This study explores the longitudinal and subgroup measurement properties of a 10-item, physical activity decisional balance scale, previously published by Plotnikoff, Blanchard, Hotz, and Rhodes (2001), within a diabetic sample of Canadian adults. Results indicated that a three-factor measurement model consistently improved model fit compared to the previously published two-factor model. Evidence of configural, metric, and scalar measurement invariance across time and among subgroups suggests that the 10-item decisional balance scale is appropriate for investigating associative relationships with other constructs and for comparing group means of the pros and cons subscales among a variety of diabetic population subgroups. (Contains 4 tables, 2 figures, and 3 footnotes.)