We investigated the accuracy and validity of clinical gait assessment, performed by experienced geriatricians viewing video clips of 10 s duration. Nineteen patients with normal or characteristic abnormal gait patterns were studied. The treating physician's diagnosis served as the gold standard. Another live assessment was then performed by a geriatrician blinded to the medical record to establish inter-rater reliability of live assessments. Subsequently, each gait video clip was examined by two independent geriatricians without any background clinical documentation. Diagnostic accuracy was tested at two levels - whether the gait was abnormal, and the specific gait diagnosis. The agreement of the video clip examination with the gold standard to identify abnormal gait from normal gait ranged from substantial to excellent among assessors (kappa = 0.68-0.85), although low agreement with the gold standard was achieved in the detection of specific gait diagnosis (average agreement between both viewing geriatricians 50%). The technique appears to be a valid screening procedure for detecting gait abnormalities (average sensitivity 100%, specificity 70%).