Radiological assessment is a valuable tool in the assessment, management and prognostication of Perthes disease. Radiological assessment, however, is not an easy task and all classification systems used in Perthes disease have some degree of interrater and intrarater variabilities. In the past, there were some isolated studies to find the reliability of the classifications used in Perthes disease. In this study, we comprehensively studied three most commonly used radiological classifications (Salter-Thompson, lateral pillar and Catterall). We had 44 patients' radiographs (anteroposterior and lateral) taken in the fragmentation stage, and two experienced observers assessed and classified the radiographs on two separate occasions. In this study, we found that the average interrater reliability of the Salter-Thompson, lateral pillar and Catterall classifications was 0.163 (0.08-0.236), 0.722 (0.581-0.824) and 0.433 (0.280-0.546), respectively. The intrarater reliability was 0.313 and 0.699 for the Salter-Thompson, 0.707 and 0.658 for the lateral pillar and 0.38 and 0.577 for the Catterall classifications. Further, we tried to determine the possible reason for the low reliability associated with the Catterall classification. We think that the quantitative method of lateral pillar has better intrarater and interrater reliabilities than other classification systems, and the reliability of the Catterall classification can be significantly improved if some radiological parameters such as metaphyseal reaction and identification of the junction of involved to uninvolved region can be optimized.