Objective: To present a critical review and meta-analysis of studies evaluating the effects of physical therapy in patients suffering from Parkinson's disease (PD), in terms of neurologic signs, activities of daily living (ADLs), and walking ability. Data Sources: Articles published from 1966 to May 1999 were compiled by means of MEDLINE, Cochrane register of controlled trials, and CINAHL using combinations of the key words Parkinson's disease, exercise, exercise therapy, physical therapy, and group training. References presented in relevant publications were also examined. Articles written in English, German, or Dutch were included. Study Selection: Studies had to meet the following selection criteria: (1) patients with PD were included in the intervention study, (2) the effects of physical therapy (PT) were evaluated, (3) the study could be classified as true or quasi-experiment, and (4) the study was published in a journal or book. Data Extraction: Two reviewers assessed independently the methodologic quality of the data of each included study. One reviewer extracted relevant meta-analysis data. Data Synthesis: For each outcome measure the estimated effect size and the summary effect size (SES) were calculated, using fixed (ie, Hedges's g) and random effects models. The meta-analysis resulted in a significant homogeneous SES with regard to ADLs (.40; confidence interval [CI] = .17-.64) and stride length (.46; CI = .12-.82). The SES with regard to walking speed showed a significant heterogeneous SES, which remained significant after applying a random effects model (.49; CI = .21-.77). The SES with regard to neurologic signs was not significant (.22; CI =-.08 to .52). The small number Of studies included and the shortcomings of the methodologic quality of these studies, however, bias the results of the present study. Conclusions: The results of the present research synthesis support the hypothesis that Parkinson patients benefit from PT added to their standard medication.