Stuart Speaker in 1989 (light gray diagonal lines). Small dots indicate locations of 80 surface collections used for the comparative study. The outline of the Middle Postclassic center of El Sauce is shaded dark gray. Also shown are the two possible loci of Middle Postclassic pottery production.

Mexico and trafcked through an extensive regional market system (Garraty 2006b; Hodge and Minc 1990; Minc 1994, 2006; Nichols et al. 2002). To date, no pottery production loci have been identied at Aztec Teotihuacan or any other site. Spence (1985) has detected a number of obsidian manufacturing loci within the TMP survey area, however. In the lower Blanco region, Curet (1993) located a Middle Postclassic community specializing in the production of comales (tortilla griddles) roughly two kilometers east of El Sauce. Stark and Garraty (2004) also detected a probable production locus for decorated bichrome bowls (local-style Black-on-red and Black-on-orange wares) roughly 12 kilometers southeast of El Sauce (see Figure 2). Both comales and bichrome bowls are well represented in collections throughout the Central Block and may have been sold in a marketplace at El Sauce. Heller (2000) identied an obsidian production locus within El Sauce, which also likely

would have been sold in the local marketplace, if one existed. The Sherd Collections The sherd categories used for calculating H scores include visually distinguishable sherd types or suites of similar types; each of these categories likely comprised a distinct class of salable market commodities. In both the PALM and TMP datasets, sherds were coded using a type-variety system (presented in Hodge and Minc [1991] for the Basin of Mexico collections and in Stark [1995] for the lower Blanco collections). I combined a number of similar types and variants from both survey areas into larger categories to bolster variable frequencies (i.e., to avoid a large number of zero cell frequencies). I previously combined types and variants for statistical analyses of both the TMP (Garraty 2006a:Table 1) and PALM sherd datasets (Garraty and Stark 2002:Table 1).

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 164

164

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

Figure 3. The TMP survey area, showing survey grid, monumental features, and approximate location of San Juan Teotihuacan. Dots indicate locations of 27 surface collections used for the comparative study.

For the lower Blanco survey, Stark (1995) identied a large number of decorated types and variants assigned to the Middle Postclassic period. In all, percentages for 11 sherd categories were used as variables used for calculating H scores. These include 10 major categories: local-style bichrome Black-on-red and Black-on-orange bowls (both of which include multiple decorative variants); three types of polychrome vessels (White and Black/Red wares, Dull Buff polychromes, and Complicated [Cholutecoid] polychromes); frieze-motif vessels (three types); interior-banded types (ve types); Fondo Sellado (stamp-based bowls); comales; and unslipped El Sauce wares (two types). An 11th, miscellaneous category includes ve lowfrequency decorated types. Other low-frequency types are excluded from this analysis because of uncertain chronological placement. Stark (1995) provides detailed descriptions and illustrations of each of these types, variants, and form classes. For Late Postclassic Teotihuacan, the variables

include 12 decorated sherd categories from Garratys (1998, 2000, 2006a) analysis of Late Postclassic surface collections from the TMP (based on Hodge and Minc [1991] and Minc [1994]:419530). The most abundant Late Postclassic decorated type in the Basin is Aztec Blackon-orange (black painted designs on an orange surface), which is subdivided here into two variants of interior-decorated tripod vessels (Variants D and E) and four variants of exterior-decorated bowls (Variants DE, F, G2, and H). A second major decorated type is Aztec Black-on-red (black painted designs on red painted exterior surface), subdivided into two bowl variants (comb-motif vessels [Variant C] and Late Prole vessels [see Minc 1994:466]) and Black-on-red copas (goblets).5 Black-and-white-on-red (black and white painted designs on a red painted exterior surface) includes two bowl variants (Variant G and Late Prole vessels). A nal, less prominent decorated type is Black-and-red-on-buff, which occurs exclusively

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 165

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

165

as bowls (Hodge and Minc 1991). Most of these categories include several combined variants and subvariants. Few plainware sherds are included in this study because it is generally more difcult to infer their probable salable units of exchange. For instance, plainware jars occur in a variety of shapes and sizes, but functional classes cannot easily be discerned without detailed analyses of vessel form, surface treatment, and paste; the salable units of exchange are thus not evident. In addition, most plainwares cannot be readily assigned to a specic time period or site component. Two exceptions are comales and El Sauce unslipped wares from the lower Blanco region, both of which are easily distinguished and occur predominantly (comales) or exclusively (El Sauce unslipped) in Postclassic collections. I used different methods to make period assignments for the PALM and TMP sherd collections. For PALM, individual collections were assigned to the Middle Postclassic period through a statistical unmixing procedure that Barbara Stark and I developed to deal with the high frequency of collections with mixed temporal components (Garraty and Stark 2002:79). The statistical procedure to unmix the Postclassic collections is based on a method developed by Kohler and Blinman (1987) to ascertain the proportionate contribution of individual temporal components in multicomponent collections. We categorized all Postclassic collections as Middle Postclassic, Late Postclassic, or mixed-component collections, which produced 611 relatively pure Middle Postclassic sherd collections. Not all of these collections are included in this study, however, given the minimum percollection size requirements outlined below. For Late Postclassic Teotihuacan, I only included collections assigned to the Late Postclassic period, a dataset I dened for a previous study (Garraty 2000). The Postclassic sherd component from the TMP could be distinguished and separated from earlier sherd components, given the lack of continuity of most pre-Postclassic types. As was the case with the PALM collections, however, most TMP collections included a mix of types dating to the Middle and Late Postclassic and Early Colonial periods (see Garraty 2006a). The Late Postclassic collections were thus dened as those in which 80 percent or more of the temporally diagnostic types date to the Late Postclassic period

(using Hodge and Mincs [1991] and Mincs [1994] period assignments for types), which generated a total of 76 Late Postclassic period collections. The collections used for the present study are drawn from this sample, but some were excluded because of the collection-size minimums outlined below. Calculating the H Scores The sherd categories outlined above were used as input variables for calculating H scores for individual collections. One problem with removing plainwares from the calculation of H scores is that it substantially reduces sample sizes per collection. I therefore had to lower the minimum sample size to achieve a robust number of collections from all areas of the survey region. For the TMP collections, the minimum sample size is 15 rims per collection, which allowed for an analysis of 27 collections (shown in Figure 3). The sample size threshold of 15 is admittedly very small, but any increase in the threshold would result in a substantial reduction of cases. The PALM data are more robust than the TMP data. I was able to analyze 80 Middle Postclassic collections with 30 or more rim sherds (shown in Figure 2). These 80 collections do not include the suspected loci of comal and bichrome bowl production (Curet 1993; Stark and Garraty, 2004). Many of the rims from these productionrelated collections are likely de facto wasters that relate to production and not household provisioning. Their inclusion would skew the H scores toward greater uniformity. Type percentages for each collection were calculated as the number of sherds for each type category divided by the total number of sherds across all categories, multiplied by 100 (rims only for the PALM dataset). The H scores are the BR coefcients of similarity (scaled to a number between 0 and 1, as explained above) between the observed per-collection percentages and the percentages over all collections. I present summary statistics for the H scoresincluding mean, median, standard deviation, and score rangeto characterize variability among domestic sherd collections within the PALM and TMP survey areas (see Table 1 and Figure 4). Comparing Homogeneity Results As indicated in Table 1 and Figure 4, the mean and median H scores and standard deviations are nearly

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 166

166

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

Table 1.Descriptive Statistics for H Scores from the Middle Postclassic Lower Blanco Region and Late Postclassic Teotihuacan. Number of Collections 27 80 Mean H score .278 .284 Median H Score .29 .27 Score Range .41 .52

Late Postclassic Teotihuacan Middle Postclassic Lower Blanco ANOVA: F (1, 105) = .075, p = .78

St. Dev. .096 .092

identical for Late Postclassic Teotihuacan and the Middle Postclassic lower Blanco region. This result tentatively suggests that most Middle Postclassic households in the lower Blanco region obtained pottery through marketplace exchange, which certainly was the case at Late Postclassic Teotihuacan. As expected, the collections in both areas do not reveal very uniform sherd compositions. The H scores of roughly .3 indicate that, on average, the cumulative variability across all categories between the individual and overall percentages accounts for about 30 percent in both survey areas. This variability likely stems from sampling vagaries as well as interhousehold differences in activities, buying power, and consumer choices. An analysis of variance (ANOVA) between the two sets of H scores suggests no signicant statistical difference and a 78 percent probability that the variance in H scores for the two datasets derive from the same population (see Table 1). In this case, the hypothetical parent population refers broadly to artifact assemblages derived from marketplace exchange in general. Future corroboration of this interpretation will require additional, independent lines of evidence and comparison against a third, known case of nonmarketplace exchange encompassing a comparable area. The range of H scores for the PALM data is somewhat larger than for the TMP data, however, which probably relates to the different number of collections from the two surveys. The PALM dataset included more than two times as many collections and, therefore, is more likely to encompass a wider range of scores. However, the disparity in ranges might also be partially attributable to differences in market integration in the two regions. As I explain below, households located in areas distant from El Sauce may have been located outside of the centers principal market service area, which might partially account for the larger score range. Given that El Sauce was likely the only market center in the region, it is reason-

able to expect a fall-off in market participation (as evidenced by the H scores) with greater distance from El Sauce. In the Teotihuacan area, conversely, even households located at a distance from San Juan Teotihuacan could patronize one of several neighboring market centers (e.g., Otumba, Acolman) offering a similar selection of market commodities. I thus would not expect a fall-off in market participation with distance from San Juan Teotihuacan. In sum, a single marketplace at El Sauce probably provisioned most households in the Middle Postclassic lower Blanco region, but the higher score range for the PALM dataset raises the possibility that some areas more distant from El Sauce were not fully integrated into the centers market hinterland. Households in these distant areas may have relied less on marketplace exchange to procure market provisions than households closer to El Sauce. I explore this possibility in the following section. Inferring El Sauces Market Service Area Unlike Late Postclassic Teotihuacan, El Sauce probably was not part of a interlocking regional system of interconnected market centers (see Smith 1974, 1976). The lower Blanco region thus offers an ideal case study for evaluating the service area radius for a single market center. To infer El Sauces market service radius, I compare distributional changes in H scores with increasing distance from the center with fall-off patterns (monotonic decrement) in obsidian prismatic blade concentrations (see Renfrew 1975). The distributional changes in H scores (for sherd data) and obsidian blade concentrations provide complementary lines of evidence for assessing the size of the market hinterland surrounding El Sauce. However, before delving into the analytical results, it is essential rst to consider pottery and obsidian as market commodities. The obsidian ana-

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 167

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

167

0. 7 M edi an 25% 75% N onO ut l i er Range O ut l i er s

0. 6 H Scor es ( Scal ed BR Coef f i ci ent s )

0. 5

0. 4

0. 3

0. 2

0. 1

LPC T eot i huacan

M PC L.Bl anco

Figure 4. Box-and-whisker plots (median and inner quartile range) of H scores for the Middle Postclassic (MPC) lower Blanco region and the Late Postclassic (LPC) Teotihuacan datasets.

lyzed here is clear gray in color and derives from the Pico de Orizaba source in highland Veracruz, located roughly 100 kilometers west of the lower Blanco region (see Figure 1). Clear-gray obsidian was heavily exploited in the lower Blanco region during the Middle Postclassic period but considerably less so during earlier and later periods (Heller 2000; Heller and Stark 1998). Obsidian was thus a nonlocal, but high-demand commodity in the Gulf lowlands and likely needed to be disseminated to consumers through a marketplace or some other centralized provisioning source. Obsidian likely was frequently sold in the form of prismatic blades in Mesoamerica. Unlike obsidian, pottery need not have been disseminated through a marketplace or centralized provisioning source, given the widespread availability of clays and production technology (Stark 1992). Even so, many potters in the lower Blanco region might have sold their wares in the marketplace in order to tap a larger consumer population and boost revenues. To help reconstruct El Sauces market service area, I created a series of thirteen 1-km ring buffers around the center to evaluate changes in H scores and clear- gray obsidian concentrations with

increasing distance (Figure 5). I inspected the variability in H scores and obsidian concentrations among collections in each of the 13 buffers. For the sherd data, if market participation dropped off, I would expect increases in mean and median H scores in the more-distant buffers. The households located farther away from El Sauce instead might have obtained pottery through local, nonmarket sources, given the added travel costs to the market center. For the clear-gray obsidian data, a fall off in obsidian concentrations should be evident in the more-distant buffer rings. To calculate H scores for this analysis, I lowered the minimum number of Middle Postclassic rims per collection to 20 to assemble a larger number of collections per buffer. This is especially pertinent for Buffers 812, which only would have included a small handful of collections per buffer if I had retained the minimum collection size of 30 rims used for the comparative study above. In this case, the H scores derive from the BR distances between the per-collection row percentages and the percentages over all collections (expected percentages). The mean and median H scores for each buffer are listed in Table 2 and

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 168

168

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

Figure 5. The thirteen one-kilometer buffers surrounding El Sauce.

illustrated in Figure 6. The results suggest a possible market service radius for El Sauce of about nine kilometers. A marked increase in mean and median H scores is evident at the 9-km buffer: all buffers within 9 km of El Sauce have H scores of roughly .3. Beyond the 9-km buffer, H scores increase to between .41 and .47 (but note the overlap in the H score ranges between Buffers 19 and Buffers 1013, as evidenced by the per-buffer standard deviations shown in Figure 6). The 9-km radius for El Sauces market service area, if valid, is slightly larger than the typical 48-km market-service radius suggested by Blanton (1996; see above). Perhaps the larger radius resulted from the absence of other nearby marketplaces in the region. Given the absence of competing market centers, consumers might have been more willing to travel longer distances to attend the El Sauce marketplace, resulting in a somewhat stretched service radius. The clear-gray obsidian data show similar, but not equivalent, results (Table 3 and Figure 6). Heller (2000; Heller and Stark 1998) identied one Mid-

dle Postclassic obsidian workshop at El Sauce (Collection 1756), which she suggests was disseminated to consumers through a marketplace. Other obsidian production loci may have been located elsewhere in the lower Blanco region. Given the existence of at least one probable obsidian workshop, I calculated obsidian concentrations as median counts of clear-gray prismatic blades per 100 Middle Postclassic rims. Median values were preferable over means to avoid inating the obsidian concentrations in the buffers that include known or possible obsidian workshops. Also, because the obsidian concentrations are calculated relative to rim counts, I only included cases with a minimum of 10 Middle Postclassic rims.6 I calculated the ratios on a percollection rather than a per-buffer basis (with all collections combined) because I am concerned with changes in interhousehold variability in obsidian consumption with greater distance from the center rather than the fall-off pattern in raw obsidian blade concentrations. The per-collection measure more directly accommodates the empirical expectation of Hirths distributional approach.

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 169

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

169

Table 2. Descriptive Statistics for H Scores From the Thirteen 1-km Buffers Surrounding El Sauce. Buffer Number of Mean H range (km) collectionsa Score 1 40 .28 2 12 .30 3 17 .32 4 12 .31 5 11 .28 6 16 .28 7 20 .29 8 7 .34 9 7 .31 10 4 .47 11 7 .41 12 4 .42 13 5 .44 All buffers 162 .31 aMinimum collection size: 20 Middle Postclassic rims. ANOVA: F(12, 149) = 4.09, p < .001 Median H Score .27 .31 .28 .30 .27 .28 .27 .28 .32 .45 .42 .41 .42 .29

St. Dev. .08 .10 .12 .08 .09 .07 .07 .08 .05 .11 .11 .11 .05 .10

Figure 7 and Table 3 shows the fall-off pattern for the number of prismatic blades per 100 rims with increasing distance from El Sauce. The median values per buffer ring fall off signicantly after the 6-km buffer, rather than after the 9-km buffer. The fall-off pattern for obsidian thus is steeper than it is for the H scores, suggesting a more abrupt edge to the market-distribution range. This result
6.0 5.5 5.0

might suggest that consumers were less willing to travel as far to procure obsidian blades as they were to purchase decorated pottery at the El Sauce marketplace. Additional research will be needed to better understand the differences in the distributional and fall-off patterns for pottery and obsidian blades. Also notable is the spike in obsidian concentrations at the 13-km buffer, which might indicate

M ean +/ -1 St .D ev.

HScor es ( Scal ed BR Coef f i ci ent s)

4.5 4.0 3.5 3.0 2.5 2.0 1.5 1 km 2 km 3 km 4 km 5 km 6 km 7 km 8 km 9 km 10 km 11 km 12 km 13 km

1 K m Buf f erRange Sur r oundi ng ElSauce

Figure 6. Plot of means and standard deviations of H scores for collections in each of the 13 one-kilometer buffers surrounding El Sauce.

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 170

170

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

Col l ect i on M i ni mum:10 Sher ds


60 Cl ear Gr ay Pr i sm at i c bl ades /100 M i ddl e Post cl assi c Ri ms

50

M edi an I nner quar t i l er ange

40

30

20

10

0
1 km 2 km 3 km 4 km 5 km 6 km 7 km 8 km 9 km 10 km 11 km 12 km 13 km

1 Km Buf f er s Sur r oundi ng ElSauce


Figure 7. Plot of medians and inner-quartile ranges of clear-gray obsidian prismatic blade frequencies per 100 Middle Postclassic rims in each of the 13 one-kilometer buffers surrounding El Sauce.

obsidian production loci in that buffer range. Another possibility is that commercial middlemen residing in the 13-km buffer accumulated for resale to households located in areas farther away from El Sauce. Obsidian is lightweight, transportable, and, therefore, amenable to trading through itinerant peddlers. Perhaps middlemen in this area pro-

cured obsidian at El Sauce and resold it locally to consumers residing relatively far away from the market center. A second possibility is that an additional, but as-yet-undetected Middle Postclassic market center or other provisioning source was located in the vicinity of the 13-km buffer. Previous indications of bichrome bowl production loci

Table 3. Median and Inner-Quartile Ranges for Clear-Gray Obsidian Prismatic Blade Frequencies per 100 Middle Postclassic Rims in the Thirteen 1-km Buffers Surrounding El Sauce. Frequency of clear-gray obsidian prismatic blades per 100 Middle Postclassic rims Buffer range (1 km) Number of collectionsa 25th percentile Median 75th percentile 1 51 17.4 33.3 57.1 2 27 6.3 14.3 28.9 3 42 6.7 17.9 35.3 4 28 10.3 23.9 35.7 5 23 10.0 20.0 50.0 6 25 4.8 17.6 38.3 7 24 4.7 8.9 22.5 8 19 0.0 10.0 23.5 9 16 0.0 10.8 12.3 10 6 0.0 9.4 12.8 11 11 0.0 12.0 20.0 12 9 5.6 10.0 16.7 13 17 19.0 26.7 35.8 All buffers 298 7.7 17.1 35.7 aMinimum collection size: 10 Middle Postclassic rims. ANOVA: F(12, 285) = 3.74, p < .001

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 171

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

171

in the same vicinity (Stark and Garraty 2004) support this possibility (see Figure 2). Additional lines of evidence, larger sample sizes, and complementary methods and approaches will be required to fully test the possibility of marketplace exchange in the lower Blanco region and better dene El Sauces market service area. Especially important will be further investigation of a possible market center or provisioning source in the vicinity of the distant 13-km buffer surrounding El Sauce. Conclusion: Reevaluating the Distributional Approach Hirth (1998, 2000:182194) has made a valuable contribution to archaeological method and theory by devising a useful approach to detecting marketplace exchange in premodern contexts where written records are unavailable. As several commentators have indicated, however, Hirths approach requires some renement. I address some of these critiques by proposing a different, complementary means of assessing marketplace exchange. First, by calculating global homogeneity scores for each site or context, one can readily compare known against unknown cases of marketplace exchange, which accommodates Smiths (1999) critique of Hirths approach. Testing the unknown lower Blanco dataset against the known Late Postclassic Teotihuacan dataset suggests that marketplace exchange may have been the primary means of obtaining decorated pottery in the lower Blanco region during the Middle Postclassic period. In the future, I hope to evaluate these results against a known case of nonmarketplace exchange. Second, the fall-off patterns in H scores and obsidian prismatic blade concentrations provides a way of assessing the relative degree of market integration (Bohannan and Dalton 1965) as a function of distance from the market center, which partially addresses Hassigs (1998) and Wilks (1998) concern that Hirth treats marketplace exchange as a simple presenceabsence phenomenon. Changes in the distributions of H scores (sherd collections) and clear-gray obsidian blade frequencies with increasing distance from El Sauce suggest a market service area between about six and nine kilometers. Had El Sauce been part of a larger, interlocking regional market system, I would expect

a relatively even distribution of H scores and obsidian concentrations over the entire region with no evidence of a fall-off. The approach used for this study offers a different and complementary method of evaluating marketplace exchange to the approach Hirth used for his study at Xochicalco. He demonstrates marketplace exchange based on evidence that both elites and commoners had access to the same assortment of imported pottery and obsidian commodities, thus producing relatively homogenous artifact assemblages among households, regardless of social rank (see also Smith 1999). Equally important to social rank, however, is the spatial scale of interhousehold heterogeneity among domestic artifact assemblages. Marketplace exchange generates more homogenous material culture assemblages over a large area relative to nonmarket forms of exchange. Hirths study of Xochicalco does not consider the important issue of scale. To conclude, clearly additional methodological inquiry will be needed to better determine the considerable variability in the size, character, and structure of ancient market systems. Even so, archaeologists should follow Hirths lead in determining which variables or indicators are sensitive to marketplace exchange and how best to quantify them. Hirths approach, as well as the method proposed in this article, has the potential to illuminate premodern economies not just in Latin America but also in all areas of the ancient world. Continued work on this issue will yield better and more nuanced theories and methods for modeling the evolution of premodern complex economies.
Acknowledgments. I am grateful to Barbara Stark, George Cowgill, Kenneth Hirth, Keith Kintigh, Alanna Ossa, Ian Robertson, Kate Spielmann, and two anonymous reviewers for comments on earlier versions of this manuscript. I also thank Lynette Heller, who assisted with the PALM obsidian data. The PALM survey was supported by grants (to Barbara Stark) from the National Science Foundation (BNS 8519167, BNS 8741867, and SBR9804738), the National Geographic Society, and Arizona State University. PALM was conducted with the permission of the Instituto Nacional de Antropologa e Historia in Mexico. Research on Late Postclassic collections from the TMP was supported by grants (to Garraty) from the Arizona State University (ASU) Latin American Studies Program, the ASU Chapter of Sigma Xi, and the ASU School of Human Evolution and Social Change. Ian Robertson created the TMP MapInfo database used for part of the analysis, with the assistance of George Cowgill, based on the maps published in Millon et al. (1973).

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 172

172

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009

I am indebted to them for permitting me to access the TMP MapInfo database and to Barbara Stark for permitting me access to the PALM MapInfo database.

References Cited
Abbott, David R. 2000 Ceramics and Community Organization among the Hohokam. University of Arizona Press, Tucson. Blanton, Richard E. 1996 The Basin of Mexico Market System and the Growth of Empire. In Aztec Imperial Strategies, coauthored by Frances F. Berdan, Richard E. Blanton, Elizabeth Hill Boone, Mary G. Hodge, Michael E. Smith, and Emily Umberger, pp. 4784. Dumbarton Oaks Research Library and Collection, Washington D.C. 1998 Comment on Hirth. Current Anthropology 39:463464. Bohannan, Paul, and George Dalton 1962 Introduction. In Markets in Africa, edited by Paul Bohannan and George Dalton, pp. 126. Northwestern University Press, Evanston, IL. Boone, James L., III 1987 Dening and Measuring Midden Catchment. American Antiquity 52:336345. Brainerd, George W. 1951 The Place of Chronological Ordering in Archaeological Analysis. American Antiquity 16:301313. Carrasco, Pedro 1998 Comment on Hirth. Current Anthropology 39:464465. Curet, L. Antonio 1993 Regional Studies and Ceramic Production Areas: An Example from La Mixtequilla, Veracruz. Journal of Field Archaeology 20:427440. Curet, L. Antonio, Barbara L. Stark, and Sergio Vsquez Z. 1994 Postclassic Changes in Veracruz, Mexico. Ancient Mesoamerica 5:1332. Garraty, Christopher P. 2000 Ceramic Indices of Aztec Eliteness. Ancient Mesoamerica 11:323340. 2006a Aztec Teotihuacan: Political Processes at a Postclassic and Early Colonial City-State in the Basin of Mexico. Latin American Antiquity 17:363387. 2006b The Politics of Commerce: Aztec Pottery Production and Exchange in the Basin of Mexico, A.D. 12001650. Unpublished Ph.D. Dissertation, School of Human Evolution and Social Change, Arizona State University, Tempe. Garraty, Christopher P., and Barbara L. Stark 2002 Imperial and Social Relations in Postclassic Southcentral Veracruz. Latin American Antiquity 13:333. Hassig, Ross 1998 Comment on Hirth. Current Anthropology 39:467. Heller, Lynette 2000 Postclassic Obsidian Workshop Debris from El Sauce, Veracruz, Mexico. Mexicon XXII:139146. Heller, Lynnette, and Barbara L. Stark 1998 Classic and Postclassic Obsidian Tool Production and Consumption: A Regional Perspective from the Mixtequilla, Veracruz. Mexicon XX:119128. Hicks, Frederic 1987 First Steps Towards a Market-Integrated Economy in Aztec Mexico. In Early State Dynamics, edited by Henri Claessen and Pieter van de Velde, pp. 9107. Brill, Leiden.

1998 Comment on Hirth. Current Anthropology 39:467468. Hirth, Kenneth G. 1998 The Distributional Approach: A New Way to Identify Marketplace exchange in the Archaeological Record. Current Anthropology 39:451476. 2000 Ancient Urbanism at Xochicalco: The Evolution and Organization of a Pre-Hispanic Society. Archaeological Research at Xochicalco, Vol. 1. University of Utah Press, Salt Lake City. Hodge, Mary G., and Leah D. Minc 1990 The Spatial Patterning of Aztec Ceramics: Implications for Prehispanic Exchange Systems in the Valley of Mexico. Journal of Field Archaeology 17:415437. 1991 Aztec-Period Ceramic Distribution and Exchange Systems. Final Report Submitted to the National Science Foundation for Grant BSM8704177, Washington DC. Johnson, Gregory A. 1977 Aspects of Regional Analysis in Archaeology. Annual Review of Anthropology 6:479508. Kintigh, Keith W. 1984 Measuring Archaeological Diversity by Comparison with Simulated Assemblages. American Antiquity 49:4454. 1989 Sample Size, Signicance, and Measures of Diversity. In Quantifying Diversity in Archaeology, edited by Robert D. Leonard and George T. Jones, pp. 2536. Cambridge University Press, Cambridge. 2002 Tools for Quantitative Archaeology: Programs for Quantitative Analysis in Archaeology. Programs manual published by the author, Tempe. Electronic document, http://tfqa.com, accessed January 27, 2009. Kohler, Timothy A., and Eric Blinman 1987 Solving Mixture Problems in Archaeology: Analysis of Ceramic Materials for Dating. Journal of Anthropological Archaeology 6:128. Kurtz, Donald V. 1974 Peripheral and Transitional Markets: The Aztec Case. American Ethnologist 1:685705. Lanata, Jos Luis 1996 La diversidad instrumental en el sudeste fueguino. Arqueologa 6:159197. Leonard, Robert D., and George T. Jones (editors) 1989 Quantifying Diversity in Archaeology. Cambridge University Press, Cambridge. Magurran, Anne E. 1988 Ecological Diversity and its Measurement. Princeton University Press, Princeton. 2004 Measuring Biological Diversity. Blackwell, Malden, MA. Millon, Ren 1973 Urbanization at Teotihuacan, Mexico: The Teotihuacan Map, Part 1: Text, Vol. 1. University of Texas Press, Austin. Millon, Ren, R. Bruce Drewitt, and George L. Cowgill 1973 Urbanization at Teotihuacan, Mexico: The Teotihuacan Map, Part 2: Maps, Vol. 1. University of Texas Press, Austin. Minc, Leah D. 1994 Political Economy and Market Economy Under Aztec Rule: A Regional Perspective Based on Decorated Ceramic Production and Distribution Systems in the Valley of Mexico. 2 vols. Ph.D. Dissertation, University of Michigan, University Microlms, Ann Arbor. 2006 Monitoring Regional Market Systems in Prehistory: Models, Methods, and Metrics. Journal of Anthropological Archaeology 25:82116.

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 173

G%((%)*$

DISTrIBuTIONAL APPrOAcH TO mArKeTPLAce e!cHANGe

173

Nichols, Deborah L., Elizabeth M. Brumel, Hector Neff, Mary G. Hodge, Thomas H. Charlton, and Michael D. Glascock 2002 Neutrons, Markets, Cities, and Empires: A 1000-Year Perspective on Ceramic Production and Distribution in the Postclassic Basin of Mexico. Journal of Anthropological Archaeology 21:2582. Pielou, Evelyn C. 1975 Ecological Diversity. Wiley and Sons, New York. 1977 Mathematical Ecology. Wiley and Sons, New York. Plattner, Stuart 1989 Markets and Marketplaces. In Economic Anthropology, edited by Stuart Plattner, pp. 171208. Stanford University Press, Stanford. Plunkett, Patricia 1998 Comment on Hirth. Current Anthropology 39:468469. Polanyi, Karl 2001 [1944] The Great Transformation: The Political and Economic Origins of Our Time. Beacon Press, Boston. Pryor, Frederic L. 1977 The Origins of the Economy: A Comparative Study of Distribution in Primitive and Peasant Economies. Academic Press, New York. Renfrew, Colin 1975 Trade as Action at a Distance. In Ancient Civilization and Trade, edited by Jeremy A. Sabloff and Carl C. Lamberg-Karlovsky, pp. 359. University of New Mexico Press, Albuquerque. Rice, Glen E., Arleyn W. Simon, and Christopher Loendorf 1998 Production and Exchange of Economic Goods. In A Synthesis of Tonto Basin Prehistory: The Roosevelt Archaeology Studies, 1989 to 1998, edited by Glen E. Rice, pp. 105130. Anthropological Field Studies 41. Arizona State University Ofce of Cultural Resource Management, Tempe. Robinson, George 1951 A Method of Chronologically Ordering Archaeological Deposits. American Antiquity 16:293301. Sahlins, Marshall 1972 Stone Age Economics. Aldine, Chicago. Smith, Carol A. 1972 Market Articulation and Economic Stratication in Western Guatemala. Food Research Institute Studies in Agricultural Economics, Trade, and Development 11:203233. 1974 Economics of Marketing Systems: Models from Economic Geography. Annual Review of Anthropology 3:167201. 1976 Regional Economic Systems: Linking Geographical Models and Socioeconomic Problems. In Regional Analysis: Economic Systems, Vol. 1. edited by Carol A. Smith, pp. 363. Academic Press, New York. Smith, Michael E. 1999 On Hirths Distributional Approach. Current Anthropology 40:528530. Speaker, J. Stuart 2001 Settlement and Agricultural Land Use in Ancient Mixtequilla, Veracruz, Mexico. Unpublished Ph.D. Dissertation, Tulane University New Orleans, University Microlms, Ann Arbor. Spence, Michael W. 1985 Specialized Production in Rural Aztec Society. In Contributions to the Archaeology and Ethnohistory of Greater Mesoamerica, edited by William Folan, pp. 77125. Southern Illinois University Press, Carbondale. Stark, Barbara L. 1991 Survey Methods and Settlement Features in the Cerro

de las Mesas Region. In Settlement Archaeology of Cerro de las Mesas, edited by Barbara L. Stark, pp. 3948. Monograph 34. Cotsen Institute of Archaeology, University of California, Los Angeles. 1992 Ceramic Production in La Mixtequilla, Veracruz, Mexico. In Ceramic Production and Distribution: An Integrated Approach, edited by George J. Bey, III, and Christopher Pool, pp. 175204. Westview Press, Boulder. 1995 Introduccin a la alfarera del postclsico en La Mixtequilla, sur-centro de Veracruz. Arqueologa 1314:1736. 2006 Systematic Regional Survey in the Gulf Lowlands in a Comparative Perspective. In Managing Archaeological Data: Essays in Honor of Sylvia W. Gaines, edited by Jeffrey L. Hantman and Rachel Most, pp. 155167. Anthropological Research Papers 57. Arizona State University, Tempe. Stark, Barbara L., and Christopher P. Garraty 2004 Evaluation of Systematic Surface Evidence for Pottery Production in Veracruz, Mexico. Latin American Antiquity 15:123143. Wanmali, Sudhir 1981 Periodic Markets and Rural Development in India. B. R. Publishing, Delhi. Wilk, Richard 1998 Comment on Hirth. Current Anthropology 39:469.

Notes
1. Following Hirth (1998), I make a distinction between market exchange and marketplace exchange. The former broadly refers to balanced and negotiated exchange transactions in which the forces of supply and demand are highly visible (Pryor 1977:104). Market exchanges can occur in any number of centralized or noncentralized contexts (via marketplaces, workshop procurement, or itinerant middlemen; see Hirth 1998:454455). Marketplace exchange refers more narrowly to exchanges in a marketplace setting which are spatially centralized (i.e., occur at specic loci) and temporally regularized (i.e., follow a predictable schedule). Dened in this way, one could posit that dyadic market exchanges have been occurring for millennia (contra Polanyi 2001 [1944]). However, the development of marketplaces and marketplace exchange marks the institutionalization of market exchange as a fundamental mechanism for economic transactions. I am concerned in this article with the archaeological detection of marketplace exchange and not market exchange in the broader sense. 2. This is a simplication. Biologists actually have devised many other methods of quantifying and studying diversity (see, e.g., Magurran 1988, 2004). 3. The differences between observed and expected proportions can be calculated in many other ways, such as Euclidean distances, City Block measures, Gowers similarity coefcients, and so on. One advantage of the BR coefcient is its conceptual simplicity and ease of calculation. BR coefcients can be simply calculated by hand or with Excel formulas. Kintighs BOONE program (2002) calculates a variety heterogeneity measures using BR coefcients and other techniques. Other scholars may develop other, better methods of quantifying interassemblage variability. 4. The TMP survey encompasses an area within and surrounding the modern settlements of San Juan Teotihuacan,

LAQ20(1) Garraty:Layout 1

3/3/09

3:42 PM

Page 174

174

LATIN AmerIcAN ANTIquIT"

# '&. 20, N'. 1, 2009$

San Martn de los Piramides, and several other small towns. For this reason, many scholars assume that large portions of the TMP survey area in the more built-up residential and commercial areas were inaccessible to survey. Had that been the case, the built-up TMP survey area and the rural PALM survey may not be suitable for comparison. However, as Millon (1973) points out, the TMP surveyors were frequently able to take collections within the built-up portions of the modern towns, primarily in house lots and small elds in between modern roads and structures. The PALM survey also encompassed several built-up areas (including the modern town of El Sauce), which restricted surface accessibility (see Stark 1991). For these reasons, I do not view the two survey areas as dissimilar, and, therefore, a comparison of data from the two areas is appropriate. 5. Note that the Aztec-style Black-on-orange and Blackon-red types in the Basin are distinct from the local-style Black-on-orange and Black-on-red types recovered in Postclassic collections in the lower Blanco region. 6. The minimum rim count was necessary to avoid distorting the obsidian blade-to-rim ratios. For example, a very

small collection might contain only one rim and three prismatic blades, indicating 300 prismatic blades per 100 rims. In very small collections, the likelihood of obtaining a collection with this ratio of prismatic blades to rims by chance is high. However, it is much less likely that a collection would include 10 rims and 30 prismatic blades by chance, therefore indicating a more reliable result. Inclusion of collections with fewer than 10 rims thus could potentially distort the mean and median ratios per buffer. However, also important is that some buffers only include a handful of collections. A high minimum rim count would reduce the number of collections per buffer. The 10-rim minimum is therefore a compromise between the need to avoid distorting the prismatic blade-torim ratios and the need to maintain an ample number of collections per buffer ring.

Submitted March 22, 2007; Revised July 17, 2007; Accepted August 2, 2007.

Das könnte Ihnen auch gefallen