Slides of 198 patients with primary malignant melanoma have been reevaluated independently by three pathologists; considered parameters were histologic type, Clark's level and tumor thickness. The agreement between pathologists is excellent, especially for histologic type and thickness. The use of the Cox's regression model, dealing with disease free interval as response criterion, shows that the main prognostic factor is tumor thickness, whatever the pathologist. A prognostic staging is thus proposed, based only on the tumor thickness, less than 2 mm, greater than 2 mm. The prognostic value of this staging is confirmed on a new sample of 145 patients. This classification is used as decision tool in a new randomized clinical trial.