Cognitive impairment is common in Parkinson's disease (PD), and many patients will eventually develop a dementia, which has a devastating impact on the patient and their family. As such, there has been much interest in identifying a prodromal state to inform prognosis and facilitate earlier management, similar to the concept of 'MCI' in the Alzheimer's field. However, grouping the early cognitive deficits of PD together as 'PD-MCI' may not be the best way forward as it implies a single aetiological basis with one clinical consequence. In this review, we argue that cognitive deficits in PD arise from a number of different pathological pathways, only some of which herald a dementing process. This has important implications both for treatment of individual patients, and for the design of future disease-modifying therapy trials.