In schizophrenia, passivity phenomenon are clinically related to an abnormal sense of agency, which has been experimentally studied through self-recognition tasks. However, Tsakiris et al. (Cognition 96(3):215–231, 2005) have recently shown in healthy controls that the sense of agency is distinct from self-recognition abilities. We propose a simple motor task to obtain an implicit indicator of the working status of the pre-reflective sense of agency in schizophrenia. Collision dynamics gave us the means to further dissociate agency from motor prediction. Twenty-four patients and a group of matched controls used a hand-held object to stop the fall of a pendulum that was released either by the Subject (task S) or by the Experimenter (task E). The objective indicator of the sense of agency was taken as the efficiency difference between tasks S and E, before the availability of afferent information from collision. Qualitative feedback was provided to assess the top-down effect of explicit information. Motor prediction was as accurate in patients as in controls in tasks E and S. Controls were more efficient in S than in E. Patients revealed similar efficiency levels in both tasks. Qualitative feedback helped but did not affect the efficiency difference between tasks. Our results suggest an impairment of a pure efferent-driven sense of agency in schizophrenia, which is (1) distinct from motor prediction and (2) not under voluntary control. The abnormal judgments previously reported in schizophrenia for self-recognition abilities might be the consequence of a low order deficit of a pre-reflective sense of agency.