In order to distinguish impairment in set-shifting in Parkinson's disease (PD) from inability to inhibit distraction by stimuli that compete for attention, 18 nondemented patients with idiopathic PD and 13 normal controls equated for age and education, were administered the Odd-Man Out (OMO) test and the Stroop Color-Word Test. PD patients were significantly impaired on the OMO test but showed no evidence of interference during the Stroop test. Analysis of error patterns during the OMO test indicated that the requirement to repeatedly switch rules, rather than the requirement to maintain steady responding between rule switches, was responsible for impaired OMO performance. It is concluded that the OMO test is fundamentally a test of set shifting, rather than a test of set maintenance in PD. In addition, analysis of a larger sample of PD patients revealed a significant positive relationship between number and severity of extrapyramidal signs and error production on the OMO, and between the latter and global mental function. These relationships were independent of each other, suggesting that impairment in set-shifting function in PD may arise from pathology of the fronto-striatal system independently of changes in cognitive ability.