Hallucinations, mainly of a visual nature, are considered to affect about one-quarter of patients with Parkinson's disease. They are commonly viewed as a side-effect of antiparkinsonian treatment, but other factors may be involved. The aim of this study was to determine the phenomenology, prevalence and risk factors of hallucinations in Parkinson's disease. Two-hundred and sixteen consecutive patients fulfilling clinical criteria for Parkinson's disease were studied. Demographic and clinical variables were recorded, including motor and cognitive status, depressive symptoms and sleep-wake disturbances. Patients with and without hallucinations were compared using non-parametric tests, and logistic regression was applied to significant data. Hallucinations had been present during the previous 3 months in 39.8% of the patients, and fell into three categories: minor forms, consisting of a sensation of a presence (person), a sideways passage (commonly of an animal) or illusions were present in 25.5% of the patients (an isolated occurrence in 14.3%), formed visual hallucinations were present in 22.2% (isolated in 9.3%) and auditory hallucinations were present in 9.7% (isolated in 2.3%). Patients with minor hallucinations had a higher depression score than non-hallucinators but did not differ in other respects. Logistic regression analysis identified three factors independently predictive of formed visual hallucinations: severe cognitive disorders, daytime somnolence and a long duration of Parkinson's disease. These findings indicate that, when minor hallucinations are included, the total prevalence is much higher than previously reported. A simple side-effect of dopaminergic treatment is not sufficient to explain the occurrence of all visual hallucinations. The main risk factor in treated patients is cognitive impairment, although sleep-wake cycle disturbances, and possibly other factors related to the duration of the disease, act as cofactors.