Prior studies of sleep in Parkinson's disease (PD) have been compromised by inadequate comparison groups, mixed medication regimens, and absence of quantitative data collection. This is the first study to compare polysomnographic sleep measures in PD patients on only dopaminergic medications with and without hallucinations. We performed two consecutive nights of polysomnography in 10 nondepressed, nondemented PD patients, 5 with and 5 without hallucinations. All patients were being treated with carbidopa/levodopa and a dopaminergic agonist only. Hallucinators and nonhallucinators were group-matched for age, PD duration, severity, and medication doses. Both groups had abnormal sleep records. In particular, there was a reduction in K-complexes and spindle formation, and the frequent occurrence of motor activation during rapid eye movement (REM) sleep consistent with REM behavior disorder. The hallucinator group had a significantly lower sleep efficiency (0.25 in hallucinators vs 0.61 in nonhallucinators, p = 0.006), a reduced total REM sleep time (mean total REM sleep time, 3 minutes in hallucinators vs 50 in nonhallucinators; p = 0.005), and a reduced REM percentage (mean, 5% in hallucinators vs 20% in nonhallucinators; p = 0.011). This study demonstrates that advanced PD patients treated with dopaminergic agents have abnormal sleep patterns and that those with dopaminergic-induced hallucinations have significantly greater REM aberrations than nonhallucinating PD patients.