This paper presents statistical tools to assess reliability in gait analysis. Gait analysis involves measuring continuous curves over a gait cycle. Three main problems are encountered in clinical practice: (1) determining if a patient's gait curves belong to a given population, (2) classifying a new subject, and (3) comparing two populations. The paper introduces intra-class correlation coefficients to measure reliability, confidence bands to define population curves, and analysis of variance with confidence bands to compare population means. These tools can help clinicians interpret gait data and determine if a patient's gait is pathological or normal.