The patterning of human behavior: A test of a prehistoric archeological subsistence and settlement model for Delaware
This dissertation presents a test of a Delaware model of prehistoric human subsistence and settlement patterns (Custer 1983a, 1984a) and a review of archeological models. The test was conducted by surveying a five percent stratified random quadrat sample of four environmental zones in the Saint Jones and Murderkill River watersheds, Kent County, Delaware. The problem which is addressed is the extent to which prehistoric human subsistence and settlement patterns and their resulting site locations can be predicted by using contemporary environmental variables and paleoenvironmental reconstructions. The research tested the following hypothesis: Assuming human adaptation to the changing environment, it is postulated that there will be concomitant changes in prehistoric subsistence and settlement patterns in the watersheds, and both these changes and the subsistence and settlement patterns can be correlated with contemporary environmental variables. Differences in numbers and locations of sites and correlations of site types with environmental variables, as predicted by a Delaware LANDSAT-generated model (Custer et al. 1986), are indicators of these changes. The null hypothesis is that: There is no correlation between prehistoric subsistence and settlement patterns and specific environmental variables as predicted by the Delaware model. The field survey tested high, medium and low probability predictions of archeological site locations based on the model. The model was determined to be 84% accurate, which is not possible by random chance. Therefore, the null hypothesis is rejected and the hypothesis is accepted. Environmental information was collected to identify any other influencing factors and recommend improvements in the model. Five recommendations for refinement of the Delaware model are offered. They are: (1) to better identify high probability areas related to marshes; (2) modify the model for the Delaware Bay Shore Zone for which the model had a low 33% accuracy rate; (3) closely analyze terraces with well drained soils as a possible predictive factor in future tests; (4) consider lower elevations (0-30 feet) as a possible predictor variable; and (5) adjust the logistical regression formula to enlarge the probability classes for high and medium probability, and reduce it for low probability.