Two hundred fifty-four patients with definite multiple sclerosis were followed up prospectively for 1 to 5 years (mean, 2.6 years). None of the patients received immunosuppressive medication. Yearly exacerbation rates and each patient's adherence to initial disease type were determined. Disease type was defined at entry and prospectively each subsequent year as stable, relapsing remitting stable, relapsing remitting progressive, or chronic progressive. Exacerbation rates determined prospectively did not decline significantly during 3 years of follow-up, even if patients were stratified by disease duration. Adherence to the initially assigned disease type was highly variable. When followed up for 2 years, 30% of patients with chronic progressive disease had conditions become stable, 32% of patients with stable disease had conditions become chronic progressive, 20% of patients with relapsing remitting disease had conditions become stable, and 20% of patients with relapsing remitting disease had conditions become chronic progressive. Patients with stable or relapsing remitting stable disease switched to one of the progressive categories as frequently (44%) as patients with progressive disease stabilized (46%). Progression of disease measured by changes in Kurtzke Expanded Disability Status Scores did not differ between the different disease types. The results challenge dogma regarding the natural history of exacerbation rates and the assumption that we can reliably assign patients to a specific disease type. The findings have important implications for understanding the natural history of multiple sclerosis and designing clinical trials.