Analysis of Bone Counts by Maximum Likelihood

Abstract

This paper describes analysis of bone counts by maximum likelihood (abcml), a new statistical method for analysis of bone counts from archæological sites. It estimates (i) the proportion of the animals in an assemblage contributed by each of several agents of deposition, (ii) the damage from such causes as gnawing by carnivores, and (iii) the number of animals represented in the assemblage. In addition to point estimates, the method provides confidence intervals and measures the goodness of fit of the model to the data. Abcml is evaluated using simulated skeletal part counts and the Hadza pattern of transport. These experiments show that the method yields accurate estimates even with severely damaged assemblages. In contrast to the minimum number of individuals (MNI), which is shown to yield severely biased estimates of the number of individuals in a data set, abcml estimates all its parameters with minimal bias. It is used to consider the hypothesis that the faunal assemblages of Gatecliff and Last Supper caves were deposited by people who transported animal parts in the manner of modern Hadza. Under that hypothesis, the data indicate that both sites are kill–butchery sites rather than home bases. The residuals from this analysis, however, are much larger than they should be. This observation makes it possible to reject the hypothesis that these sites were accumulated by people who behaved in accord with the Hadza small-animal transport model.