Background: Postpartum women are a population at risk for sedentary living. Physical activity (PA) prior to pregnancy may be effective in predicting similar behaviour in the postpartum period. Objective: To test a composite version of the extended transtheoretical model (TTM) by adding "past behaviour" in order to predict PA behaviour among postpartum Iranian women. Methods: Participants were 300 primiparous postpartum women from health clinics in Arak, Iran who completed the Iranian versions of the TTM, past behaviour and PA behaviour questionnaires. The mean age of the participants was 24.85 years (SD = 4.4, range = 15-41). Data were analysed using bivariate correlation and path analysis. Results: The first revised model accounted for 43% of the variance in current PA and represented a good data fit (?2 = 7.45, df = 4, p = 0.114). There were significant direct pathways between PA and stage of change (ß = 0.51), pros (ß = 0.30), and cons (ß = -0.31), and significant indirect effects of self-efficacy (ß = 0.39) and processes of change (ß = 0.34). The final model, which included the direct effects of past behaviour (ß = 0.28), explained 50% of the variance in PA and represented a good data fit (?2 = 8.30, df = 5, p = 0.14). Additionally, in this model pros (ß = 0.25) and cons (ß = -0.26) are directly related to PA, and self-efficacy (ß = 0.39) and processes of change (ß = 0.34) are indirectly related to PA. Conclusion: The findings of this study support the application of TTM to PA behaviour change among Iranian postpartum women. However, adding past behaviour to the model increases our understanding of the population's PA and may help the planning of more efficient interventions for promoting PA.