Two hundred women presenting with primary breast carcinoma were studied to find the most accurate single or combination of methods to assess breast tumour size. Correlations of the maximum clinical, mammographic and ultrasound tumour diameter were made with maximum histological diameter. Tumour size could be assessed clinically in all 200 patients, and overestimated the size of small tumours and underestimated large tumours (P less than 0.001). Mammographic measurement, which was possible in 145 (72.5%), underestimated the size of large tumours (P less than 0.01). Only 100 women underwent ultrasound examination (size assessed in 86%) and this modality tended to underestimate the size of all tumours (P less than 0.05). All methods of measurement showed similar correlations with histological size. Stepwise linear regression showed that the most accurate and practical estimation could be made using the formula: Histological size = 0.5 x mammographic size + 0.5 x clinical size. We conclude that clinical measurement of breast cancer size is as accurate as that from mammography or ultrasound. Accuracy can be improved by the use of a simple formula of both clinical and mammographic measurements.