The 122 psychology student subjects were grouped according to reported trait anxiety as measured by Lykken's Activity Preference Questionnaire and then engaged in either a mock crime or a neutral activity. The 'crime' gave subjects 'guilty knowledge,' and the neutral task provided 'innocent association' to relevant stimuli imbedded in each scenario. Subjects were then interrogated using Lykken's Guilty Knowledge Technique. Anxiety classification, guilty vs. innocent treatment, and type of stimulus were manipulated by a 2 x 2 x (2) factorial design. Results for differential responsivity scores showed significant main and interaction effects. Low-anxiety subjects showed almost no effects, but highly anxious subjects responded strongly in the 'guilty' rather than the 'innocent' treatment. Contrary to previous findings, results showed a substantial but qualified relationship between self-reported anxiety and electrodermal responsivity. In addition, the Guilty Knowledge Technique with rank scoring yielded 97.5 percent correct classification. It is suggested that information on a subject's self-reported trait anxiety would be useful to field polygraph operators and to researchers. Two tables,1 reference note, and 15 references are provided. (Author abstract modified)