OBJECTIVE: To review the epidemiologic literature on evaluating the effectiveness of Tai Chi, an aged Chinese exercise regime, for improving balance and preventing falls and to provide recommendations for future research design. METHODS: Literature on Tai Chi, published after 1966, as identified by Medline, was reviewed. Studies were selected if they met the following criteria: 1) the research was related to balance and fall prevention; and 2) observational study designs, including: case series, ecologic, cross-sectional, cohort, case-control and quasi-experiment. RESULTS: Thirteen studies were identified, based on the two criteria above, from a total of 187 articles. Findings from these studies are inconsistent because of a number of inherent limitations and potential biases in the study design. These limitations include: selection bias due to non-randomization of study subjects in quasi-experimental study; no convincing causal relationship established using cross-sectional study; selection bias due to poor baseline demographic data collection and assessment; failure to identify the difference between people who are compliant to the study and those who drop off; lack of external validity due to including only one gender or one age group in the study; failure to define intervention assessment in the study (e.g., the style, intensity, frequency of the Tai Chi intervention) and no standard measurement for improving balance across different studies. CONCLUSIONS: The evidence of the effectiveness of Tai Chi on improving balance and preventing falls based on existing observational studies is inconclusive. In the future, randomized clinical trial should be used in order to control confounding and eliminate bias. More systematic studies are needed to understand the therapeutic nature of Tai Chi. Based on this, the consensus about the terms of use and core measures that help describe and compare participants, characterize Tai Chi interventions, and describe relevant outcomes has to be achieved.