Objects were assigned to groups on the basis of highest absolute factor loadings, with the minimum loading required for assignment systematically varied. Rotational methods did not differ significantly in either accuracy or coverage of the resulting classifications. Paradoxically, setting the number of factors equal to the number of underlying populations resulted in less accurate solutions than determining the number of factors empirically by Cattell's screen test. The inverted factoring technique was found to be as accurate as the best hierarchical clustering algorithms previously tested on these mixtures. Thus, inverted factor analysis appears to be a useful taxonomic tool. Tables, figures, and about 50 references are provided. (Author abstract modified)