Correctional researchers have increasingly focused on social bonding as a key pathway by which parolees desist from crime after release. Most work to date has focused on levels of bonds, either at reentry or as a function of events occurring in the community. However, few have assessed whether the magnitude of change in bonds during incarceration has an additional effect on desistance. Distinguishing between levels and change with respect to bonds may have important implications for understanding how bonds impact behavior. This paper addresses this gap by drawing on survey data from a sample of inmates at the start and end of their 6-month prison terms. Recidivism is assessed as a function of change in social bonds (attachments and beliefs) from entrance to exit from prison, as well as levels of bonds at release. The findings indicate that changes in social relationships predict recidivism, whereas improvements in prosocial beliefs do not. The data also suggest that the level of prosocial belief at release is significantly related to recidivism, whereas the level of attachment is not. The implications of these findings and directions for future research are discussed. Abstract published by arrangement with Taylor and Francis.