Abstract The validity of methods used to score the severity of radiographic findings (RFs) in horses is uncertain since only one or two joints are usually studied, classification criteria are heterogeneous and the internal validity is not assessed. The aim of this study was to assess the internal validity of a severity scoring system (SSS) of RFs by repeated scoring of a sample of radiographs. This SSS of RFs is based on four criteria that can be applied to every type of RF observed in limb joints. It consists of five weighted severity indexes (0, 1, 2, 4, 8) and was used to assess RFs found on the limbs of 392 young horses. The internal validity of the SSS was assessed using Kappa coefficients calculated on a subsample of 137 horses whose radiographs were interpreted twice. The final RF severity indices from these radiographs were obtained after three experienced veterinarians had reached a consensus, similar to the procedure used at foal and yearling sales. The majority of RFs from the 392 horses were scored either 1 or 2, while scores of 8 were only observed in the stifle and tarsus. Among the subsample, the overall agreement on the presence or absence of RFs was good (?=0.63; 95% Confidence Interval [CI], 0.56-0.69), and was excellent for the severity of RFs (weighted ?=0.82; 95% CI, 0.75-0.87). Most disagreements involved RFs that scored 1. The fore fetlock and the carpus experienced the lowest agreement. The SSS was a stable and reliable procedure applicable to any RF on any limb joint of the horse. It will be of potential interest in clinical practice and in the pre-purchase evaluation of young horses and could also be used in additional studies on the evolution or risk factors of RFs.