Our accuracy in diagnosing 464 solid breast masses by fine-needle aspiration was enhanced by a statistically based algorithm for distinguishing between benign and malignant epithelial cells. Epithelial cells were obtained from 378 breast masses by fine-needle aspiration; nonepithelial cells considered diagnostic of benign conditions were obtained from 66 breast masses, and 19 aspirations were considered unsatisfactory. Excluding one benign cystosarcoma, the algorithm gave 3 false-negative diagnoses and 16 false-positive diagnoses with 211 benign and 167 malignant samples. The overall clinical performance measures for the 444 masses from which diagnostic fine-needle aspirations were obtained and excluding the cystosarcoma were 0.98 sensitivity, 0.94 specificity, and 0.92 positive predictability. Masses diagnosed as benign by fine-needle aspiration can be followed up clinically. Intraoperative frozen section is needed before definitive surgery to determine invasion and confirm the diagnosis of some cytologically malignant masses.