The pattern of axillary lymph node involvement was analyzed in a review of 377 cases of T1-4 breast cancers. Clinical judgment of the axillary status proved to be wrong in approximately one-third of the cases. In univariate analysis, a strong correlation (P < .01) between the number of involved nodes, tumor size, and blood vessel invasion was found. Other features of the primary tumor (lymphatic invasion, degree of differentiation, presence of necrotic areas) were related to a lesser degree (P < .05). While others (age, site) were not at all significant. However, the number of nodes resected proved to be the most important determinant of all (P = .003). Also, the simple distinction between node-negative and node-positive cases is strongly dependent on the extent of axillary dissection (P = .009). In multivariate analysis, only the number of resected nodes and T stage showed a good relationship with the number of positive nodes. Skip metastases above levels 1 and 2 were seen in only 2% of the cases. A clear influence of the number of invaded nodes on survival could be demonstrated. These findings are discussed, especially as concerns the technique, prognostic significance, and therapeutic usefulness of axillary dissection.