Recently, a modular organisation has been proposed to simplify control of the large number of muscles involved in human walking. Although previous research indicates that a single set of modular activation patterns can account for muscle activity at different speeds, these studies only provide indirect evidence for the idea that speed regulation in human walking is under modular control. Here, a more direct approach was taken to assess the synergistic structure that underlies speed regulation, by isolating speed effects through the construction of gain functions that represent the linear relation between speed and amplitude for each point in the time-normalized gait cycle. The activity of 13 muscles in 13 participants was measured at 4 speeds (0.69, 1.00, 1.31, and 1.61 ms-1) during treadmill walking. Gain functions were constructed for each of the muscles, and gain functions and the activity patterns at 1.00 ms-1 were both subjected to dimensionality reduction, to obtain modular gain functions and modular basis functions, respectively. The results showed that 4 components captured most of the variance in the gain functions (74.0% ± 1.3%), suggesting that the neuromuscular regulation of speed is under modular control. Correlations between modular gain functions and modular basis functions (range 0.58-0.89) and the associated synergistic muscle weightings (range 0.6-0.95) were generally high, suggesting substantial overlap in the synergistic control of the basic phasing of muscle activity and its modulation through speed. Finally, the combined set of modular functions and associated weightings were well capable of predicting muscle activity patterns obtained at a speed (1.31 ms-1) that was not involved in the initial dimensionality reduction, confirming the robustness of the presently used approach. Taken together, these findings provide direct evidence of synergistic structure in speed regulation, and may inspire further work on flexibility in the modular control of gait.