In recent years, there has been a major shift in our understanding of the course of Parkinson's disease (PD) from a disease of the brain to a disease of long latency, characterized by the progressive emergence of multiple non-motor symptoms, including hyposmia, constipation, depression, anxiety, rapid eye movement (REM) sleep behavior disorder and excessive daytime sleepiness, as well as subtle motor signs, before the typical motor signs appear. Epidemiological studies have made major contributions by allowing better characterization of subsequent PD risk in relation to non-motor symptoms. Such findings have profound implications for the conduct of epidemiological studies examining risk and protective factors in PD, and the interpretation of their findings. Given the length of the prodromal period, reverse causation in particular is a major concern with many reported associations. One striking feature of PD etiology, compared with other diseases, is the presence of numerous inverse associations. If these associations are truly causal, they would have major implications for disease prevention and for slowing disease progression. However, whether these associations are truly causal remains to be demonstrated in future studies. Experimental studies play an important role by offering a better understanding of the underlying mechanisms. Well-designed epidemiological studies using innovative approaches will also be key in elucidating whether these intriguing associations are causal or a consequence of reverse causation.