Many deterrence researchers have based their research on the assumption that if persons commit crimes because they have not been deterred and if persons refrain from crimes because they have been deterred, then those who commit crimes tend to perceive punishment as less certain or severe than those who obey laws. In reaching this conclusion about the deterrent properties of perceived risk, many deterrence researchers have used cross sectional designs, correlating current perceptions of punishment risk with self-reports of past behavior. Negative correlations between current perceptions and prior behavior have been uniformly interpreted by deterrence researchers as a deterrent effect; however, low perceptions of punishment risk or severity may be a consequence rather than a cause of involvement in illegal behavior. The study reported here examines the association between perceptions of the certainty-severity of punishment and subsequent behavior, controlling for extralegal conditions. Three hundred college students were selected randomly from a list of freshmen and interviewed at time periods 1 year apart, using a scale that measured deterrence, social bonding, and social learning. Involvement in five specific offenses was self-reported. Findings showed that when other constraining factors were considered (beliefs, informal sanction, attachment to parents, attachment to peers, etc.), the deterrent effect disappeared. Future research should focus on replicating these findings with more representative samples and with different offenses. Tabular data, figures, 19 footnotes, and 38 references are provided.