The purpose of the present study was to determine whether muscle synergies are constrained by changes in the mechanics of pedaling. The decomposition algorithm used to identify muscle synergies was based on two components: "muscle synergy vectors," which represent the relative weighting of each muscle within each synergy, and "synergy activation coefficients," which represent the relative contribution of muscle synergy to the overall muscle activity pattern. We hypothesized that muscle synergy vectors would remain fixed but that synergy activation coefficients could vary, resulting in observed variations in individual electromyographic (EMG) patterns. Eleven cyclists were tested during a submaximal pedaling exercise and five all-out sprints. The effects of torque, maximal torque-velocity combination, and posture were studied. First, muscle synergies were extracted from each pedaling exercise independently using non-negative matrix factorization. Then, to cross-validate the results, muscle synergies were extracted from the entire data pooled across all conditions, and muscle synergy vectors extracted from the submaximal exercise were used to reconstruct EMG patterns of the five all-out sprints. Whatever the mechanical constraints, three muscle synergies accounted for the majority of variability [mean variance accounted for (VAF) = 93.3 ± 1.6%, VAF (muscle) > 82.5%] in the EMG signals of 11 lower limb muscles. In addition, there was a robust consistency in the muscle synergy vectors. This high similarity in the composition of the three extracted synergies was accompanied by slight adaptations in their activation coefficients in response to extreme changes in torque and posture. Thus, our results support the hypothesis that these muscle synergies reflect a neural control strategy, with only a few timing adjustments in their activation regarding the mechanical constraints.