When many lymph nodes are found by using lymphoscintigraphic techniques performed to detect the sentinel lymph nodes (SLNs) in breast cancer, it is usual to find that the 'hottest' SLN is not always the node that is pathologically positive (pN+). Various criteria have been proposed to define which radioactive lymph nodes should be removed. In order to determine the frequency with which the hottest SLN 'fails' to be pN+, and to determine which criteria best define the radioactive lymph node to be removed, we reviewed and analysed our cases in which more than one SLN was detected and where there was also at least one pN+ node. From a series of 181 patients, 40 were selected. In 11 of these 40 cases (27.5%), the hottest SLN was not pN+. Radioactivity levels in the pN+SLN of these 11 patients ranged from 2% to 94% of the activity of the hottest SLN. Twenty-one patients (52.5%) showed only micrometastatic (pN1a) disease in one or more SLNs. In four of these patients (19%) the pN1a SLN was not the hottest node. Two of the patients had radioactivity levels in the pN+SLN which were more than 50% of that of the hottest SLN. In another two of these patients (9.5%), radioactivity levels were lower than 50% of that of the hottest node (respectively, 38% and 2%). However, in these two last cases, the first and hottest SLN removed surgically was found, by the pathologist, to consist of six nodes. Macrometastases (dimensions greater than 2 mm) were found in 19 patients. In 12 of these patients, the hottest SLN was macrometastatic although macrometastases and/or micrometastases were found in other 'cooler' SLNs in four of them. In another seven of these patients (36.8%), macrometastases were found in SLNs with radioactive levels lower than 51% of that of the hottest node. One patient (with three SLNs) out of the 40 (2.5%) had one SLN pN+ with less than 10% of that of the hottest. In fact, it contained only one micrometastasis and its activity was equal to 2%. Upon pathological examination, however, the hottest lymph 'node' was found to consist of six nodes. It is concluded that, with four intra-mammary and peritumoural injections of 99mTc labelled nanosized colloids of Human Serum Albumin (Nanocoll R: Sorin: 74 MBq and 0.05 mg per injection) performed 18-24 h before using a gamma probe to detect the SLNs, the hottest SLN was not the pathologically positive node in 27.5% of patients in our series. By using the activity in the hottest SLN as the reference point, and 10% of this activity as the lower threshold for removing active SLNs, the sensitivity of the technique is 97.5%.