Moderated regression is a commonly used technique within the behavioral sciences. The power of such analyses, however, is dependent on the strength of the moderator relationship and the distribution of the moderator variable. This study compares the traditional moderated technique with a technique designed to increase the probability of the indication of a moderator variable. The results indicate that, often, the detection of moderator variables is not so much dependent on their existence but, rather, is dependent on their strength. The results also indicate that the ability to detect moderators also depends on the distribution of the moderator variable. Finally, the results indicate that a higher probability of detecting a moderator exists if the interaction is entered into the regression first.