To determine the positive predictive value (PPV) of the American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) categories 0, 2, 3, 4, and 5 by using BI-RADS terminology and by auditing data on needle localizations. Between April 1991 and December 1996, 1,400 mammographically guided needle localizations were performed in 1,109 patients. Information entered into the mammographic database included where the initial mammography was performed (inside vs outside the institution), BI-RADS category, mammographic finding, and histopathologic findings. A recorded recommendation was available for 1,312 localizations in 1,097 patients, who composed the study population. The 1,312 localizations yielded 449 (34%) cancers (139 [31%] were ductal carcinoma in situ [DCIS]; 310 [69%] were invasive cancers) and 863 (66%) benign lesions. There were 15 (1%) category 0 lesions; the PPV was 13% (two of 15 lesions). There were 50 (4%) category 2 lesions; the PPV was 0% (0 of 40 lesions). There were 141 (11%) category 3 lesions; the PPV was 2% (three of 141 lesions). The three cancers in this group were all non-comedotype DCIS. There were 936 (71%) category 4 lesions; the PPV was 30% (279 of 936 lesions). There were 170 (13%) category 5 lesions; the PPV was 97% (165 of 170 lesions). Placing mammographic lesions into BI-RADS categories is useful for predicting the presence of malignancy. Perhaps, most important, a lesion placed into BI-RADS category 3 is highly predictive of benignity, and short-term interval follow-up as an alternative to biopsy would decrease the number of biopsies performed in benign lesions.