Alexander disease is a rare but often fatal disease of the central nervous system. Infantile, juvenile and adult forms have been described that present with different clinical signs, but are unified by the characteristic presence in astrocytes of Rosenthal fibers-protein aggregates that contain glial fibrillary acidic protein (GFAP) and small stress proteins. The chance discovery that mice expressing a human GFAP transgene formed abundant Rosenthal fibers suggested that mutations in the GFAP gene are a cause of Alexander disease. Sequencing results from several laboratories have indeed now identified GFAP coding mutations in most cases of the disease, including both the infantile and juvenile forms. These mutations have been found in the 1A, 2A and 2B segments of the conserved central rod domain of GFAP, and also in the variable tail region. All changes detected are heterozygous missense mutations, and none has been found in any parent of a patient that has been tested. This indicates that most cases of Alexander disease arise through de novo, dominant, GFAP mutations. Many of these mutations are homologous to ones described in other intermediate filament diseases. These other diseases have been attributed to a dominant loss of function, as the intermediate filament network is usually disrupted and a similar phenotype is observed in mice in which the corresponding intermediate filament gene has been inactivated. However, astrocytes of Alexander disease patients have normal appearing intermediate filaments, and GFAP null mice do not display the symptoms or pathology of Alexander disease. Thus, Alexander disease likely results from a dominant gain of function. Drawing upon the homology of many of the Alexander disease mutations to those found in other intermediate filament diseases, it is suggested that the gain of function is due to a partial block of filament assembly that leads to accumulation of an intermediate that participates in toxic interactions.