Previous international studies of science achievement put the data through a process of winnowing to decide which variables to keep in the final regressions. Variables were allowed to enter the final regressions if they met a minimum beta coefficient criterion of 0.05 averaged across rich and poor countries alike. The criterion was an average across all countries because the original idea was to identify those variables which were thought to be "important" across all societies taken together. The question is whether this process tended to omit school variables which may have had strong effects within one country but not across the average of many countries. We have re-entered variables for each country separately, using the same minimum entry criterion of 0.05 as before, but within each country separately. Using only variables which have been found to be "important" in that particular society, we get different results, and the changes are of two sorts: they tend to significantly increase the variance explained by school effects, and this increase tends to be greatest in the "poorer" countries. (Author)