The purpose of this study was to evaluate a large number of patients with cutaneous melanoma who had or who were at risk for lymph node metastases to contribute to the understanding of the behavior of and appropriate management of draining nodes. A major goal of the study was to reassess the clinical impact of elective lymph node dissections (ELND) in a large patient population. Large retrospective studies suggest that ELND may improve the prognosis of patients with intermediate thickness melanomas; however, that improvement has not been observed in two randomized prospective controlled trials. The charts of 4682 patients treated at a single institution for localized or regional disease were reviewed individually. The median follow-up was 4.7 years, with 814 patients followed more than 10 years. The data were tabulated and evaluated with the aid of a computer data base system. Among patients with nodal metastases, 10% of nodal metastases were to contralateral nodes, and 6% were to nodal basins that would not be predicted by classic models of lymphatic drainage; in 13% of patients, nodal metastases occurred to greater than one nodal basin (3% of the entire study group). For all thickness ranges, the incidence of nodal metastases was comparable to the incidence of distant metastases; intermediate-thickness lesions had no relative predilection for nodal metastases. At the initial evaluation, regional nodal basins were clinically negative in 3550 patients, of whom 911 (25.7%) underwent ELND. Stratified into five thickness groups (< 0.76 mm, 0.76 to 1.5 mm, 1.5 to 2.5 mm, 2.5 to 4 mm, and > 4 mm), pathologically positive nodes were identified in 0%, 5%, 16%, 24%, and 36%, respectively (16% overall). Among the 911 patients who underwent ELND, 214 (23%) had nodal metastases, 143 at the time of ELND and 71 at a later date. Of these 71 patients, 31 (44%) had nodal metastases in a previously dissected nodal basin, and 40 (56%) had them in basins not previously dissected. The survival of patients with clinically negative nodes treated with and without ELND were compared. The two groups were well matched for major prognostic factors. Stratified by Breslow thickness and primary site, no significant improvement in survival was observed with ELND. Because of the significant incidence of metastases to contralateral and atypical nodal basins, lymphoscintigraphy may be justified for the preoperative evaluation of patients for ELND. However, the therapeutic value of ELND is questionable as a result of (1) the finding that the risk of nodal metastases is not relatively more common than is that of distant metastases among patients with intermediate-thickness melanomas, (2) the fact that only 16% of ELND were positive, (3) the finding that ELND may not prevent recurrent nodal disease in the dissected basin, and (4) the absence of any apparent impact on survival among patients who underwent ELND.