Multivariate Approaches to Site Location on the Northwest Coast of North America

Article excerpt

The American Northwest Coast, famously rich as an environment for hunter-gatherers, is not an easy landscape to inhabit. Field survey of the Tebenkof Bay region, Kuiu Island, southeast Alaska, identifies the pattern of site positions. Mathematical modelling explores which considerations directed the placing of settlements in that landscape.


For over 30 years one of the critical issues in archaeological method and theory has been the development of techniques to model and characterize prehistoric site locations (Cart 1985b; Clark 1968; Hodder & Orton 1976; Kohler & Parker 1986; Kvamme 1985). Archaeological methods to accomplish this task have centred primarily around the relationship between food resources and seasonality, and settlement locations (Kvamme 1985; Limp 1991). Statistically, a number of methods for the construction of site location models have been described; logistic regression has been shown to be the most successful (Kvamme 1985; Warren 1990a). Logistic regression procedures have a firm foundation in the archaeological literature (Kvamme 1983; 1989; Warren 1990a), and a recent overview by Warren (1990b) provides the necessary background adequately to evaluate the procedure's use in the context of archaeological analysis. Here we present an example of a logistic regression analysis that is both robust and open to field evaluation and statistical replication. Yet certain questions are left unanswered by logistic regression analyses, so we address the structure of the relationships among the site location variables through log-linear modelling. These methods are applied to the Tebenkof Bay region of the northern Northwest Coast of North America, the traditional homeland of the Tlingit [ILLUSTRATION FOR FIGURE 1 OMITTED].


Site location models have played only a minor role in the history of archaeological research on the Northwest Coast. Hobler attempted to find correlations between distance to salmon streams and the location of archaeological sites, but with little success (1982). Acheson, surveying sites in the Kunghit Haida region, found they tended to be on the exposed outer coast but in places that were sheltered from the main surf by micro-geographic characteristics. He also recognized that habitation sites were not near salmon streams and had very poor water sources (1991).

C. Rabich-Campbell (1980) made the first attempt to isolate important environmental determinants of site location in southeast Alaska. From a univariate approach, she recognized a tendency for sites to have southerly exposures, good beaches, ice-free zones, protection from southeasterly storms, and fresh water (1980: 17). Yet there was a considerable amount of variability in the data presented for her sample of 24 house-depression villages.

From early ethnographic accounts we can identify some variables that influenced where the historic Tlingit placed their settlements. These include southerly exposures for winter warmth, sheltered coves to protect houses and canoes (Krause 1956), fresh water, beaches of pebbles and small rocks so as not to damage the cedar canoes (Niblack 1970), level and well-drained terrain and defensibility (Emmons 1991). As would be predicted, proximity to food resources is also important (Krause 1956; Langdon 1979; Niblack 1970; Oberg 1973).

In order to investigate the decision-making processes that influence site location, a complete archaeological survey of Tebenkof Bay, Kuiu Island, southeast Alaska was conducted as part of a brooder study to investigate the development of ranked hunting and gathering societies on the northern Northwest Coast (Maschner 1991; 1992). A goal of the Tebenkof Bay Archaeological Project was to build a statistical model of site location, based purely on environmental variables, in order to establish a screen on which to project social and political processes. Over 150 sites were found on the Tebenkof Bay survey, of which 94 were identified as villages or temporary habitations based on the presence of shell middens [ILLUSTRATION FOR FIGURE 2 OMITTED]. …