Problems of identification and quantification in archaeozoological analysis, Part I: Insights from a blind test

In archaeozoology, counts are generally considered as replicable data that accurately represent the initial abundances of elements, individuals, or taxa, although perhaps only at the ordinal scale. However, few studies have tested these assumptions with control data. To improve our knowledge of these issues, we conducted a blind test that involved the analysis of two large experimental samples composed of modern ungulate specimens of known element and taxon. Because the samples differed in level of fragmentation, the blind test provides substantial information on the impact of bone processing on faunal identification and quantification. Our results suggest that Number of Identified SPecimens (NISP) and Minimum Number of Elements (MNE) provide measures of abundance for whole assemblages and for samples limited to non-long bones that are both replicable and accurate at the ratio scale. However, the same metrics generally failed, even at the ordinal level, to predict abundances in analyses restricted to long bones and long bone portions. Given these mixed results, it seems judicious, in agreement with the current majority view among archaeozoologists, to treat faunal tallies as ordinal-level information. Despite issues of reproducibility and the difficulty of aggregating counts with MNE, the blind test also indicates that this measure is more robust at predicting skeletal abundances than NISP. Substantial variations in rates of long bone fragmentation and identification probably explain the poorer performance of NISP in the blind test