With respect to the ongoing discussion of "sleep attacks" in Parkinson's disease (PD), we sought to estimate the prevalence of sudden onset of sleep (SOS) with and without preceding sleepiness in PD, to identify associated factors, and to define the role of antiparkinsonian medication in SOS. We sent a questionnaire about SOS, sleep behaviour, and medication to 12,000 PD patients. The response rate was 63%, from which 6,620 complete data sets could be analysed. A total of 42.9% of our population reported SOS, 10% of whom never experienced sleepiness before the appearance of SOS (4.3% of all), and we identified the administration of all dopaminergic drugs as a risk factor for SOS. However, SOS occurred earlier after introduction of nonergoline dopamine agonists (DA) and was more strongly associated with nonergoline DA in younger patients (below 70 years) with a shorter disease duration (up to 7 years) but, actually, medication was less efficient in predicting SOS than most other factors considered such as higher age, male sex, longer disease duration, and the report of sleep disturbances. This survey strongly suggests that SOS is a multifactorial phenomenon. Some subgroups are at particular risk of experiencing SOS under nonergoline DA, especially at the beginning of this therapy. Our results support the current notion that SOS, in part, can be attributed to PD-specific pathology because disease duration and subjective disease severity have been shown to be predictors of SOS. We recommend the development of a standardised question to recognise SOS and to facilitate the comparison of prevalence estimates.