To assess the reliability of bone scintigraphy, a random sample of 100 bone scans was reviewed twice by each of two physicians. Observer variation in the description and interpretation of bone scintigrams varied by diagnosis. Good to excellent k values were obtained for inter- and intraobserver variation in relation to metastasis or normal scans. For degenerative bone disease, as well as the specific agreement on major pathologies other than metastases, k values were found to be moderate. The agreement on the need for further radiographic studies was poor to moderate. The interpretation of bone metastases or normal scintigrams was found to be more reliable in a research setting than in the usual clinical framework, and the latter requires improvement. The interpretation of bone scintigraphy as consistent with degenerative changes is not reliable. The diagnosis should be evaluated by radiography.