| Most archaeological arguments assert a connection between typological units and some inferred property. In the case of functional inferences in historical archaeology, two approaches dominate Alaskan archaeology: one proposed by Catherine Spude, another by Rick Sprague. Spude's model assumes normal or multivariate normal distributional patterns, but this is shown to be unsupportable. Using the data she upon which constructed her family assemblage model, goodness-of-fit tests show that a majority of her cases fail to match model expectations.; Sprague's approach is based on an assumed functional connection between the scale of an individual object and that of aggregated objects, an assemblage. If his classification monitors function at the assemblage level, a series of expectations of within- and among-group variability can be deduced. Checking the fit between observations and expectations requires an independent control of function. To this end, Twentieth Century assemblages from Fairbanks, Alaska are used for the controlled cases. For these, documentary and photographic information are used to determine functional similarity and differences among locations. Assemblages associated with these known circumstances are used to see if expectations are met when examining frequencies of Sprague's functional classes. Within- and among-group relationships at primary and secondary classificatory scales are examined using categorical and multivariate non-parametric techniques. The analyses show that his classification does not identify function. It is argued that the reasons for its failure stem from non-linear patterns within the data. This supports the idea that any single monolithic typology is incapable of diagnosing function. |
