This study aims to longitudinally examine relationship patterns among attitude, criminal associates, and recidivism among Black (n = 109) and White men (n = 107) released from Allegheny County Jail in Pittsburgh, Pennsylvania. Attitude and criminal associates belong to the “big four” risk factors for recidivism (Andrews & Bonta, 1998). Review of studies reveals that different dimensions of attitude predict different measures of recidivism in different offender populations. This study examines two new attitudinal dimensions – autosuggestion and attitude toward community-based services (CBS attitude in the following). Autosuggestion measures the reported likelihood of ex-inmates’ future offending. CBS attitude is a meaningful measure among jail populations given short jail stays and the critical role played by community-based services in jail ex-inmates’ reintegration.

The original path model with three-wave data was split into four hypotheses because of inadequate bivariate correlations among focal variables. Longitudinal relationships between attitude and recidivism, and criminal associates and recidivism, and longitudinal reciprocal relationships between attitude and criminal associates were investigated. Each hypothesis was tested in the entire sample (with interactions) and in each subgroup by race, age, and offense type.

Findings indicated that criminal associates predicted recidivism and attitude, but attitude failed to predict recidivism and associates with an exception that CBS attitude predicted recidivism in some groups. This latter finding illuminates the importance of the community-based services and CBS attitudes. Autosuggestion interacted with age and CBS attitude with race in predicting recidivism. Results suggest that “very likely” response of autosuggestion may contain two different meanings – criminal intention and acknowledgement of vulnerability, possibly leading to two different recidivism results. Improvement of the two attitude measurements is suggested necessary considering the double meaning contained in autosuggestion and cultural competency of CBS attitude measure. In addition, attitude was shown to change over time, and attitude change may make a better predictor for recidivism and criminal associates than attitude measured at a time. The original path model may be tested with attitude change as a predictor. Factors for attitude change, possibly including criminal associates, should also be investigated. Other points of discussion, and significance and limitations of the study are discussed.