Compared the traditional moderated regression technique with a technique designed to increase the probability of indication of a moderator variable. Results indicated that detection of moderator variables is dependent on their strength. A higher probability of detecting a moderator exists if the interaction is entered into the regression first. (WAS)