This study investigated the effect of cognitive training aimed at improving shifting ability on Parkinson's disease (PD) patients' performance of prospective memory (PM) tasks. Using a double-blind protocol, 17 PD patients were randomly assigned to two experimental arms. In the first arm (n=9) shifting training was administered, and in the second (placebo) arm (n=8), language and respiratory exercises. Both treatments consisted of 12 sessions executed over 4 weeks. PM and shifting measures (i.e., Trail Making Test and Alternate Fluency Test) were administered at T0 (before treatment) and T1 (immediately after treatment). A mixed analysis of variance was applied to the data. To evaluate the effects of treatment, the key effect was the interaction between Group (experimental vs. placebo) and Time of Assessment (T0 vs. T1). This interaction was significant for the accuracy indices of the PM procedure (p<.05) and for the performance parameters of the shifting tasks (p ≤.05). Tukey's HSD tests showed that in all cases passing from T0 to T1 performance significantly improved in the experimental group (in all cases p ≤.02) but remained unchanged in the placebo group (all p consistently>.10). The performance change passing from T0 to T1 on the Alternate Fluency test and the PM procedure was significantly correlated (p<.05). Results show that the cognitive training significantly improved PD patients' event-based PM performance and suggest that their poor PM functioning might be related to reduced shifting abilities.