BackgroundChildren with Cerebral Palsy (CP) walk with an uncoordinated gait compared to Typically Developing (TD) children. This behavior may reflect greater muscle co-activation in the lower limb; however, findings are inconsistent, and the determinants of this construct are unclear. Research objectives(i) Compare lower-limb muscle co-activation during gait in children with, and without CP, and (ii) determine the extent to which muscle co-activation is influenced by electromyography normalization procedures and Gross Motor Function Classification System (GMFCS) class. MethodsAn electromyography system measured muscle activity in the rectus femoris, semitendinosus, gastrocnemius, and tibialis anterior muscles during walking in 46 children (19 CP, 27 TD). Muscle co-activation was calculated for the tibialis anterior-gastrocnemius (TA-G), rectus femoris-gastrocnemius (RF-G), and rectus femoris-semitendinosus (RF-S) pairings, both using root mean squared (RMS)-averaged and dynamically normalized data, during stance and swing. Mann-Whitney U and independent t-tests examined differences in muscle co-activation by group (CP vs. TD) and GMFCS class (CP only), while mean difference 95% bootstrapped confidence intervals compared electromyography normalization procedures. ResultsUsing dynamically normalized data, the CP group had greater muscle co-activation for the TA-G and RF-G pairs during stance (p < 0.01). Using RMS-averaged data, the CP group had greater muscle co-activation for TA-G (stance and swing, p < 0.01), RF-G (stance, p < 0.05), and RF-S (swing, p < 0.01) pairings. Muscle co-activation calculated with dynamically normalized, compared to RMS-averaged data, were larger in the RF-S and RF-G (stance) pairs, but smaller during swing (RF-G). Children with CP classified as GMFCS II had greater muscle co-activation during stance in the TA-G pair (p < 0.05). SignificanceGreater muscle co-activation observed in children with CP during stance may reflect a less robust gait strategy. Although data normalization procedures influence muscle co-activation ratios, this behavior was observed independent of normalization technique.