The purpose of this study was, using composition data analysis, to compare urban black and white women's physical activity (PA) patterns. A total of 480 urban women (177 Whites, 293 Black/African-Americans, and 10 others, aged from 40 to 79 years) were recruited for the original study, and their exercise/PA participations (MET-min/day) were measured by a 97-item, 4-week PA recall questionnaire (Ainsworth, 1997). The participants were asked to recall their participation (days per month, times per day, and amount of time) in PA within the seven major categories. For the purpose of this study, only 200 participants (100 Black and 100 White) were randomly selected. Further, the category of "Walking/Going Place" was excluded and the rest categories were combined into three categories: "Household," "Leisure," and "Working for Pay." Based on the responses on these new categories, the total minutes of exercise/PA participation were computed, according to which each participant's Total MET-min/day was further computed. The proportions of each category (in MET-min/day) were then computed by dividing them with the Total MET-min/day. Because the statistical procedure used could not deal with the "zero" scores in a category (i.e., the participants did not report any activity in this category), 31 black and 36 white participants, who had the "zero" scores in one of three categories, were excluded from the further analysis. The retained data were analyzed using the composition data analyses, which were completed using the CODA software (Aitchison, 1986). Overall, some differences were found in urban black and white women's PA patterns. Black women participated slightly more PA in their work (Black: M=43.91%, SD=1.81%; White: M=40.62%, SD=2.25%), but less in their household (Black: M=43.94%, SD=1.89%; White: M=45.46%, SD=2.20%) and leisure activities (Black: M=12.15%, SD=1.27%; White: M=13.92%, SD=1.30%). The observation was verified by the ternary diagrams generated from CODA. The results of the Anderson-Darling, Cramer-won Mises, and Watson tests indicate that the data met the assumption of the logistic normal distribution. The black women were different from their white counterparts in the PA patterns (Likelihood: M & Covariances equal=3.31, p=.004; Covriances equal=10.39, p=.018). It is concluded that urban black and white women illustrated some differences in their exercise and PA participation. The study was partially funded by NIH WHI-SIP #22W-U48/CCU409664-03.