AbstractHuman movements are associated with the feeling that sensory consequences are self-generated (sense of agency). Though it has shown that the sense of agency for our movements is sensitive to the temporal coherence between our actions and their outcomes, previous research has focused on a particular temporal ordering, namely that movements (action) precede sensory consequences (outcome). Here, we wondered whether SOA could be felt for movements in the artificial case where this ordering is inversed; when sensory effects precede their corresponding movement. To test this, we predict voluntary motor movements (finger deflections) and provide visual sensory consequences that precede them (negative delay). Furthermore, in the same participants, we test the standard temporal ordering, investigating trials where motor movements precede their visual consequences (positive delay). By performing hundreds of trials per participant and utilizing methods from psychophysics, we fully characterize SOA as a function of the temporal offset between visuo-motor actions and effects, offering a comprehensive view of the dependence of SOA on the temporal coherence of action and effect.IntroductionMaterials and MethodsThe principle of the experiment is to let the participants perform a self-paced simple hand movement (index finger deflection) within a XX seconds trial and to show them, in place of their real hand, a virtual hand performing the same movement (3D animation) but at a random time within the same interval. Two cases can therefore occur: either the participant is first (movement onset precedes animation onset), either the animation is first (animation onset precedes movement onset). For each trial the participants answered a forced choice question about whether the movement they saw corresponded to the movement they made. The system was precisely tuned to guarantee an optimal visual correspondence between real and virtual hands, a precise recording of the timings, and to minimize the time interval between the two events by influencing the randomization of the animation onset time.ParticipantsFor the main experiment 10 healthy, right-handed participants were recruited (ages XX.X ± X.XXX mean±SD; XXX females). For the pilot behavioral study, we recruited an additional 14 healthy participants (XX handed; ages XX.X ± X.XXX mean±SD; XXX females). Both studies were undertaken in accordance with the ethical standards as defined in the Declaration of Helsinki and were approved by the local ethics research committee at the University of Lausanne. Material and procedure Participants sat at a table and placed their right hand underneath a computer monitor, holding a block containing a touch sensor. The monitor occluded vision of their forearm and, in correspondence with their real hand, displayed in stereoscopic 3D a virtual hand holding a virtual block (Fig. 2A). Head movements were restrained with a chin rest and the experiment took place in a darkened room.The touch sensor, an Arduino™ microcontroller with a 16MHz sampling rate, allowed for the detection at millisecond precision of the precise moment the index finger broke contact with a conductive surface when lifting (movement onset time). The 3D graphics were rendered using OpenGL on an nVidia Quadro 2000 graphics card using Quad-Buffer extension and nVidia 3DVision glasses for stereoscopic display on a XXX monitor at 120Hz. The constant latency of the graphics hardware was measured to be 30ms and accounted for in all estimations of the visual onset times (interval from the trigger time by the CPU to the actual visual onset of the event on the LCD monitor).In a short training block, participants learned to make index finger deflections that mirrored the velocity and amplitude of an animated finger in the virtual scene. During these 20 trials, no delay was introduced between movement and visual onset.The following question was then presented and participants were asked to remember it: “Did the movement that you did correspond to the movement that you saw?” No specific explanation was given about what could correspond or not. Answers to the question were given with the left hand by pressing buttons of a gamepad (index for yes, middle finger for no, reaction times ignored).For each trial of the actual experiment, the participants saw the virtual scene appear (T0) and were instructed to lift their finger at the time of their choosing (TM, movement onset time). The animation of the virtual finger occurred at a random time (TV, visual onset time). A question screen was displayed after both events have occurred and the trial ended after participants’ pressed a button to respond. If participants did not lift their finger before Tmax (XXs), the protocol moved forward and the trial was rejected. Timings (movement and visual onsets) and participants’ answers were logged for all valid trials.Finally, because using a purely uniform distribution for the randomization of the visual onset times (between T0 and Tmax) would require an extremely high number of trials for statistical analysis of data points with very close timings (TM and TV at less than 100ms apart), our system had a double strategy. First it tried to anticipate the movement onset time with the aim of providing visual consequences just prior to that moment by using a dynamic predictive algorithm based on per-subject movement history profiles. Second, if participants moved prior to the visual onset of the predictive algorithm, the visual consequence were presented with a delay shuffled within a uniform distribution in a small window of interest (0 to 750 ms).Experimental design and statistical analyses The first experiment consisted in 600 trials executed in 4 blocs, preceded by XX training trials. Here we manipulated the coherence (ΔT) between visual and motor events using a continuous design. To analyze these continuous data, we binned SOA responses for 20ms ΔT intervals. As our predictive algorithm used to anticipate movement onset, the number of trials for each bin was not balanced (Fig. ; Fig. SXX). To assess the electrical brain activity associated with the individual visual and motor events, two additional baseline condition blocks were recorded while participants saw the same virtual scene. In a first baseline condition, participants were instructed to relax while watching the fixation cross as the virtual finger moved (random time onset; vision only). In the second condition, participants were instructed to perform a voluntary finger deflection as in the main experiment, but no virtual visual counterpart accompanied the movement (motor only). Each experimental block consisted of XXX trials and was repeated XXX times, resulting in XXX trials. Pilot Study: Agency for virtual hands versus virtual objects· Refer to supplemental materials for our confirmation / “negative” finding that object vs body doesn’t matter? · It’s smartest if we save these extra findings for the last section in the results, briefly mention that we did it in the introduction, and describe here that it’s the same procedure as above, just with 300 trials with hand; 300 with object.