Two factors affecting multi-element quantification of commingled human skeletal assemblages

Skeletal quantification is critical to the analysis of skeletal assemblages. The use of multiple skeletal elements to estimate numbers of individuals can increase viability of estimation methods. Quantification, however, is both necessitated and constrained by skeletal attrition. Two factors involved in skeletal attrition that are identified here are independence of recovery of elements within the skeleton and the interaction between specific physical properties of skeletal elements and their recovery probabilities. The hypotheses tested were that (1) dependence of recovery of skeletal elements would be correlated to anatomical proximity and (2) element-specific recovery rates would be correlated with bone mass, length, and mineral density. Skeletal attrition is viewed here as a probabilistic process. Skeletal inventories of 15 appendicular skeletal elements were collected from the Forensic Data Bank (FDB), CSU, Chico Human Identification Lab (HIL), Gibson Md. 2 from the Lower Illinois River Valley (LIV), Santa Clara Valley Medical Center historic cemetery (VMC), and unidentified forensic anthropology cases from the New York City Office of the Chief Medical Examiner (NYC). Mass and length data were collected from non-overlapping cases from the HIL and NYC as well as the CSU, Chico anthropological teaching collection. Bone mineral density values were drawn from Kendell and Willey (2014). Independence of all possible 2-element combinations (n = 600) was tested for FDB, NYC, HIL, VMC, and LIV using Holm’s adjusted Fisher’s exact tests. Anatomical patterns were uncovered using Wilcoxon rank sum tests of phi values from all but VMC. Spearman’s correlation was used to test relationships between element-specific recovery probabilities and mass, length, and mineral density for NYC, HIL, VMC, and LIV. The results indicate that there is a relationship between anatomical proximity and independence of recovery. When selecting elements for inclusion in assemblage size estimation it is therefore advisable to choose elements on both limbs and avoid those that are attached at a joint. This will maximize agreement between actual and expected degrees of freedom, generating a reliable estimate of confidence in assemblage size approximation. Testing of hypothesis 2 showed a distinct pattern in significant skeletal properties for each dataset. This means that there is no one most important property for determining patterns of attrition, but rather that the most important property is the case-by-case result of taphonomic variables. This study indicates the need for further testing using collections affected by different taphonomic conditions and with a wide range of assemblage recovery rates.