BACKGROUND The purpose of this study was to test the ability of the Transtheoretical Model (TTM) to predict exercise stage transition of individuals in a large, untreated-population-based, random sample of Canadian adults (18-65 years of age) over two consecutive time periods. METHODS Assessments of TTM's stage of exercise behavior change, self-efficacy, pros, cons, experiential processes, and behavioral processes were made at baseline (time 1), 6 months (time 2), and 1 year (time 3). Six hundred eighty-three men and women, identified through random-digit telephone dialing, completed all measures across the three time points. Within each time period (time 1-2; time 2-3) participants were categorized as having regressed (moved back at least on stage), remained (no stage change), or progressed (moved forward at least one stage). Baseline TTM constructs were analyzed for their ability to predict change transition across the two time periods. RESULTS Of 40 possible predictions (20 for each time period) 18 (45%) were supported. CONCLUSIONS Overall, the findings demonstrate partial support for the internal validation of TTM in the exercise domain. Implications of the findings are discussed and future directions for researchers, practitioners, and program planners are provided.