Mammography and ultrasonography are traditional for preoperative estimation of breast cancer size; magnetic resonance imaging (MRI) is more recent but not as well studied. We compared ultrasonography, mammography, and MRI for preoperative imaging of primary breast cancer presenting as a mass in patients treated at our center over a 2-year period. Of the 61 breast cancers with all three imaging modalities performed, 52 were infiltrating ductal cancer, 5 were infiltrating lobular cancer, 2 were ductal carcinoma in situ, and 2 were other histologic types. When pathologic size was used to determine the accuracy of imaging assessments, the Pearson correlation coefficient was better for MRI (r = 0.80) than ultrasonography (r = 0.57) or mammography (r = 0.26). Mean tumor size was 2.1 cm by mammography, 1.73 cm by ultrasonography, 2.65 cm by MRI, and 2.76 cm by pathology. MRI-based tumor size was within 1 cm of pathologic size in 44 (72%) tumors, > 1 cm above pathologic size in 6 (10%) tumors, and > 1 cm below pathologic size in 11 (18%) tumors. We conclude that MRI is more accurate than either ultrasonography or mammography for assessment of the size of primary breast cancer presenting as a mass.