In this review, data is summarized supporting the hypothesis that axonal loss is a major pathologic process responsible for irreversible neurologic disability in patients with multiple sclerosis. Pathologic studies implicate inflammatory demyelination as a principal cause of axonal transection and subsequent axonal degeneration. Axonal degeneration caused by chronic demyelination in the absence of active inflammation may also contribute to progressive disability in the later stages of the disease. Studies using magnetic resonance spectroscopy suggest that axonal loss begins at the onset of the disease, and studies using magnetic resonance imaging have documented brain atrophy in the earliest stages of multiple sclerosis. Brain atrophy increases during the relapsing-remitting disease stage without concurrent disability progression. This suggests that compensatory mechanisms maintain neurologic function, despite progressive brain tissue loss during the early stages of the disease. Beyond a threshold, however, further axonal loss leads to continuously progressive neurologic disability. We hypothesize that the rate and extent of axonal loss during relapsing-remitting multiple sclerosis determines when a patient enters the secondary progressive stage of the disease. This view of disease pathogenesis has several important implications. First, surrogate markers of axonal loss are needed to monitor the disease process for patient care and for clinical trials. We propose brain parenchymal fraction, a precise measure of whole-brain atrophy, as an attractive candidate for this purpose. Second, disease-modifying therapy should be used early in multiple sclerosis patients, before extensive axonal loss has occurred. Third, neuroprotective drugs should be tested in combination with anti-inflammatory drugs in multiple sclerosis patients. Finally, studies of the time course of axonal loss, and its mechanisms are critical for effective therapeutic intervention.