PART IV

THE MIDDLE PALEOLITHIC

The first of these three chapters treats the origins of the Mousterian and shows that well-excavated assemblages can and do intergrade. For that reason and others, the interpretation of the Mousterian facies as non-overlapping, mutually exclusive sets of related industries can no longer be maintained, nor can the idea that they were the stylistically distinctive products of separate, identity-conscious socio-cultural groups. In Chapter 9 I present some of the evidence suggesting that there are differences between Middle and Upper Paleolithic adaptations and speculate about their causes. Chapter 10 attempts to summarize still more evidence about those differences and to indicate some of the research errors we have committed in the past.

Among the techniques used to reach these conclusions was a pair of multivariate tests: the Kolmogorov-Smirnov two-sample test and a principal components analysis based on rank-order correlations, whose solution was subjected to Varimax rotation. My use of these techniques has been criticized for reasons suggesting that the critics lacked mathematical sophistication. One objected that the Kolmogorov-Smirnov test needed to “take account of sample size,” but that is something that is built into its formulas.

The idea that discrepancies in sample size can overdetermine correlation has led others to use a variety of means to eliminate these discrepancies, such as data transformations, when the best such means has always been the use of rank-order correlation. I have used Spearman’s r rather than the Pearson’s r statistic for correlation, since the two coefficients are closely related and Spearman’s coefficient is part of every major statistical package for home computers. As it happens, the concern is more theoretical than real, and results using discrepant sample sizes and Pearson’s r are not that different from mine. (Having in fact used both techniques, I am sure of my ground.) Last, some people still fail to understand that rotation does not really alter the Principal Components solution in any way; it just spins its axes to a position that makes the solution easier to grasp without considerable effort.