This study attempted to determine if behaviorally based measures were more effective than cognitive ones in producing larger associations between self-control and criminal/deviant behavior. Data were from the 1994 Oklahoma City Survey consisting of a simple random sample of adults. Various measures of self-control, based on cognitive and behavioral indicators, were compared in their ability to predict eight measures of crime/deviance. The results show that both cognitively based and behaviorally based measures of self-control produce evidence favorable to self-control theory, a finding consistent with previous research. Behaviorally based scales of self-control produce no advantage over cognitively based ones in the prediction of criminal/deviant behavior. This finding contradicts the theory that behaviorally based measures should show much larger associations with criminal behavior than cognitive type measures. It challenges the notion that if behaviorally based measures were used, predictive coefficients for self-control would be large enough to establish low self-control as the key, almost exclusive cause of offending. This data suggest that the general failure of self-control measures to predict criminal behavior to the degree envisioned by its proponents is not mainly because of the use of cognitive type measures. It may instead stem from the neglect of theoretical specifications of the contingencies under which self-control is likely to have more or less effect. Improvements in the theory itself, particularly the incorporation of contingencies, appears to offer more promise. 5 footnotes, 7 tables, 56 references