SAMPLING ORGANIZATIONS AND GROUPS OF UNEQUAL SIZES.

this model is taken as accurate, it leads to the conclusions that occupational and educational rank are negatively related to symptom level, that racial-ethnic rank is positively related to symptom level, and that all forms of sharp status inconsistency produce more or less equivalent increments in symptom level. The second alternative model, Model II, was based on the hypothesis that racialethnic status per se has no effect on symptom level. It included terms for the additive effects of occupation and education and terms for effects of sharp inconsistency, but not for additive racial-ethnic effects. Like Model I, this analysis leads to conclusions that educational and occupational rank are negatively related to symptom level and that sharp occupation-education inconsistencies increase symptom levels. Unlike the previous model, Model II indicates that sharp (and perhaps moderate) inconsistency between high ascribed status and low achieved status has a much greater impact on symptom level than the converse form of discrepancy. Of course, these two alternate models fall far short of exhausting the explanatory possibilities.22 But they do constitute two plausible hypotheses for further test. What are the grounds for choosing between them at the present time? Model I, involving all three status dimensions and four types of sharp inconsistency, accounts for somewhat more variation than Model II, at the cost of adding two terms to the prediction equation. We prefer Model II, however, because we find the assumption that inconsistency between high ascribed and low achieved status is especially likely to lead to a symptomization response, theoretically more palatable, for reasons explained in the earlier report,23 than the assumption that high racial-ethnic status is itself positively related to symptom level. A last conclusion is the the method of analysis used herein should be admirably suited to much sociological research, in which low levels of measurement and an interest in interaction are frequently combined.24