Use of Cluster Analysis with Anthropological Data

A LITTLE more than twenty-five years ago I published, with two coauthors, a description of a statistical device applied to the analysis of certain ethnographic data from Polynesia (Clements, Schenck and Brown 1926). So far as I know this was the first attempt to use statistical methods in cultural anthropology since the efforts of Hobhouse and his associates (Hobhouse, Wheeler and Ginsberg 1915, and the work of Czekanowski (1911) on certain African tribes. They, of course, were antedated by Tylor's (1889) classic endeavor. Our 1926 paper was attacked by W. D. Wallis (1928) who took issue both with the method of analysis and certain historicoethnographic conclusions we had drawn. In my reply to these criticisms, I availed myself of the opportunity to develop a number of points which had not been adequately covered in the original paper. At this time, I specified several principles which I still believe are fundamental to the proper use of statistical analysis of anthropological data (Clements 1928). These are: 1. The traits used must be either all the traits in the statistical universe