Dishman (2001) has described the need to better understand the “natural history” of exercise patterns and how they are developed and maintained. To that end, this investigation examined the level of association of stage of exercise (Transtheoretical Model) with exercise identity and exercise efficacy. Theory suggests that exercise identity and exercise efficacy should explain progression and maintenance of exercise behavior in later stages of the change process. Data were collected from 78 elderly women living in the Tr-City area of Tennessee. Nineteen participants were black and 59 were white. Average age was 70.8 years with a range of 59 to 98 years. Stage of exercise behavior was assessed using a five-item instrument related to regular exercise behavior and intentions (Marcus, Selby, Niaura, & Rossi, 1992). Exercise identity was measured by the sum of nine Lickert-type items (Anderson & Cychosz, 1994) designed to measure the extent to which exercise was descriptive of the concept of self (alpha=.90). Exercise efficacy (Marcus, et al., 1992) was assessed by the sum of seven Likert-type items designed to measure confidence in one’s ability to overcome barriers to exercise participation (alpha=.66). A MANOVA with race and exercise stage utilized as independent variables and age, exercise identity, and exercise efficacy as dependent variables revealed statistically significant main effects for race (Wilks’ Lambda=.14, F(3,64)=3.36, p=.02) and exercise stage effect (Wilks’ Lambda=.46, F(9,156)=6.61, p<.001). The interaction effect was not significant. Univariate results for race were only significant exercise identity. Black participants displayed a higher mean exercise identity score (40.70) than white participants (30.94). Univariate results for exercise stage were significant for both exercise identity and exercise efficacy. Exercise identity scores and exercise efficacy scores both increased with exercise stage. A stepwise regression analysis was performed with exercise stage as the dependent variable and exercise identity, exercise efficacy, and age as predictor variables. Results indicated that all three predictor variables were significantly associated with exercise stage (R2=53%). Exercise identity score entered first and explained 40% of the variance in exercise stage, exercise efficacy entered next and made a significant addition of 8%, and age entered last and made a significant addition of 5%. These data provide insight into the sociopsychological process underlying maintenance and expansion of exercise activities. These data suggest that exercise identity increases through the change process, peaking in the “maintenance” stage. This is consistent with the theory that the development of this social identity and self-perceptions of exercise efficacy may combine to sustain and perpetuate behavior.