We have used ultrasound of the breast to define four parenchymal patterns in which increasing proportions of the breast are replaced by densely echogenic tissue. A series of 452 symptomatic women examined by both ultrasound and conventional X-ray mammography was reviewed to determine whether these ultrasonographic images could predict the breast parenchymal pattern denned mammographically. A very strong correlation was demonstrated between the breast pattern on ultrasound and the volume of the breast replaced by either dysplasia (Kendall's tau-b = 0.731+0.026, p < 0.0001) or ductal prominence (Kendall's tau-b = 0.641 +0.049, p < 0.0001). This was seen both on initial reporting and on a blind re-reading of a random sample of 100 cases. The strength of correlation was similar for subgroups denned by family history of breast cancer, age, menopausal status, and history of benign breast disease, and the breast parenchymal pattern assessed by mammography or ultrasound showed similar associations with these variables. Ultrasonographic parenchymal patterns of the breast can predict the tissue patterns defined mammographically and may therefore be useful as a marker of breast cancer risk. An association between the parenchymal pattern of the breast and risk of breast cancer has been described by several authors (Wolfe, 1976; Krook et al, 1978; Gravelle et al, 1980). In previous work we have shown that radiologists can reliably identify mammographic dysplasia (Boyd et al, 1986), and that the relationship of mammographic dysplasia to risk of breast cancer is influenced both by the age of the subjects and by the proportion of the breast volume that is occupied by the radiological changes of dysplasia, as well as by the source of control subjects (Boyd et al, 1982a, b). These findings notwithstanding, the literature con- tains many contradictory reports of the association of mammographic dysplasia and risk of breast cancer. On investigation, however, most of the apparent variations