The aim of this study was to assess the efficacy of electromyography (EMG) normalization methods for a high-speed 20-m sprint. Comparisons were based on intra-individual reliability and magnitude of normalized EMG signals from three repeat sessions separated by 1 day (between days) and 1 week (between weeks) from the initial test. Surface EMGs were recorded (n=16) from the medial and lateral gastrocnemius and soleus during the normalization methods (isometric: maximum/sub-maximum/body weight; isotonic: maximum/sub-maximum/body weight; isokinetic: 1.05 rad · s–1, 1.31 rad · s–1, 1.83 rad · s–1; squat jump). The EMG data from the 20-m sprint were normalized using each method and using the within-sprint peak EMG (sprint peak). Intra-individual reliability of the EMG was assessed using typical error of measurement as a percentage of intra-individual coefficient of variance (TEMCV%). Sprint peak normalization improved intra-individual reliability of EMG (soleus: <4.91CV%; medial gastrocnemius: <6.2CV%; lateral gastrocnemius: <7.1CV%) compared with un-normalized EMG (soleus: <13.3CV%; medial gastrocnemius: <16.5CV%; lateral gastrocnemius: <16.3CV%) both between days and between weeks. Squat jump normalization improved the soleus (<11.2CV%) and medial gastrocnemius (<15.7CV%) reliability between days and weeks and provided a representative measure of triceps surae muscle activation. The intra-individual reliability of the medial gastrocnemius EMG data was improved both between days and weeks when using isotonic normalization. Isometric and isokinetic normalization showed no improvement in intra-individual reliability either between days or weeks for any muscle. The method of normalization influenced the between-stride muscle interaction during the 20-m sprint. The results of this study suggest that peak normalization can be used to normalize high-speed muscle actions, while normalizing EMG to a squat jump may provide an alternative method to represent relative muscle activation.