Résumé : During penalty shootouts, a soccer goalie knows the probabilities associated to each player’s shooting side preference. Together with sensory information, this previously cued knowledge is used by the goalie to decide towards which side to dive in order to catch the ball. In this situation, deciding what action (i.e. which side to dive for) to perform has traditionally been explained within the framework of information processing (e.g. Sternberg, 2011). In this view, the soccer goalie would, based on sensory input and probability knowledge, first compute a decision in prefrontal areas, which would subsequently act as input to motor processing yielding the observed action (the dive). Importantly, this view suggests that relevant task factors (e.g. probability of an upcoming action) are integrated in the decision process through a cost/benefit computation implemented at the OFC/vmPFC level (Hare, Camerer, Knoepfle, & Rangel, 2010; Padoa-Schioppa, 2011; Wallis & Miller, 2003). Other authors have emphasized an ‘embodied’ view on decision-making (Cisek & Kalaska, 2010; Shadlen, Kiani, Hanks, & Churchland, 2008). This evolutionary perspective suggests the involvement of sensorimotor areas in decision-making. Stemming from single-cell recordings in monkey studies and formalized in the affordance competition hypothesis (Cisek, 2007), this competition is suggested to be limited to premotor areas. Recent models of human action selection advocate such a view (e.g. Domenech & Koechlin, 2015). However, in humans, it is not clear up to what level this competition takes place. To formally address this question in humans, we designed a functional Magnetic Resonance Imaging (fMRI) experiment. Participants were shown cues indicating the probabilities associated with upcoming button presses. If action selection (i.e. pressing the left or right button) is performed through a biased competition in sensorimotor regions, these same regions are expected to integrate relevant task factors (e.g. probability cues). Our results show that the biased competition induced by probability cues is represented up to the primary motor cortex. Specifically, activity in the contralateral primary motor cortex increased with action probability; instead, activity in the ipsilateral motor cortex decreased. This finding strongly suggests that human decision-making derives from a biased competition taking place up to the primary motor cortex.