Significant bivariate correlations between the LSI-R total score, the Drug and Alcohol Problem subscale, and the outcome of interest (reincarceration) suggest that the LSI-R is a valid prediction instrument for drug-involved offenders. The study examined several control variables, including age, criminal history, instant offense, time-at-risk in the community, type of release, and three treatment-related variables. In analyses not shown here, race was an insignificant predictor of incarceration in all regression models examined. It was excluded from the final analyses because including it would have substantially reduced the sample size. After accounting for these control variables, both the Drug and Alcohol Problem subscale and the LSI-R total score remained significant predictors of reincarceration. Data came from a sample of 276 offenders who participated in in-prison substance abuse treatment at the State Correctional Institution at Chester, PA, and been released to the community. More than half of the sample (55.1 percent) had been sentenced to their current term of incarceration for a drug offense. All participants in the final sample were male. Both outcome data (reincarceration) and LSI-R data were available for the sample. The LSI-R was administered either shortly before the inmate’s release from custody or immediately after his release. While in prison, all participants had been involved in 12 months of substance abuse treatment. 3 tables, 3 notes and 41 references