This study was designed to compare the effects of exposure to two different sets of soluble discrimination problems, an easy set composed of only two- and three-dimensional problems and a more difficult set composed of problems ranging from two to seven dimensions, both immediately after training and at a 10-day posttreatment follow-up. The subjects were 32 depressed male inmates of a federal correctional institution. It was hypothesized that as a result of meeting and mastering progressively more difficult problems, the group given progressively more difficult problems would show a greater reduction in depression and a greater enhancement of performance on a variety of cognitive measures, both immediately after treatment and at the 10-day posttreatment follow-up. The results failed to support these hypotheses. Depression scores decreased significantly from pretreatment to posttreatment, but did so equally for the two groups. One of the cognitive measures, the WAIS Digit-Symbol subtest, showed significant improvements from pretreatment to posttreatment, but did equally for the two groups. Significant relationships were found between the subjects' performances on the cognitive tasks, and measures of their tendencies to attribute successes and failures to stable or unstable factors. Unexpected significant positive relationships were found between depression and performance on the cognitive tasks. The differential effect of the prison environment upon people differing in their intelligence was discussed as a possible explanation of these findings.