Studies of criminal decision making commonly rely on college students’ self-reported intentions to commit a hypothetical offense. The current study evaluates the predictive validity of these intentions to offend.Undergraduate students (n=726) read a fictitious but seemingly realistic newspaper article describing an illegal opportunity for acquiring digital music files, and then reported their intentions to act upon the opportunity. Afterward, participants’ real world attempts to follow-through on the opportunity were monitored covertly.Findings reveal that participants who reported weak intentions to offend typically refrained from the act, resulting in a low false negative rate. However, those who reported strong intentions to offend also typically refrained from the act, thereby resulting in a high false positive rate.These findings suggest that while participants’ predictions of criminal abstention are generally accurate, their predictions of criminal involvement are more problematic. Such faulty intentions have important implications for research on criminal decision making.