Previous meta-analyses have estimated that the intention–behaviour association in physical activity (PA) is large in magnitude. However, these prior meta-analyses have also revealed a large degree of heterogeneity, suggesting the presence of moderating variables. This study examines the impact of one such moderator, testing the hypothesis that the magnitude of the association between intention and behaviour decreases as the temporal separation between the two increases. A systematic literature search was used to identify published and unpublished studies that met the inclusion criteria. A random-effects meta-regression was conducted to test the study hypothesis. A total of 78 journal articles and 11 unpublished dissertations were identified, yielding 109 effect sizes. The mean number of weeks between the measurement of intention and behaviour was 5.4 (SD = 6.6, range = .43, 26). The average correlation between intention and behaviour was r = 0.51. In line with theoretical predictions, temporal separation was a significant moderator of the intention–behaviour correlation (B = −.014, p < .001) and explained 24% of the between-study variance. This result remained unchanged when entered simultaneously with several control variables. The results of this analysis have important implications both for researchers and for intervention designers aiming to increase rates of PA.