Some considerations regarding the use of counts vs. percentages are presented in light of criticisms of the author's previous work voiced in a paper by Freeman, Brown, and Thomas at the 1968 meeting of the Society for American Archaeology. The controversy concerns the effect of using percentages as opposed to raw counts in calculating Pearson's r, the coefficient of linear correlation. The author agrees with his critics that the closed array problem can create spurious correlation when percentages are used, but maintains that use of raw counts on assemblages of differing sizes leads to inflation of positive correlations. An example is presented in which the latter phenomenon is illustrated. The author goes on to suggest that part of the solution to these problems is to consider using coefficients other than Pearson's r in comparing assemblages. The point is illustrated by brief discussion of a seriation of pottery types in 1330 Argentinian gravelots. A coefficient of similarity based on the number occurrences of two pottery types together divided by their total occurrence was used to obtain a multidimensional scaling seriation which agreed quite well with radiocarbon dates. Seriations of the same data with correlation-based techniques introduced varying degrees of distortion, and were judged inferior.