Both in real life and experimental settings, increasing response speed typically leads to more error-prone actions. Processes underlying such a " speed-accuracy trade-off " (SAT) are usually assumed to be purely decisional: cautiousness would be determined only by the amount of sensory evidence required to select a response. The present data challenges this largely accepted view, by directly showing that motor processes are speeded up under time pressure. In a choice reaction time task where emphasis was put either on response speed or accuracy, motor processes were investigated through the analysis of muscular activity related to response execution. When response speed was emphasized , the time between electromyographic onset and behavioral response (motor time) was also speeded up (contributing for more than 20 % of the total effect on global reaction time). This speeded execution (likely due to a more efficient motor command) may also explain why participants are less able to interrupt incorrect response execution once started (Burle et al 2014), leading to more overt errors. Pointing to a speed-accuracy exchange within motor processes themselves, the present results call for a re-evaluation of widely accepted assumptions about SAT, and more generally, decision-making processes. They are discussed in the context of recent extensions of the drift diffusion model framework, questioning the strict separation between decisional and motor processes.