Three methods of precisely determining onset and cessation times of gait EMG were investigated. Subjects were 24 normal adults and 32 individuals with gait pathologies. Soleus muscle EMG during free speed level walking was obtained with fine wires, and was normalized by manual muscle test (%MMT). Linear envelopes were generated from the rectified, integrated EMG at each percent gait cycle (%GC) of each stride in individual gait trials. Three methods were used to generate EMG profiles for each tested subject. The ensemble average (EAV) was determined for each subject from the mean relative intensity of the linear envelopes. Low relative intensity or short duration EMG was removed from the ensemble average to create the intensity filtered average (IFA). The packet analysis method (PAC) created an EMG profile from the linear envelopes in successive strides whose respective centroid %GC locations were within +/- 15%GC of each other. Control values for onset and cessation times of individual gait trials were calculated after spurious outliers were removed. Mean onset and cessation times across subjects for control values and the experimental methods (EAV, IFA, and PAC) were calculated. Dunnett's test (p less than .05) was performed to compare control and experimental groups in patient and normal trials. EAV differed from control values for onsets (p less than .01), cessations (p less than .01), and durations (p less than .01) in both normal and patient trials. IFA and PAC had no significant differences from control value means. IFA was selected for clinical use as automatic analysis could be performed on all trials and a minimum number of decision rules were needed.