The general purpose of this study was to determine the effects of similarities between stems and keyed choices on test difficulty. Unlike previous investigations of this undesirable characteristic of some multiple-choice items, the present study employed items that were unintentionally faulty and samples of examinees who were highly experienced test-takers. The first phase of the study indicated that the presence or absence of similarities did not significantly affect examinees' scores. The second phase, however, indicated that examinees easily may be trained to recognize and use such similarities to advantage. (Author)