The aim of this study was to make prognostic comparisons between the modified scheme of Pedersen et al. the definitions of Tavassoli and the Ridolfi criteria for medullary carcinomas. Sixty breast carcinomas primarily diagnosed as medullary carcinomas were reclassified into typical medullary carcinoma (TMC), atypical medullary carcinoma (AMC) and non-medullary carcinoma (NMC) according to the three classifications. The Ridolfi classification proved to be superior to the two other schemes in discriminating survival differences between the three groups TMC, AMC and NMC. All 13 patients with TMC are still alive indicating an excellent prognosis, while 29% and 39% of the 47 patients in the AMC and NMC category, respectively, have died of their disease. In the simplified system of Pedersen et al. the survival at 10 years for TMC patients decreased to 75% and no significant survival difference between the three groups could be demonstrated. As the prognosis for AMC proved to be worse compared to TMC and in fact was similar to NMC with values of 43% at 10 years in the Ridolfi classification, we find no reasons to maintain this category. We conclude that as long as no alternative and more easily applicable diagnostic method exists, pathologists should still apply the Ridolfi criteria on these tumours with medullary features leaving two diagnostic possibilities: TMC or NMC (i.e. poorly differentiated ductal carcinoma). Only lesions that fulfil all six criteria without any doubt should be diagnosed as TMC, thus avoiding overdiagnosis and a resulting risk of undertreatment.