The relationship between tumour size and lymph node metastases was examined in screening-detected and clinically detected breast cancers. The data used were from a randomized trial of breast cancer screening with mammography. 964 cancers were reviewed, in both arms of the trial, in women aged 40-74. Lymph node status was significantly related to detection method (P less than 0.001), metastases being less common in screening-detected cancers. Node status was also significantly related to tumour size (P less than 0.001), metastases being commoner in larger tumours. Similarly, tumour size was significantly associated with detection method (P less than 0.001), smaller tumours being detected by screening. No significant interaction was observed among all three factors, indicating that the relationship between node status and tumour size did not change with detection method. When detection method was replaced with randomly allocated study (invited to screening) and control (not invited to screening) groups, the same results were observed. It is concluded that if screening detects tumours with a different natural history to that of those which surface clinically, this is not reflected in the relationship between tumour size and lymph node metastases.