PURPOSE: The magnitude of intra-individual variation, as determined by three methods of physical activity (PA) assessment, was estimated to determine the optimal number of repeated measures needed to reliably measure an individual's usual PA habits. METHODS: Data were obtained from 77 healthy adults enrolled in the Survey of Activity, Fitness, and Exercise (SAFE). PA was measured by: 1) Caltrac accelerometers worn 48-hours every 26 days for one year; 2) concurrent 48-hour PA records (PARs); and 3) a corresponding series of four-week PA histories (FWH). Repeated measures regression analyses were used to analyze the data. RESULTS: PA tended to be slightly higher in the summer months as compared with the winter months, but this seasonality explained little of the intra-individual variation. The intra-individual standard deviation (SD) estimated for the FWH (180 MET-min d−1 ) was 39% of the mean annual MET-min d−1, as compared with the Caltrac (163 MET-min d−1 or 9%) and PAR (259 MET-min d−1 or 11% ). Given the intra-individual variation, three repeated observations of the FWH, six Caltrac, or nine PAR over the course of the year were needed to achieve 80% reliability in estimation of mean annual PA. CONCLUSIONS: Intra-individual variation in PA is substantial and must be accounted for in estimates of usual PA.