Replicability in Lithic Analysis

Abstract The ubiquity and durability of lithic artifacts inform archaeologists about important dimensions of human behavioral variability. Despite their importance, lithic artifacts can be problematic to study because lithic analysts differ widely in their theoretical approaches and the data they collect. The extent to which differences in lithic data relate to prehistoric behavioral variability or differences between archaeologists today remains incompletely known. We address this issue with the most extensive lithic replicability study yet, involving 11 analysts, 100 unmodified flakes, and 38 ratio, discrete, and nominal attributes. We use mixture models to show strong inter-analyst replicability scores on several attributes, making them well suited to comparative lithic analyses. Based on our results, we highlight 17 attributes that we consider reliable for compiling datasets collected by different individuals for comparative studies. Demonstrating this replicability is a crucial first step in tackling more general problems of data comparability in lithic analysis and lithic analyst's ability to conduct large-scale meta-analyses.