The relationship between exercise and anxiety has been extensively examined over the last 15 years. Three separate meta-analysis were conducted to quantitatively review the exercise-anxiety literature for state anxiety, trait anxiety and psychophysiological correlates of anxiety. Such a procedure allows tendencies of the research to be characterised. The results substantiate the claim that exercise is associated with reductions in anxiety, but only for aerobic forms of exercise. These effects were generally independent of both subject (i.e. age and health status) and descriptive characteristics. Numerous design characteristics were different, but these differences were not uniform across the 3 meta-analyses. For state anxiety, exercise was associated with reduced anxiety, but had effects similar to other known anxiety-reducing treatments (e.g. relaxation). The trait anxiety meta-analysis revealed that random assignment was important for achieving larger effects when compared to the use of intact groups. Training programmes also need to exceed 10 weeks before significant changes in trait anxiety occur. For psychophysiological correlates, cardiovascular measures of anxiety (e.g. blood pressure, heart rate) yielded significantly smaller effects than did other measures (e.g. EMG, EEG). The only variable that was significant across all 3 meta-analyses was exercise duration. Exercise of at least 21 minutes seems necessary to achieve reductions in state and trait anxiety, but there were variables confounding this relationship. As such, it remains to be seen what the minimum duration is necessary for anxiety reduction. Although exercise offers therapeutic benefits for reducing anxiety without the dangers or costs of drug therapy or psychotherapy, it remains to be determined precisely why exercise is associated with reductions in anxiety. Since several mechanisms may be operating simultaneously, future research should be designed with the idea of testing interactions between these mechanisms.