Objectives: The aims of this study were to: 1. Evaluate the performance of arthroscopy for the diagnosis of chondropathy and to compare it to that of direct non-arthroscopic assessments; 2. Determine intra-observer reliability of arthroscopic assessments; 3. Evaluate the effects of the arthroscopic video quality and probing upon diagnostic performance. Design: The ovine medial meniscectomy (MMx) model of early osteoarthritis (OA) was used assuming that pre-MMx articular cartilage (AC) was "normal" and post-MMx AC "chondropathic". Video recordings of arthroscopic assessments of each stifle compartment were evaluated. Scores were given for the quality of the video and the amount of probing. The diagnostic performances of dynamic shear modulus (G), light microscopic assessment and superficial zone collagen birefringence assessments were evaluated and compared to that of arthroscopy. Intra-observer reliability of arthroscopic assessments was also evaluated. Results: Arthroscopic assessments had high sensitivity (91-100%), specificity (62-88%) and accuracy (75-93%) for the diagnosis of chondropathy 16 weeks after MMx. Arthroscopy compared favourably with the direct non-arthroscopic assessments in the lateral compartment and was found to have extremely high intra-observer reliability (kappa 0.78-1.00). The quality of arthroscopic video recordings and the amount of probing did not significantly influence accuracy or reliability. Conclusions: Arthroscopy performs as well as direct non-arthroscopic assessments of AC for diagnosis of early OA. These results suggest that arthroscopy can be used as a "gold standard" for the validation of non-invasive assessments like magnetic resonance imaging and that arthroscopic diagnosis can be based on small amounts of video footage without AC probing.