The specificity of the Wisconsin Card Sorting Test (WCST) as a marker of frontal lobe pathology remains controversial. One problem is the lack of a well established correspondence between WCST errors and specific cognitive or neural processes. The conventional scoring of non-perseverative WCST errors does not discriminate between errors related to the efficient test of hypotheses during set shifting (‘efficient errors’), and random failures to maintain set (‘random errors’). This inherent confusion in the non-perseverative error score probably minimizes the relative importance of random errors in frontal lobe pathology. In this study, we used a WCST version sensitive to differences between ‘efficient’ and random errors to examine set shifting deficits in patients with focal lesions to their lateral prefrontal cortex. As expected, patients showed abnormally high rates of perseverative errors. Interestingly, patients also showed enhanced rates of random errors suggesting constant shifts or fluctuations in their choice of sorting principle. These results suggest that more sensitive tests are needed to elucidate the association between a specific type of set shifting error and a particular type of frontal lobe pathology.