Objective. The purpose of this systematic review is to evaluate the evidence on the effect of Tai Chi for Parkinson's disease (PD). Methods. Six electronic databases up to June 2014 were searched. The methodological quality was assessed with PEDro scale. Standardised mean difference and 95% confidence intervals of random-effects model were calculated. Results. Nine studies were included in our review. The aggregated results are in favor of Tai Chi on improving motor function (P = 0.002) and balance (P < 0.00001) in patients with PD. However, there is no sufficient evidence to support or refute the value of Tai Chi on improving gait velocity (P = 0.11), stride length (P = 0.21), or quality of life (P = 0.40). And there is no valid evidence in follow-up effects of Tai Chi for PD. Conclusion. The current results suggest that Tai Chi can significantly improve the motor function and balance in patients with PD, but there is indeed not enough evidence to conclude that Tai Chi is effective for PD because of the small treatment effect, methodological flaws of eligible studies, and insufficient follow-up. Consequently, high-quality studies with long follow-up are warranted to confirm current beneficial findings.