The purpose of this study was to determine the reliability of investigating electromyography (EMG) of selected leg muscles during walking. Tibialis posterior and peroneus longus EMG activity were recorded via intramuscular electrodes. Tibialis anterior and medial gastrocnemius EMG activity were recorded with surface electrodes. Twenty-eight young adults attended two test-sessions approximately 15 days apart. Relative and absolute measures of reliability were calculated for EMG timing and amplitude parameters during specific phases of the gait cycle. Maximum contractions and sub-maximal contractions were obtained via maximum isometric voluntary contractions and a very fast walking speed, respectively. Time of peak EMG amplitude for all muscles displayed relatively narrow limits of random error. However, reliability of peak and root mean square amplitude parameters for tibialis posterior and peroneus longus displayed unacceptably wide limits of random error, regardless of the normalisation reference technique. Whilst some amplitude parameters for tibialis anterior and medial gastrocnemius displayed good to excellent relative reliability, the corresponding values for absolute error were generally large. Timing and amplitude EMG parameters for all muscles displayed low to moderate coefficient of variation within each test session (range: 7-25%). Overall, between-participant variability was minimised with sub-maximal normalisation values. These results demonstrate that re-application of electrodes results in large random error between sessions, particularly with tibialis posterior and peroneus longus. Researchers planning studies of these muscles with a repeated-test design (e.g. to evaluate the effect of an intervention) must consider whether this level of error is acceptable.