Fifty-three breasts resected for a biopsy diagnosis of duct carcinoma in situ were studied with a serial subgross and correlated radiographic method of examination designed to permit quantitation of the extent of the noninvasive lesion in the breast. Overall frequencies of occult invasion and multicentricity were 21 and 32%, respectively. Among 24 lesions 25 mm or larger in extent (average, 63; median, 56 mm) 11 showed occult foci of invasion, 13 had multicentric foci and six had nipple involvement. Among 29 lesions less than 25 mm in extent (average, 10; median, 8 mm) there were no instances of occult invasion, four were multicentric and two had nipple involvement (P = less than 0.05 for multicentricity and occult invasion). Twenty patients with lesions averaging 8 mm in extent are being followed after excision only in an experimental program. There have been three local recurrences at an average follow-up of 44 months. All recurrences occurred ipsilaterally, two were within the prior biopsy site. All patients with recurrence are free of disease following local resection in two and modified radical mastectomy in one. For lesions with associated microcalcifications, the distribution of the mammographic microcalcifications closely approximates the extent of disease as confirmed histologically. These findings suggest that an important predictive factor for the presence of occult invasion and multicentricity in the resected breast is the extent of the noninvasive lesion.