To evaluate the intra- and inter-observer measurement variability of an existing osteoarthritis (OA) stifle scoring system.  Paired caudocranial and mediolateral canine stifle radiographs were selected randomly. A total of 15 assessment points were evaluated independently and graded twice (integer numeric scale: 1-4) at an interval of 2 weeks by three observers with different levels of experience. The grades for each of the 15 factors were summed to obtain the OA score for each patient.  The 15 independent assessment points measured by the three observers showed high reproducibility and low intra-observer variability. Inter-observer variability was also low (mean: 1.09 ± 4.99, 95% CI [confidence interval]: -0.35 to 2.55). The most discordant ratings among the three observers involved sesamoid bones of gastrocnemius muscle (assessment point 11 of 15) and popliteal surface of femur (assessment point 10 of 15).  A validated and feasible OA scoring method is prerequisite for reliable radiographic assessment of OA progression. The low overall inter- and intra-observer variabilities among the 15 independent measures of the OA scoring system presented herein support its feasibility for application in clinical practice as an objective tool for radiographic scoring of stifle OA.