To evaluate the usefulness of recently proposed schemes of classification for systemic sclerosis an extensive cross-sectional study of a series of 164 consecutive patients with long-term systemic sclerosis was undertaken. There were 47 cases of proximal sclerosis, 93 of distal sclerosis and 24 of complete CREST syndrome. The study included clinical, visceral, immunological and follow-up data. In addition, a quantitative clinical score was calculated for each patient, thus providing indications for prognosis. Data were expressed according to three conventional systems of classification: The ARA system, the diffuse versus limited systemic sclerosis system and the early cutaneous involvement system. The most reliable indications of severe outcome were: proximal sclerosis, trunk skin involvement, presence of anti Scl 70 autoantibody, pulmonary and/or heart involvement and age. Diagnosis and prognosis were not generated by the same items. Prognosis indicators proved more accurate for groups than for individuals. Mortality was 1 death per 149 patient X years of follow-up from diagnosis. We conclude that the ARA criteria for classification should be recognized as a standard, but patients with complete CREST syndrome should be included in the distal group. Other systems of classification, principally 2-way versus 3-way criteria, allow different subsets of patients that correlate with prognosis and the severity of the disease, and could be used for therapeutic purposes.