The proximal poles of five patients with presumed avascular necrosis of the scaphoid, mostly the result of nonunion, were analyzed histologically for true, pathologically confirmed avascular necrosis of bone. Each specimen was reviewed both qualitatively by an experienced bone pathologist and quantitatively with the use of a computerized image analysis system. Avascular necrosis was found to be variable from one specimen to another as well as within each specimen. Midcentral levels showed a statistically significant greater degree of avascular necrosis than either adjacent palmar or dorsal zones. Both viable and nonviable areas of avascular bone were found in an unpredictable, irregular (patchy) distribution, often in direct apposition to one another. It is apparent from this study that random biopsy alone cannot accurately predict the histologic status of the entire specimen; the clinician must rely on other methods to determine scaphoid viability.