Quantitative gait analysis using computer aided videomotion analysis, force plates and electromyography is of recognized value in assessment of gait disabilities and in quantitative evaluation of treatment. But, despite dramatic improvements in techniques, gait analysis still lacks widespread clinical utility because of uncertainties relating to data selection, manipulation, and analysis. Since data acquisition is usually limited to a single test session over several gait cycles, one vital problem concerns selection of a gait cycle for analysis that is representative of the patient's gait and determination of its reliability as a basis for clinical decision making. Another problem relates to the manipulation of large quantities of data generated by measurements of various kinetic, kinematic and electromyographic parameters over several gait cycles to detect clinically significant patterns of performance. This process is further complicated by the fact that gait patterns vary among patients with the same syndorme and to a certain extent, vary even among normals. Interpretration may be simplified by using statistical pattern recognition techniques, but for a pattern recognition approach to be successful, the enormous quantity of data must be reduced to a parsimonious set of features which describe gait patterns accurately. Furthermore, representation of graphic patterns associated with various gait parameters in terms of a discrete set of variables would make numerical comparison more meaningful. The first part of this study will investigate repeatability and clinical reliability of selected kinetic, kinematic and electromyographic parameters obtained from repeated gait analyses on normal subjects and two representative groups of orthopaedic patients, using a statistical criterion (variance ratio). The feasibility of using the variance ratio as a criterion for averaging gait cycles to yield a representative cycle also will be examined. The second part of the study will deal with the application of principal component analysis for representing gait patterns in a concise manner, as an initial step to application of pattern recognition techniques. The effectiveness of the derived features in representing gait abnormalities will be evaluated using the same normal and orthopaedic patient subjects. It is anticipated that the results of the proposed study will enhance the clinical effectiveness of quantitative gait analysis by defining practical computer algorithms for selecting and assessing reliability of raw gait data, and by defining techniques for feature selection that can be used effectively in pattern recognition techniques.