Wolfe has suggested that the mammographic parenchymal patterns can be used to identify a group with high incidence of breast cancer. To evaluate this claim, mammograms of women with breast cancer that was detected at the University of Michigan Breast Cancer Detection Demonstration Project have been classified and compared with a randomly selected control group from the same project. The basic mammographic classifications as defined by Wolfe were used with further refinements made in the DY and QDY groups. The mammographic classifications have been grouped according to the degree of density and age. For all ages combined, our dense classifications (DY1, DY2, DYC-, QDY2) show a higher incident rate, 22/1000,than the lucent classification (N1, P1, and QDY1),9/1000,(P less than .01), although not to the degree suggested by Wolfe. This difference is statistically significant (P less than .01); the higher risk is markedly increased for women under 50 years of age (P less than .005). After age 50 the higher risk associated with the dense breast seems to disappear (P less than .13). However, this might be a consequence of women who were in a dense classification at an earlier age who subsequently changed to a lucent classification later in life. Because the percentage of lucent breasts increases with age there is a higher absolute number of cancers (55%) in this group of women past 50; these women cannot be neglected in screening. Our results suggest the following guidelines for clinical evaluation: (1) Careful mammographic and clinical follow-up for any woman with a dense breast at any age. (2) Careful mammographic and clinical follow-up past 50 regardless of breast classification. (3) Women with lucent breasts under age 50 represent a low risk category and may not require as frequent a follow-up as the other mammographic types.