One hundred and nine breast lesions were examined by dedicated-waterpath ultrasonography unit with 7.5MHz polymer transducer (copolymer of vinylidene fluoride and trifluoethylene). These cases were also evaluated by physical examination and pathological examination. Overall accuracy by ultrasonographic diagnosis was 72.5%, while that of physical diagnosis was 61.5%. Accuracy in carcinoma of ultrasonographic diagnosis was 80.7%, while that of physical diagnosis was 77.1%. Irregular margin, irregular internal echoes, hyperechoic corona (hyperechoic belt surrounding mass lesions), attenuation of back echo and connective tissue sign were recognized as malignant findings. By utilizing these criteria, 11 cases of false negative carcinoma were correctly interpreted in retrospect. Two of 9 cases of false positive carcinoma were also correctly interpreted, but the rest of 7 cases were diagnosed as carcinoma even in retrospect. Overall accuracy of retrospective examination of ultrasonography was 86.2%, while that in carcinoma was 92.7%. Connective tissue sign was particularly useful, being a major contributing factor of the greater diagnostic accuracy in the retrospective study. Three quarters of all cases were within 2 cm in diameter. Accuracy of physical diagnosis in the lesions less than 1 cm in diameter was 75.1%. Whereas that of ultrasonographic diagnosis was 91.7% in the same group. Ultrasonography was effective in such small lesions that were hardly palpated. Eighty-nine point six percent (53/58) in non-malignant lesions such as fibro-adenoma or mastopathy, was horizontally elongated in configuration. Ninety-six point eight percent (30/31) of vertically oriented masses were malignant. Seven point five mega-hertz transducer had greater resolving power than 5 MHz transducer, whereas the findings like acoustic shadow and/or enhancement were not clearly imaged by the high frequency transducer.