The practice of parole is commonly understood as a way to release offenders from prison prior to the completion of sentence. The practice of parole is steeped in the ideal of rehabilitation and the medical model. Although law violators are thereby presumed to be 'sick,' they are also presumed to recover at different rates. The task of the parole board is to judge who has recovered sufficiently to leave prison and live in the outside world. Data used to explore the parole decisionmaking process were gathered in 1972 at a small maximum-security prison for adult felons. Institutional records of incarcerated inmates, which are compiled and submitted to the parole board before inmate hearings, were evaluated. The research sought to identify the relative power of variables contained in the institutional records that predict the parole process. Stepwise multiple regression was used to construct a simple block model by which the effect of a number of predictor variables on three different dependent variables could be examined simultaneously. Analysis reveals that inmate records contain few variables consistent with the medical model philosophy, e.g., involvement with prison educational programs. In addition, the predictive power of these variables ranges from negligible to nonexistent. To the extent that any pattern in decisionmaking can be statistically reconstructed, the design probably proceeds from stereotypical labels that identify the 'kind of person' likely to succeed or fail, rather than sick-well distinctions as the model suggests. More than anything else, the analysis calls into question the difference between what we say we do, why we say we do it, and what we do in fact. One figure, 1 table, 10 notes, and 42 references are included.