Abstract: | Comparisons of factor analytic results based on a matrix of total (individual) correlations and three matrices of between-groups (ecological) correlations revealed fundamental differences in factor patterns when the unit of analysis shifted from individuals to aggregates. The major difference involved the marked tendency for a highly individualistic General Ability Factor derived from the matrix of total correlations to become more and more sociological in content as individuals were aggregated. Results are discussed in terms of their bearing on the problem of determining the "appropriateness" of a given unit of analysis, individuals vs. aggregates, for certain levels of inquiry, i.e., psychological vs. sociological. It is concluded that whether the researcher should heed the traditional warning about the use of aggregate data relationships as a basis for making inferences concerning individuals ought to depend in part on whether his interest is in psychological or sociological propositions about indivduals. |