In the present study we tested the hypothesis that learned irrelevance underlies the frequently observed poor performance of Parkinson's disease (PD) patients on card sorting tests. If learned irrelevance accounts for the poor performance of PD patients on card sorting tests, PD patients and control subjects (CS) will not differ in the acquisition phase, during which basic concept formation is assessed, but they will differ in the subsequent shifting phases. We presented three distinct card sorting tests with an identical format to 51 PD patients and 24 normal controls. The groups did not differ with respect to intelligence, memory, or attention. PD patients showed a slightly better performance in the acquisition phase. In the first shifting phase, the performance of PD patients was significantly poorer than that of control subjects after correction for basic concept formation. In the second shifting phase this difference disappeared. We conclude that learned irrelevance does not account for the poor performance of PD patients in card sorting tests. The results are discussed in terms of self-generation of problem solving strategies.