Cognitive deficits in Parkinson's disease (PD) have increasingly been recognized over the last decade and reversal learning in particular has received a great deal of attention. In a classical reversal-learning paradigm, participants learn to associate various stimuli with specific responses (i.e. A →Positive; B→ Negative), and subsequently learn to associate the same stimuli with opposite outcomes (i.e., A→Negative; B→ Positive). Prior studies have revealed that medicated PD patients have a selective impairment with learning from negative, but not positive, outcomes, even when both reward- and punishment-related stimuli were equally relevant. The aim of the present study was to further understand this impairment by applying a novel reversal-learning paradigm, which enables the differentiation between positive/negative and cue/context reversal impairments. Twenty-seven medicated PD patients and twenty-nine healthy individuals matched for age, gender and education completed the cue-context reversal learning paradigm. The results revealed no significant differences in context reversal learning between individuals with PD and healthy controls. However, in cue reversal learning, healthy controls were significantly better at performing positive-to-negative reversal trials compared to individuals with PD, while individuals with PD were significantly better in negative-to-positive reversal trials compared to healthy controls. As such, the present study distinguishes between different types of reversal learning and suggests that different neural circuits are responsible for context and cue learning. These results improve our understanding of the possible effects of dopaminergic medications and may have important clinical implications.