Four hundred and sixteen invasive breast cancers, detected initially by mammography, were compared with 929 presenting symptomatically, all treated at a South Australian teaching hospital. Predictable differences included lower stages and grades, less vascular invasion and proliferative activity, and more hormone-receptor expression among the mammographically detected. Unpredicted differences included significantly higher survivals for mammographically detected cases throughout the 9 year follow-up period after adjusting for stage and the Nottingham Prognostic Index. In a multivariable analysis, differences in stage, grade, and hormone receptor expression accounted for only about half the survival advantage of mammographically detected tumours. Accounting for additional person and tumour characteristics had only a marginal effect on this result. This suggests that detection by mammography has independent favourable prognostic significance beyond that explained by conventional indicators. If confirmed, this finding would have important implications for the prognostic advice given to women and may merit further investigation into its underlying biological mechanisms.