Objective To identify the effect of the repeatability of muscle activations on extraction of consistent motor modules across trials. Methods The activities of sixteen muscles in twelve subjects who consistently walked at a self-selected speed were recorded. The intraclass correlation coefficient (ICC) was used to identify inter-stride repeatability of muscle activities and motor modules. Based on the repeatability of muscle activation, three types of muscle sets were organized. Results The reliable set containing the muscles showed high ICC (>0.4), but the whole-body and mixed sets containing the muscles showed poor ICC (＜0.4). When motor modules were extracted from each set, the reliable set showed the highest repeatability of motor module extraction, but the whole-body and mixed sets presented significantly lower repeatability. Conclusions Greater repeatability of muscle activations result ed in consistent motor modules. Extraction of consistent motor modules was a critical issue, especially in real-time motion recognition based on muscle patterns.