Routine mammography (MX) practice based on self-referral is considered not to be susceptible to monitoring at the population level. Invasive breast cancer cases notified to the Romagna Cancer Registry between 1989 and 1991 (total 654) were stratified by MX experience and were compared for the proportion of lesions greater than 2 cm in size (T2+) with cases observed in the study group of the Swedish Two Counties Screening Trial (TCST). Any MX more than 6 months prior to diagnosis was considered a previous routine examination. The proportion of cases registered in Romagna within 6-11 months of the most recent MX was quite similar to that shown by the interval cases of the TCST (50% vs 53%). T2+ cases registered in Romagna 12-23 months (27%) and 24-35 months (19%) after previous MX were similar in frequency to those detected at second and later screens in the TCST (22%). For longer intervals, the percentage of large tumours in the Romagna series increased, up to 45% among the cases registered more than 48 months since last MX (p < 0.0001). The whole group of cases with previous MX in Romagna and its counterpart in the TCST (i.e. interval cases + cases detected at second and later screens) had the same size distribution (35% vs 34%). The frequency of large tumours in the total series from Romagna was greater than in the total study group of the TCST (45% vs 36%, p < 0.0001). The difference was entirely accounted for by the cases with no previous MX.