An accurate interpretation of radiographs is crucial in determining therapeutic choices in Legg-Calvé-Perthes disease. The aim of this study was to measure the interobserver reliability of a group of surgeons in this condition. Twenty-three radiographs were shown to nine pediatric orthopedic surgeons at nine different medical centers. Differences in coding between members of the group was assessed using a two-way random factor analysis of variance. Reliability was excellent for the Catterall classification [intraclass correlation coefficient (ICC) = 0.94], and good for the stage of disease and Head at Risk signs (ICC ranging from 0.74 to 0.85) except for lateral subluxation, whose reliability is fair (ICC = 0.68). Analysis of results shows that lateral subluxation should be expressed in a quantitative measurement rather than a binary answer. The reliability of Gage's sign would be improved if multiple definitions did not exist in the literature. In experienced hands, Catterall's classification can be used with reliability. Difficulties are still noted in separating group II from group III.