The research reported here is based on a comparison of active residential burglars and a matched control group regarding their willingness to commit a burglary at varying levels of certainty of arrest, severity of penalty, and anticipated reward. Initial analyses revealed that few controls were willing to offend regardless of risk, penalty, or reward and that offenders were not influenced by penalty on its own. Consequently, responses of the oflenders only were further analyzed in relation to the impact of risk, penalty, and reward. The results of a logit analysis indicated that both risk of being caught and prospect of increased gain had a significant influence on the offenders' decision making.Deterrence is essentially a psychological process that involves the balancing of personally held beliefs about possible punishment and anticipated gain. Strictly speaking, then, there can be no direct relationship between sanctions and criminal action; the two must be linked through the intervening variable of subjective perceptions of the risks and rewards of committing an offense (Beyleveld, 1980; Gibbs, 1975). Perceptual deterrence research typically investigates the effect of perceived certainty and severity of punishment in preventing criminal behavior. However, this research suffers from a dearth of evidence about known offenders (Paternoster, 1987). In addition, the deterrence models examined usually have been misspecified because they did not include the "positive" side of the deterrence equation-anticipated gain. The calculus of deterrence envisioned by classical theorists such as Bentham (1948)-and extended to contemporary statements of rational choice (Cornish and Clarke, 1986)-sought to find the appropriate balance between punishment and gain to prevent criminal acts from occurring; thus, studies failing The research on which this paper is based was funded by grant 89-1J-CX-0046 from the National Points of view or opinions expressed in this document are those of the authors and do not necessarily represent the official position or policies of the U S . Department of Justice. We are grateful to Paul Garthwaite, Department of Statistics, University of Aberdeen, for his assistance with the logit analysis and to Richard Rosenfeld for his helpful comments on the manuscript.CRIMINOLOGY VOLUME 31 NUMBER 1 1993 135 1. A larger sample obviously would have reduced the chances of making a type I1 error. Given the difficulties of recontacting active offenders, however, obtaining such a sample was patently impractical.