We investigated the utilization of mammography as a screening test for breast cancer in a middle-income Connecticut suburban community of 30,000 people. The sampling frame was community-dwelling women aged 30 years and over who had telephones. Random digit telephone survey methods were used to identify a sample of 470 eligible subjects. Of those eligible to be included, 350 or 74.4% completed the interview. Analysis of data from the 171 respondents aged 50 years or greater indicated that women aged 65-80 years had a significantly lower rate of screening mammography than did women aged 50-64 years (means 2 = 6.6, P = .01). When further analysis was done to take into account the effects of education and of income on these rates, the association of age with mammography utilization was no longer statistically significant. Among women who recalled their physician advising a mammogram, 88% had had one performed. Among women who could not recall their physician advising a mammogram, 7% had had one. The impact of physician advice was statistically significant (means 2 = 110.3, P less than .001). Physicians recommended screening mammography less for patients with low level of education (means 2 = 21.6, P less than .001), low income (X2 = 7.8, df = 2, P = .02) and greater age (means 2 = 14.2, P = .003). We conclude that utilization of screening mammography in the community studied is related more strongly to education and to income than to age. The bivariate association of mammography utilization with age may be attributable to a cohort effect, rather than an age effect.(ABSTRACT TRUNCATED AT 250 WORDS)