We evaluated the importance of 14 clinical and pathologic variables as determinants of prognosis in patients with malignant melanoma and positive regional lymph nodes at axillary dissection. The records of 197 patients operated on between 1974 and 1984 were reviewed. Univariate analysis indicated as prognostically significant the number (p less than 0.001) and percentage (p less than 0.001) of positive nodes, highest nodal status (p less than 0.001), macroscopic or microscopic nodal metastases (p = 0.002), presence or absence of extranodal disease (p = 0.003), clinical stage (III versus less than III, p = 0.015), and site (considered as trunk versus other locations, p = 0.02). However, by multivariate analysis, only three variables were shown to be independent determinants of survival: percentage of positive nodes (p = 0.004), presence or absence of extranodal disease (p = 0.012), and site (trunk versus other locations, p = 0.019). Combining these three variables, subsets of patients with markedly different prognoses could be generated. It is possible to predict a favorable outcome for patients with less than 10% positive nodes, no extranodal disease, and a primary lesion at a site other than the trunk. It is also possible to recognize that the prognosis is very poor for patients with extranodal disease and truncal primary lesions, regardless of the percentage of positive lymph nodes. Finally it was verified that the prognosis is always unfavorable when the percentage of positive lymph nodes is very high.