Between March 1980 and April 1981, 1,140 women underwent physical examination, xeromammography, and whole-breast ultrasound (US) using a whole-breast water path system. Results of each study were interpreted independently by separate observers in a blind fashion. Biopsy revealed 125 cancers in 127 breasts. Findings were considered to be suspicious for carcinoma in 199 women based on physical examination, in 201 based on mammograms, and in 255 based on US scans. Physical examinations were able to reveal 91% (115/127) of the cancers, reflecting the referral nature of the population, and failed to detect 12 lesions. Mammograms disclosed 94% (119/127) of the cancers, including 12 clinically occult lesions, but did not show eight palpable cancers. US scans disclosed only 64% (81/127) of the cancers, all of which were palpable. In a 4-year follow-up, no cancers have appeared in the group that had suspicious findings by US only. We conclude that US should not be used routinely to screen women for breast cancer.