Perceptual deterrence researchers have used simple cross-sectional correlations between prior behavior and current perceptions to study the effect of legal threats on social control. Such designs are inadequate because they: (1) confuse the causal ordering of perceptions and behavior, and (2) fail to take into account other inhibitory factors in an explicit causal model. In an analysis of panel data, the methodological simplicity of earlier studies is shown to have led researchers to reach erroneous conclusions. Our data suggest that past studies report an experiential effect, not a deterrent effect, and that the effect of perceived sanctions on criminal involvement is minimal once social definitional factors (moral commitment, informal sanctions) are controlled.