The expression of p53 protein was examined in a series of 136 primary breast carcinomas, 106 of which were analysed with a panel of four monoclonal antibodies (MAbs 1801, 240, DO7 and DO1). p53 expression was detected with at least one antibody in 40 tumours (38%), whereas only 15 tumours (14%) were positive with all four antibodies. Some variability in the immunostaining could be observed depending on the antibody used. This was noticeable both for the number of positive cells within a section and for the intensity of staining. We therefore selected a panel of 17 tumour sections (nine were highly positive, three with medium to low staining and five with low to negative staining), which we analysed by polymerase chain reaction-single-strand conformation polymorphism analysis (PCR-SSCP) for the presence of a p53 mutation at the molecular level. Mutations were identified in 15 cases. Therefore the proportion of p53-stained cells does not seem to be an exact representation of the number of cancer cells bearing a mutation within a tumour. A statistically significant correlation was observed between p53 expression, regardless of the number of positive antibodies, and grade III disease (P < 0.0001), oestrogen (P < 0.0001) or progesterone receptor negativity (P = 0.0061), increased Ki 67 index (P = 0.0018), epidermal growth factor receptor (EGFR) positivity (P = 0.0076) and aneuploidy (P = 0.037). No correlation was observed with tumour size or lymph node involvement. In univariate analysis p53 expression was not correlated with disease-free survival, in contrast to the classical prognostic parameters, which were statistically correlated. In this series p53 expression was not a marker of early recurrence.