In a study of invasive breast cancer, multiple correspondence analysis (MCA) revealed clustering of eight pathobiological variables. Two different phenotypes were distinguished by an index calculated on the basis of the variables (histologic grade, necrosis, lymphoid infiltration, number of mitosis and expression of c-erbB-2, p53, progesterone receptor and Bcl-2). Phenotype A lesions share most of the features of normal breast tissue. Phenotype B looks more malignant, has a higher early recurrence rate and is more frequently seen in younger patients. Our aim was to see if ductal breast carcinoma in situ (DCIS) could be divided into the same phenotypes. One hundred and eighty DCIS were investigated. Association between the eight variables was studied in 2 x 2 models. The phenotype index was calculated by summing weights for the variables in the MCA. All variables were associated, except Bcl-2. DCIS was divided in two phenotypes. Thirty-three tumours were Phenotype A and 147 Phenotype B. The mean age at diagnosis was 65.5 and 58.4 years for Phenotypes A and B, respectively (p = 0.0012). No difference regarding local relapse free survival was seen. Two phenotypes were distinguished in DCIS, similar to invasive breast cancer. In an earlier study, 45% of the invasive cancers were classified as Phenotype B. In this study, 82% of DCIS were Phenotype B. This may indicate that invasive breast cancer of Phenotype B is derived from DCIS of Phenotype B. The distribution of DCIS phenotypes with a small proportion of Phenotype A DCIS may be due to that Phenotype A DCIS is less likely to be detected by mammography, or that some invasive breast cancers of Phenotype A progress to invasiveness without passing the in situ phase.