An analysis and application of the size distribution of waste flakes from the manufacture of bifacial stone tools
Stahle and Dunn (1982) presents a method of anaylsis that can be used to interpret stages of biface reduction present in a debitage assemblage. This article pays particular attention to devising ways to use frequency distribution of waste flake size categories as an analogy for unbiased samples stemming from biface reduction and estimate how many stages of biface reduction are present in a particular debitage sample through comparison with experimentally replicated flakes. The hypothesis is that, as waste flake size decrease from initial to final stages in biface production, systematic changes in flake size can be used to identify stages of reduction in anonymous debitage samples through comparison with experimental assemblages.
Methods and Materials
To begin, the authors experimentally define four stages of biface reduction: cortex, bedding plane, and imperfection removal; heat treatment; bifacial flaking, and bifacial shaping of the margin and hafting area. All debitage was seived through ten screens ranging in size from 1/8 - 2 inches with 1/8th inch intervals. Both broken and whole flakes were included in analysis. Distributions of flakes based on reduction stage followed assumed frequencies. Next, the authors found theoretical distribution that would most accurately model flake size distribution to smooth the data and remove bias. Out of the mathematical models tested, the Weibull distribution was determined most accurate, as it’s estimated values produced the highest correlation among the samples and the coefficiet of determination for debitage over all stages was 0.9950, the highest of all models. To estimate the relative contirbution of each stage of production to the debitage pile, Stahle and Dunn apply least squares analysis of their data. Their logic was that if each stage is defined by the curve that represents its cummalitive distribution function (cdf), then plotting the cdf of an unknown flake collection could show what stage of biface reduction produce a standard curve most similar. They tested this with samples from two stages of biface reduction and plotted the results with least square analysis. The median values suggest that more than half of the data is closer than 0.310 to the true proprtion, indicating a relatively high correlation between estimated and expected flake size proportions.
Stahle and Dunn’s experiment shows that a Weibull distribution is the most accurate model for predicting flake size distributions of debitage from biface reduction. The experiment also agrees with the four reduction stages defined earlier and determines that least square analysis can be used with relatively high accuracy to estimate which stage of reduction a particular debitage sample results from.
This article is only one of a number focused on finding empirical ways to confirm visual or logical information. In archaeology, I’ve found that this is one of the hardest factors of the discipline to reconcile, so these articles, while usually mathematically over my head, are really helpful. In particular, Stahle and Dunn’s article is reminiscent of Spaulding’s (1953) paper on statistical techniques for finding meaningful characteristic typology in artefact assemblages. Like Stahle and Dunn, Spaulding finds a useful statistical model and compares data from samples and experiments. However, Spaulding, more than Stahle and Dunn, find that the statistical methods require a somewhat exhaustive process and are not particularly useful on a smaller scale.
This article is very dense and mathematical. Without a complex math background, I was unable to understand most of the reasoning in this article, and had to assume that their calculations were sound. Overall, the general ideas behind the math seem accurate to me. I think this article is a good example of using calculations and empirical data to verify visual information; that is, logically it makes sense that smaller flakes result from a later stage of biface reduction, but this article does a good job at giving an empirical reason why.
Spaulding, A. C. (1953). Statistical Techniques for the Discovery of Artifact Types. Society for American Archaeology, 18(4), 305–313. Stahle, D. W., & Dunn, J. E. (1982). An analysis and application of the size distributio of waste flakes from the manufacture of bifacial stone tools, 14(1), 84–97.