Site-level Data vs. Regional-scale Survey Data for Cultural Analysis in the Central Andes Presented at the 64th annual meeting of the Society for American Archaeology, Chicago, March 1999. Read by Cristiana Barreto.

Alvaro Higueras, University of Asmara. Presented at the General Symposium "Archaeological Landscapes."

To papers list ......... To bookshelf

Today I will review three specific concerns I have on settlement survey methodology in the Central Andes. I hope my remarks on research cases in this region will match themes in this general session, and will be useful for researchers who strive for comparative examples for studying prehistoric landscapes.

The Central Andes region is located on the mountain chain lying on the western edge of South America, in today's countries of Peru and Bolivia. Three topographic settings dominate this region: fertile and deltaic valleys cross-cutting a desertic coastline; highland valleys bounded by higher ridges at an average elevation of 2500 meters above sea level; and a plateau or "altiplano" at about 3800 meters, which includes Lake Titicaca. The archaeological cases I will refer to are set in coastal and highland valleys.

The first modern settlement pattern survey in archaeology was carried out by Gordon Willey in the coastal valley of Viru in the mid 1940s (Willey 1953; Figure 1). However, innovations in surveying techniques and concepts were made elsewhere (Ammerman 1981:66). Survey research in this region was framed in orthodox full-coverage survey methdologies (e.g., Wilson 1988). Similarly, use of exploratory tools for analyzing settlement patterns has not been common, even if conducting full-coverage surveys. Recent survey research in the Tiwanaku Valley has systematically used spatial tools for exploring Tiwanaku settlement patterns (Albarracin-Jordan 1992, 1996; McAndrews et al 1997). But other problems remain in survey methodology.

Here I argue (1) that it is often difficult to use surveys made in contiguous areas when provided with scarce information on methodology and data; (2) that the relevance of site-level data for regional problems may be of temporary value while a regional analysis is not available; (3) that little consideration is made of applying survey methods other than full-coverage strategies; and, (4) that a full-coverage survey strategy does not insure that data analysis is made with regional considerations in mind.

I will review survey and site-level research in the Mantaro, the Cochabamba, and the Moquegua valleys to illustrate the above points. These regions belong to highland settings, except Moquegua which is a coastal valley. In Mantaro, the upper valley was studied by Earle and team in the early 1980s. The Mantaro middle valley, was surveyed by Browman in the late 60s (Browman 1970). In Moquegua, Goldstein studied the middle valley site of Omo in the late 1980s. And in Cochabamba I recently studied two small valleys surrounding the main Central Valley.

Two of the above cases, Earle's et al. Mantaro and Goldstein's Omo research, are obligatory case studies for Andean archaeologists. They stand among the most productive and influential research in the Central Andes. The former because it produced outstanding monographic research on political interaction between the Inka polity and local populations. The latter for it documented in Omo a Tiwanaku ceremonial/administrative settlement of the altiplano Tiwanaku polity. This site has proven crucial to understand regional interaction patterns of the Tiwanaku polity.

The third case of Browman's Mantaro survey is mostly cited in MacNeish et al.'s (1975) volume on interaction spheres in the Central Andes . Finally, my Cochabamba research is not well known, as a synthesis of it remains unpublished (an account of the research is posted at Its results are less widely accepted, as I used an unorthodox survey strategy (i.e., random sampling), and my conclusions differ from most popular assumptions that suggest Tiwanaku populations were colonizing and restructuring agricultural production in the Cochabamba region. This conclusion challenges a traditional state/imperial character attributed to the Tiwanaku polity.

FIRST POINT: heterogeneous survey methodologies preclude building larger regional data sets.

An ideal scenario for researchers in a region is to see a continuous growth in settlement data gathered in a systematic and controlled fashion. Not all survey projects aim at covering the complete area of a valley or natural region as Wilson achieved in the Santa Valley. It is then positive to have research in previously unsurveyed portions of a region, or have more extensive surveys in previously surveyed areas. A case in point is West's Viru valley study, after Willey's "intensive sampling survey" as Parsons put it. The case in the Mantaro region is about this: Earle's at al. survey coming after Browman's survey expanding the data set of settlement distribution for the valley. However, differences in defining site size, site boundaries, and distinct mapping codes, produce data sets that are not fully complementary. I came across this observation as I was planning research in the southern Mantaro valley. I deemed it advantageous to have previous survey data for the region: they provided a relevant research problem. And further survey would increase the settlement data set for the region.

Earle's et al. and Browman's full-coverage surveys overlapped in about 200 km2 of the Mantaro valley. Earle et al. have published several maps of settlement distribution, but they often emphasize the northern Yanamarca portion of their survey. The most detailed map by D'Altroy (1994) shows 80 sites recorded for the Late Horizon in the overlapping area. Fifty sites are local Wanka settlements contemporaneous to thirty Inka sites. Of these latter, 27 are storage sites. Sites are plotted as circles, where circle size is adapted to site size ranges. Absolute site size estimations are not provided. In this same area, Browman records 29 Inka sites, that include nine storage sites. The difference in storage sites is important despite a close figure for total sites recorded (Figure 2). Browman argues that storage sites also have non- conspicuous domestic evidence; Earle's et al. storage sites would also have had habitational functions which were not identified. In contrast to Earle at al., Browman does not consider occupation of local sites for the Inka period. Browman's maps consist in point-plotting sites, with data on UTM coordinates for a site's grid. No site size estimations are provided either.

Differences in site identification are broader for the Late Intermediate Period. Hastorf (1993), as part of Earle's team, maps sites for the two consecutive phases of this period: 21 and 7 sites, respectively, for a number of 22 sites, as six sites are occupied for both phases. Browman reports 42 sites for this period. The difference in site number may be due to principles of defining a site. In several cases, Browman's clusters of several adjacent sites seem to match the blobs shaping site boundaries in Hastorf's maps (Figure 3). Visual comparison of Hastorf and Browman maps show occupation of the same clusters, but differences in the number of sites plotted; it also shows gaps in occupation in Hastorf's maps.

In a sense, maps in Earle's team publications need not be so precise given that spatial analysis of settlement distribution, despite having conducted a full-coverage survey, was not an objective of the project. However, it precludes spatial analysis of this data by other researchers. On the other hand, while plotting all his sites, Browman produced little spatial analysis beyond visual identification of shifts in clusters occupation. He identified an exclusive concentration of Inka occupation to the northern end of the valley, an important fact studied later on by Earle et al.

SECOND POINT: site-level analysis is of temporary value until survey data provides regional scale data.

The site is a commonly studied unit of analysis in archaeological research. The wealth of information available on domestic, elite, and workshop activities in a site-level study makes of it a very attractive research unit. Whether the rationale for analyzing a specific site is based on artifact evidence on surface, on architectural remains, or, in a best case, on a comparison of a set of sites recorded in a initial survey stage, is crucial for the scope of conclusions derived from a site-level study.

Therefore, it is problematic when site-level data are used to suggest regional patterns of socio-political organization. Are architectural, mortuary, and portable elements found in an important, yet single, site sufficient to assess its position in a regional hierarchy? Or its administrative and political affiliation towards the core site? Or its role towards smaller, local sites, or Tiwanaku sites in the hinterland? These are questions that apply to the role of Omo, the Tiwanaku administrative site in Moquegua.

This does not imply that the colonial or administrative status attributed to Omo is not accurate. It just means that its position in a regional hierarchy and the scope of its regional control need to be supported with regional data. A regional survey of the lower Moquegua Valley failed to find any Tiwanaku occupation. Reports by Goldstein on survey data from the Middle Valley do not mention shifts in Omo's role or in the site hierarchy. In contrast, recent survey of the upper Moquegua valley by Owen should produce new settlement data that may not strip Omo of a paramount role, but will help construct a more comprehensive regional hierarchy. A good example of how regional interpretations vary with site-level or with regional data is available for the Nazca polity in Peru's South Coast (Silverman 1993).

THIRD POINT: is there room for applying other survey methods than full-coverage strategies?

The strategy of random sampling survey applied in the Cochabamba region is not common in the Andes. The survey strategy was aimed at recording the spatial distribution of settlements in relationship to particular topographic and soil zones; these latter define spots of agricultural productivity. A sampling strategy was ideal to cover an extensive survey universe needed for this human-land relationships analysis. This survey strategy is not aimed at finding all a region's sites or its most important site, but rather at recording the range of correlations between agricultural potential and location of most common sites, such as villages and homesteads.

The research in Cochabamba tested the nature of political and economic interaction between altiplano and Cochabamba populations during Tiwanaku times (from AD 400-1000). It is commonly assumed that the altiplano Tiwanaku state established colonies in Cochabamba to ensure access to warm-land crops. To test this hypothesis I predicted that either external or local populations in this period, concerned with increasing surplus of agricultural production, would settle on the best soils, or, in areas of higher agricultural productivity in the region. Two areas with contrasting agricultural potential were selected for sampling survey; studied under a same methodology, these areas produced comparable samples for spatial analysis. This method, as suggested by Struever, allows definition -and comparison- of the selective pressures impinging on the settlement system. In addition, with a diachronic comparison human-land patterns of one period are measured against prior settlement distributions.

How was sampling data used for documenting shifts in occupation size and in location between regions and through time? Occupation size by period and by environmental strata, based on field estimation of site size, are key for the analysis. Occupation size produces more significant comparisons than using the number of sites recorded. The sampled occupation size for each strata was used to estimate occupation size for the total strata in the survey area, at 95% confidence levels. These total occupation estimates were then standardized by using percentages of occupation of each topographic and soil zone, and plotted in bullet graphs, as used for representing C14 dates. Environmental preferences of human occupation are defined by departures from a random pattern, when occupation is closely proportional to the area of each zone. The zone with the highest percentage of its area occupied had preferential occupation provided that its error ranges do not overlap the estimated percentage of any of the other two zones. This analysis allowed definition of significant temporal shifts in occupation of soil and topographic zones within a region and between regions.

Survey sampling is not a popular approach, and the Central Andes are no exception. Sampling survey is, however, a powerful strategy for addressing research questions requiring the study of large regions. Proponents of full-coverage survey have hammered the utility of the strategy for "completeness" of the data. However, research questions relating to human-land relationships may often need larger universes than what is available in a contiguous area. Sampling survey strategies allow therefore analysis of a broader range of correlations between human occupation and features such as soil diversity and topographic variation. But more importantly, sampling allows for statistically controlled inferences made on observed correlations to be extrapolated to entire populations, whereas sampling problems in "full-coverage" surveys are often disregarded.

FINAL POINT: not all settlements surveys are aimed at producing regional analysis.

Survey strategies have an overall objective: to produce a systematized and controlled record of settlements and environmental features in a pre-defined region. Research questions guiding the survey can vary widely, and therefore the resulting data set is handled differently. Whichever the ultimate objective of surveying, a standard result should be explicit methodology, full data descriptions, and adequate maps. If this occurred, perhaps comparison of survey data made in adjacent regions could be more accessible for secondary analysis.

Survey can be crucial for locating appropriate sites for site-level excavations. Earle et al. and Goldstein, having started their research on regional and site-level scales, respectively, share a same spatial unit of analysis in the archaeological site. Earle et al. conducted a full-coverage survey aimed at obtaining a complete record of sites in order to do a proper selection of sites for excavation purposes. No spatial analysis was attempted with the survey data. Goldstein, on the other hand, started his analysis at the site of Omo, identified it as a paramount site in Tiwanaku's realm, and then moved to conduct survey in Omo's hinterland.

Survey research could be, on the other hand, directed at spatial analysis with questions needing documentation of site distribution for exploring spatial relationships between sites, for producing a site hierarchy, or for defining relationships of site location vs. environmental features. This diversity in research questions and objectives should generate a greater flexibility in choosing fitting survey strategies for regional settlement analysis.

References cited

Albarracín-Jordan, Juan
1992 Prehispanic and Early Colonial Settlement Patterns in the Lower Tiwanaku Valley, Bolivia. Unpublished Ph.D. dissertation, Department of Anthropology, Southern Methodist University, Dallas.
1996 Tiwanaku Settlement System: The Integration of Nested Hierarchies in the Lower Tiwanaku Valley. Latin American Antiquity 7:183-210.

Ammerman, Albert
1981 Survey and Archaeological Research. Annual Review of Anthropology 10:63-88.

Browman, David
1970 Early Peruvian Peasants: The culture history of a central highlands valley. PhD dissertaion, Department of Anthropology, Harvard University.

D'Altroy, Terence
1994 Factions and political development in the central Andes. In Factional Competition & Political Development in the New World, E. Brumfiel and J. Fox, eds., pp. 171-187. Cambridge University Press.

Hastorf, Christine
1993 Agriculture and the Onset of Political Inequality Before the Inka. Cambridge University Press, Cambridge.

MacNeish, Richard, Tom Patterson and David Browman
1975 The central Peruvian prehistoric action sphere. Papers of the R.S. Peabody Foundation for Archaeology, vol. 7.

McAndrews, Tim., Juan Albarracín-Jordan, and Marc Bermann
1997 Regional Settlement Patterns in the Tiwanaku Valley of Bolivia. Journal of Field Archaeology 24:67-83.

Silverman, Helaine
1993 Cahuachi in the Ancient Nasca World. University of Iowa Press, Iowa City.

Willey, Gordon R.
1953 Prehistoric Settlement Patterns in the Viru Valley, Peru. Bureau of American Ethnology, Bulletin 155, Washington D.C.

Wilson, David
1988 Prehispanic Settlement Patterns in the Lower Santa Valley, Peru: A Regional Perspective on the Origins and Development of Complex North Coast Society. Smithsonian Institution Press, Washington, D.C.


To papers list ......... To bookshelf

Tiwanaku and Andean Archaeology
Copyright © 1998-01, Alvaro Higueras. Derechos Reservados/All rights reserved.
Please send comments on content and presentation to
URL of this document:
Revised: 29 May 1999.