The frequency of oncogene amplification described in the literature shows a large fluctuation, which could be attributed to the study of relatively small series of tumours, to selection of subgroups of patients, or, especially in retrospective studies, to selection of tumour material from the tumour-bank. To address this question, we have studied amplification of c-myc, HER2/neu and int-2/bcl-1 genes in a series of 1052 collected human breast tumours. The retrospective and prospective subgroups in this collected series of tumours were of equal size. c-myc was amplified in 17.1%, HER2/neu in 18.7% and int-2/bcl-1 in 14.1%, of all breast cancer specimens studied. In the retrospective subgroup the prevalence of amplification was 18.1% for c-myc; 22.6% for HER2/neu and 11.6% for int-2/bcl-1, whereas in the prospective subgroup an incidence of amplification of 16.1%, 15.1% and 16.3% for c-myc, HER2/neu and int-2/bcl-1, respectively was observed. HER2/neu amplification was negatively correlated with oestrogen receptor (ER) and progesterone receptor (PR) status (P less than 0.0001; for both), c-myc amplification was more prevalent in the PR-negative subpopulation (P less than 0.05) and int-2/bcl-1 amplification was positively correlated with ER status (P less than 0.001).