Abstract
This article provides a commentary on Bryant MacLean, and Bradley (1990). It is argued that their results do not differ greatly from those reported by Bowey and Patel (1988), with the discrepancies in the outcomes of multiple regression analyses reflecting the relative size of simple correlations, which are probably attributable to differences in the design of the two studies. These differences can only be settled by future replication work. Methodological suggestions are made for maximizing the generalizability of such work.