Attenuation of visual reafferent signals in the parietal cortex during voluntary movement

The present work shows that early visual event-related potentials over the parietal cortex are attenuated when vision of the moving limb is presented in real time compared with when it is experimentally delayed. These findings bridge a gap in the literature, because they show that akin to somatosensory and auditory signals, sensory predictions associated with descending motor commands influence the magnitude of the cortical response to visual feedback.

Abstract

It is well established that the cortical processing of somatosensory and auditory signals is attenuated when they result from self-generated actions compared with external events. This phenomenon is thought to result from an efference copy of motor commands used to predict the sensory consequences of an action through a forward model. The present work examined whether attenuation also takes place for visual reafferent signals from the moving limb during voluntary reaching movements. To address this issue, EEG activity was recorded in a condition in which visual feedback of the hand was provided in real time and compared with a condition in which it was presented with a 150-ms delay, thus creating a mismatch between the predicted and actual visual consequences of the movement. Results revealed that the amplitude of the N1 component of the visual event-related potential evoked by hand visual feedback over the parietal cortex was significantly smaller when presented in real time compared with when it was delayed. These data suggest that the cortical processing of visual reafferent signals is attenuated when they are correctly predicted, likely as a result of a forward model.

NEW & NOTEWORTHY

The present work shows that early visual event-related potentials over the parietal cortex are attenuated when vision of the moving limb is presented in real time compared with when it is experimentally delayed. These findings bridge a gap in the literature, because they show that akin to somatosensory and auditory signals, sensory predictions associated with descending motor commands influence the magnitude of the cortical response to visual feedback.

the cortical processing of sensory feedback is influenced by whether it results from self-generated movements or from external events, such as passive movements. To account for this phenomenon, it is thought that the brain uses an efference copy of descending motor commands to predict the sensory consequences of an action through a forward model (Shadmehr and Krakauer 2008; Von Holst 1954; Wolpert and Miall 1996). According to this account, when the actual sensory consequences of the movement match the sensory predictions, they are attenuated.

Sensory attenuation has been extensively studied in the somatosensory domain (Bays et al. 2005; Bernier et al. 2009; Blakemore et al. 1998b, 1999a, 2000; Brooke et al. 1997; Claxton 1975; Shergill et al. 2013; Weiskrantz et al. 1971). Indeed, studies using functional magnetic resonance imaging (fMRI) showed that the blood oxygen level-dependent (BOLD) response in the somatosensory cortex is reduced when tactile stimuli are caused by one's own movement compared with when they are externally triggered (Blakemore et al. 1998b; Shergill et al. 2013). This attenuation is thought to underlie the percept that people cannot tickle themselves (Claxton 1975; Weiskrantz et al. 1971). Furthermore, it was found that if tactile reafferent signals are temporally delayed through experimental manipulation by ∼100 ms, sensory attenuation is alleviated and perceived sensation is consequently increased (Bays et al. 2005; Blakemore et al. 1999a; Shergill et al. 2013).

Sensory attenuation is also ubiquitous in the auditory domain (Bäss et al. 2008; Curio et al. 2000; Hirano et al. 1996, 1997a, 1997b; Houde et al. 2002; Hughes et al. 2013; Lange 2013; Wang et al. 2014). Indeed, numerous electroencephalography (EEG) and magnetoencephalography (MEG) studies have shown that early event-related potentials (ERP) generated in the auditory cortex are suppressed when people hear their own voice while talking compared with simply listening to a tape playback of it (Houde et al. 2002; Wang et al. 2014). It is thought that a copy of the motor command to speak is transferred from the frontal lobe to the auditory cortex, where it is used to attenuate the auditory response to the spoken sound as represented by the reduced ERP amplitude (Lange 2013; Wang et al. 2014).

In the visual domain, sensory attenuation has been investigated in the context of eye movements, with reduced visual sensitivity during the saccade (i.e., saccadic suppression; Bremmer et al. 2009; Krock and Moore 2014). Other studies have probed the cortical processing of visual stimuli triggered by voluntary movements. For instance, Hughes and Waszak (2011) compared EEG activity evoked by a visual grating that was either triggered by a participant's finger movement or triggered externally, and they found that the former led to reduced activation in a frontoparietal network. Similarly, Roussel et al. (2014) showed that after participants were trained to associate left- and right-hand keypresses with the appearance of specific letters, the amplitude of the N1 component of the visual ERP was significantly attenuated during congruent action effects (appearance of the expected letter) compared with incongruent action effects (appearance of the unexpected letter). Although this line of work clearly shows that visual attenuation can occur when action effects are correctly anticipated, many of these studies have used visual stimuli that were “arbitrarily” linked to actions (i.e., a grating triggered by a button press), rather than the actual visual feedback of the limb in motion (but see Leube et al. 2003). In this regard, the neural mechanisms mediating the cognitive association between an action and its perceptual effect, framed in the theory of ideomotor learning (e.g., Elsner and Hommel 2001; Greenwald 1970; Herwig et al. 2007; Prinz 1997; Waszak et al. 2012), may differ from those underlying the sensory reafferent predictions based on an efference copy of descending motor commands. For instance, the anticipation of action effects has been shown to implicate primarily the medial prefrontal cortex (Nachev et al. 2008; Waszak et al. 2012). In contrast, sensory reafferent predictions involve forward modeling through cerebellar circuitry, necessary to account for the mechanical properties of the musculoskeletal apparatus (Blakemore et al. 1998a, 1999b; Shadmehr and Krakauer 2008; Wolpert and Miall 1996). Furthermore, unlike stimuli such as gratings or letters, visual feedback from the moving limb is functionally related to the control of an ongoing movement, which may render attenuation suboptimal (see Lebar et al. 2015).

As a result, it remains unclear whether the actual visual reafferent signals from the moving limb are also attenuated at the cortical level during voluntary arm reaching movements. The present work addressed this issue by recording EEG activity during a reaching task in which visual feedback of the hand was provided either in real time or with a 150-ms delay, thus creating a temporal mismatch between expected and actual visual reafferent signals. Based on what has been shown in the somatosensory domain, it was hypothesized that early visual ERPs would be attenuated when visual feedback of the hand is provided in real time compared with when it is delayed.

MATERIALS AND METHODS

Participants

Fourteen healthy participants (4 women), between 18 and 31 yr old (22.9 ± 3.7 yr) took part in the experiment. They were naive as to the purpose of the experiment. All participants were right-handed, had normal or corrected-to-normal vision, and had no neurological disorders. All participants read and signed consent forms validated by the ethical committee of the Centre Hospitalier de l'Université de Sherbrooke. They were encouraged to ask questions if any part of the consent form was unclear.

Apparatus

The experimental setup consisted of a table supporting a computer monitor that projected visual stimuli on a semireflective mirror, preventing participants from seeing their hand (Fig. 1, A and B). The monitor (20-in. Dell P1130; resolution: 1,024 × 768; refresh rate: 150 Hz) was mounted face down 29 cm above the horizontal mirror. The mirror itself was mounted 29 cm above the table. With this setup the visual stimuli appeared to be projected directly onto the surface of the table on the same plane as the hand. The starting base consisted of an L-shaped piece of aluminum fixed to the table. It was located 15 cm in front of the participant's chest along the midline and was defined as coordinates (0, 0). A two-joint planar manipulandum was placed on the table and was held by participants via a stylus located at its mobile end. The manipulandum was custom-built with two lightweight metal rods (48 and 45 cm, respectively), with the fixed end attached to the upper left side of the table. A thin sheet of smooth plastic was put on the table surface, and foam pads were installed under the hinges, allowing the manipulandum to be moved everywhere on the table with minimal inertia and friction.

Fig. 1.

Experimental apparatus and time course of task events. A: lateral view of the participant and the experimental apparatus. B: overhead view of the participant and the experimental apparatus. C: a go cue (auditory tone) instructed participants to initiate the reaching movement 1,500 ms after the hand had returned to the starting base. In the real-time condition, the visual cursor corresponding to the hand was provided in real time. In the delayed condition, it was provided at the same physical location but with a 150-ms delay.

Two potentiometers positioned in the joints of the manipulandum allowed us to measure the angle of each segment at 1,000 Hz, from which we estimated the kinematics of the stylus in the X (left, right) and Y (near, far) dimensions. This information was then used to project a cursor corresponding to participant's hand in real time (see below). During recording, raw kinematic data were spatially smoothed with a Kalman filter to estimate hand position in real time. With this procedure, the total time necessary to collect the X and Y coordinates of the hand and present the corresponding visual cursor in real time was estimated to be ∼7–9 ms, as determined in separate piloting using a high-speed camera.

A single target was used, consisting of cyan circle of 0.75-cm diameter positioned 20 cm straight ahead of participants along their midline. The hand visual cursor consisted of a green circle of 1-cm diameter. The starting base where participants had to return after each trial was depicted by a pink 0.5-cm × 0.5-cm diamond.

Experimental Task

Participants had to perform reaching movements with the right hand toward the visual target. They were asked to produce straight movements with no discrete corrective submovements in a movement time of ∼800 ms. This targeted movement time was chosen so that there would be enough time for the online processing of hand visual feedback in both the real-time and delayed conditions, making vision relevant in both conditions. The experimenter continuously monitored movement kinematics and provided participants with verbal feedback between the experimental blocks if needed. Participants were asked to fixate the target at all times throughout the experiment. This was done to minimize ocular artifacts in the EEG recordings and to ensure that the retinotopic location of the cursor being presented would be similar across trials. Before the experiment, participants trained for ∼20 min to become familiarized with the task and produce consistent movement times and kinematics while keeping fixation. Participants reported no difficulty complying with the requirement to fixate the target, which constitutes the natural behavior during goal-directed reaching.

The visual target and the starting base were permanently shown throughout the experiment. Before each trial, participants were asked to bring the cursor to the starting base, at which point it was extinguished. After a 1,500-ms delay, an auditory tone instructed participants to initiate a reach (Fig. 1C). The first 5 cm of the movement trajectory were produced without vision of the cursor. The main experimental manipulation was the timing at which the hand visual cursor was presented. In the real-time condition, the hand visual cursor was provided in real time. Specifically, when the hand crossed a 5-cm radius (i.e., 15 cm away from the target), the cursor was shown for 150 ms (corresponding to a trace of ∼6 cm in the Y direction). In the delayed condition, the hand visual cursor was provided at the same physical location as in the real-time condition (i.e., crossing of a 5-cm radius with a trace of ∼6 cm), but with a temporal delay of 150 ms. Importantly, given that only the timing of cursor presentation was manipulated and not its physical location, it was possible to directly compare the cursor-evoked EEG response in the two conditions. A suddenly appearing cursor was chosen instead of a continuously presented cursor to evoke a more phasic EEG response. At movement's end, participants were asked to remain stationary for 300 ms, at which point their final position was indicated with a red dot, prompting them to come back to the starting base.

The real-time and delayed conditions were carried out jointly in six blocks. Each block comprised 80 trials in the real-time condition and 20 trials in the delayed condition, presented in pseudorandom order. The greater proportion of trials in the real-time condition was meant to establish an expectation that the hand visual cursor would be presented in real time. It was reasoned that establishing strong temporal expectancies would maximize the likelihood of capturing differences in cortical responses when visual feedback would be delayed. Participants were not informed of the manipulation of cursor timing, and verbal debriefing after the experiment confirmed that none of them consciously perceived it. Overall, the entire experiment comprised 480 trials in the real-time condition and 120 trials in the delayed condition and lasted ∼2 h. Given the known sensitivity of visual ERPs to retinotopic features of the stimulus (Clark and Hillyard 1996; Hillyard and Anllo-Vento 1998), the large number of trials allowed us to have stringent rejection criteria for reaction time, movement time, and endpoint accuracy so that the EEG analysis would be carried out on trials with very similar movement kinematics.

Kinematics Data Acquisition and Processing

The onset of the movement was defined as the moment the manipulandum left the starting base. The end of the movement corresponded to the moment when the Y velocity first crossed the X-axis on the velocity profile (0 m/s).

Reaction time.

Reaction time (RT) was calculated as the time between the auditory tone and movement onset. Trials with RT <100 ms and >600 ms were rejected. This stringent rejection criterion minimized variability in participant's vigilance/attentional state, which is known to influence ERP amplitude (Brandt and Jansen 1991). On average, this threshold led to the rejection of 38 ± 10 trials per participant in the real-time condition and 9 ± 2 trials in the delayed condition (8% of the data in each condition). The corresponding EEG epochs were removed.

Movement time.

Movement time (MT) was calculated as the time between movement onset and movement end. Each participant's mean MT was calculated over all trials and those for which MT was beyond ±2 SD were excluded. On average, this threshold led to the rejection of 16 ± 2 trials per participant in the real-time condition and 4 ± 2 trials in the delayed condition (3% of the data in each condition). The corresponding EEG epochs were removed.

Endpoint position.

Endpoint position was defined as the X and Y coordinates corresponding to movement end. From these coordinates, we measured X and Y accuracy with respect to the target. X accuracy corresponded to the X coordinate at movement endpoint minus the target X coordinate. Y accuracy corresponded to the Y coordinate at movement endpoint minus the target Y coordinate. The endpoint X and Y coordinates were also used to calculate a radial error, which was used to reject outliers. When radial error was >3 cm, we excluded the trial from further analysis. This stringent cutoff was justified by the need for similar cursor kinematics to compare the ERPs reliably. On average, this threshold led to the rejection of 5 ± 1 trials per participant in the real-time condition and 1 ± 1 trials in the delayed condition (1% of the data in each condition). The corresponding EEG epochs were removed.

EEG Recording and Processing

EEG data were acquired with a 64-channel BrainAmp system (Brain Products, Munich, Germany) along with the BrainCap electrode cap (Falk Minow Services, Herrsching-Breitbrunn, Germany). When placing the cap, we made sure that the Cz electrode was at the vertex. The electrodes were ring-type sintered nonmagnetic Ag-AgCl electrodes and were positioned in accordance with the extended 10/20 system. The reference electrode was located at FCz. Vertical eye movements and blinks were monitored with frontal electrode FP1 (positioned above the left orbit). The EEG signals were digitized online (sampling rate 5 kHz), and impedances were kept below 20 kΩ.

EEG data were analyzed offline using the BrainVision Analyzer software (version 2.0; Brainproducts). Data were downsampled to 250 Hz. They were digitally bandpass filtered offline (0.5–100 Hz, notch at 60 Hz, 12 dB/octave) and transformed to the average reference (Gwin et al. 2010; Gwin and Ferris 2012a, 2012b). Participants were encouraged to delay their blinks until the intertrial interval, yet remaining ocular artifacts (as monitored with electrode FP1) were subtracted from the EEG signal using the statistical method of Gratton et al. (1983). Given that the signal at electrode FP1 becomes null after this subtraction procedure, it was not included for further analyses.

The data were epoched within a time window between −2 s and +1.5 s around the onset of the visual cursor in both conditions. In other words, for the real-time condition, the 0-ms time point corresponded to the moment when the hand crossed the 5-cm radius, whereas in the delayed condition, it corresponded to 150 ms after the hand crossed the 5-cm radius. The epochs were visually inspected and removed if they contained stereotypical artifacts such as large deflections or bursts of EMG activity. On average, 41 ± 7 epochs for the real-time condition and 16 ± 2 epochs for the delayed condition were removed on the basis of artifacts (9% and 13% of the data for the real-time and delayed conditions, respectively). They were then exported into MATLAB (The MathWorks, Natick, MA) using the EEGLAB toolbox (Delorme and Makeig 2004) for further analyses.

Monopolar EEG recordings were transformed to current source density (CSD) profiles using the Laplacian transformation as implemented in the CSD toolbox (Kayser and Tenke 2006). The signal was interpolated with a spherical spline interpolation procedure to compute second-order derivatives in two dimensions of space (order of splines, 4; approximation parameter λ, 1.0e-5) (Perrin et al. 1989). Because the CSD transform represents “true” estimates of current activity at the scalp, the polarity of the signal is unambiguous by itself and is therefore more informative than the polarity of the surface potential (Kayser and Tenke 2006). CSD data are also much less affected by far-field generators than monopolar recordings (Manahilov et al. 1992), thus enhancing the spatial and temporal resolution of the recordings (Burle et al. 2015; Vidal et al. 2015).

After the various kinematic and EEG artifact rejections, the mean number of trials analyzed per participant was 381 ± 15 for the real-time condition (21% of data rejected in total for this condition) and 90 ± 2 for the delayed condition (25% of data rejected in total for this condition).

Visual ERPs.

In light of previous work that has identified the N1 component as being modulated by anticipation of visual action effects (Gentsch and Schütz-Bosbach 2011; Roussel et al. 2014), the main hypothesis focused on this component. Two regions of interest (left parietal, right parietal) were created by pooling the electrodes overlaying the left and right posterior parietal regions (P1, P3, PO3 and P2, P4, PO4, respectively; see Homan et al. 1987 for the cerebral topography of the electrodes). This was based on a number of previous studies that have characterized the N1 component at or near these electrode sites (Buzzell et al. 2013; Di Russo et al. 2002; Foxe and Simpson 2002; Manning et al. 1988). For completeness, the early P1 component (recorded at O1 and O2, which were pooled; Clark and Hillyard 1996; Johannes et al. 1995) and the P2 component (recorded at left parietal and right parietal regions; Di Russo et al. 2012) were also analyzed. As a marker of novelty detection, we also measured the P300 component at parietal electrode Pz (Correa et al. 2006; Miniussi et al. 1999; Patel and Azzam 2005).

The timing of the P1 component was taken as the peak positive deflection of the visual ERP within a temporal window between 90 and 130 ms (Clark and Hillyard 1996; Di Russo et al. 2002; Johannes et al. 1995). The amplitude of the P1 component was defined as the difference between the P1 peak and the baseline, defined as the average ERP signal between 0 and 50 ms. The timing of the N1 component was taken as the peak negative deflection of the visual ERP within a temporal window between 125 and 200 ms. The amplitude of the N1 component was calculated as the difference between the N1 peak and the baseline (i.e., 0 to 50 ms; Buzzell et al. 2013; Correa et al. 2006; Di Russo et al. 2002). The timing of the P2 component was taken as the peak positive deflection of the visual ERP within a temporal window between 170 and 320 ms (Zalar et al. 2015). The amplitude of the P2 component was defined as the difference between the P2 peak and the N1 peak. The timing of the P300 component was defined as the peak positive deflection of the visual ERP within a temporal window between 250 and 450 ms. Its amplitude was measured relative to baseline (i.e., 0 to 50 ms; Correa et al. 2006; Miniussi et al. 1999; Patel and Azzam 2005).

Statistical Analysis

For kinematic data, paired-samples t-tests were used to compare RT, MT, radial error, and X and Y accuracy between the real-time and delayed conditions. Paired-samples t-tests were also used to compare the time necessary to reach the 5-cm radius and movement velocity at that point, which corresponds to where the cursor was presented. Alpha threshold was fixed at 0.05.

For EEG data, repeated-measures ANOVAs with the factors hemisphere (left parietal, right parietal) and condition (real time, delayed) were used to compare the timing and amplitude of the N1 and P2 components. t-tests were used to compare the timing and amplitude of the P1 and P300 components across the two conditions. Alpha threshold was fixed at 0.05.

RESULTS

Kinematic Results

There was no difference in RT across conditions (real time: 302 ± 13 ms; delayed: 297 ± 12 ms; means ± SE; t = 1.53, P = 0.15). In contrast, there was a significant difference in MT (Fig. 2A), which was significantly longer in the delayed condition (852 ± 55 ms) than in the real-time condition (814 ± 56 ms; t = 5.82, P < 0.01).

Fig. 2.

Behavioral results. A: average movement time in each condition. B: average radial error at movement endpoint in each condition. C: average endpoint position relative to the target position in each condition. Error bars denote SE. *P < 0.01.

Concerning radial error at movement endpoint (Fig. 2B), the t-test revealed a significant difference across conditions (t = 3.59, P < 0.01). Radial error was significantly larger in the delayed condition (0.62 ± 0.08 cm) compared with the real-time condition (0.26 ± 0.05 cm). More specifically, movement amplitude tended to be larger in the delayed condition, as indexed by a significant difference in Y accuracy (real time: −0.07 ± 0.06 cm; delayed: 0.55 ± 0.09 cm; t = 6.88, P < 0.01). There was no difference in X accuracy across conditions (real time: 0.03 ± 0.06 cm; delayed: 0.14 ± 0.08 cm; t = 1.33, P = 0.21). Hence, participants overshot the target in the delayed condition compared with the real-time condition (Fig. 2C). Postexperiment debriefing with participants revealed that none of them consciously perceived the delay in the feedback or the slight accuracy differences across conditions.

The time to reach the 5-cm radius was nearly identical across conditions (real time: 257 ± 15 ms; delayed 255 ± 16 ms), although it reached statistically significant levels (t = 2.49, P = 0.03). Also, movement velocity at the 5-cm crossing did not differ significantly across conditions (real time: 46.08 ± 3.29 cm/s; delayed: 46.08 ± 3.08 cm/s; t = 0.01, P = 0.99). Furthermore, there was no significant difference in the X direction at the 5-cm crossing across conditions (real time: −0.40 ± 0.02 cm; delayed: −0.39 ± 0.03 cm; t = 0.75, P = 0.47). Together, these data suggest that the retinotopic features of the visual cursor being presented were very similar across conditions.

EEG Results

The P1 component was observed at occipital scalp sites with a mean peak timing of 103 ms. Its timing did not differ significantly across conditions (real time: 104 ± 4 ms; delayed: 101 ± 3 ms; t = 0.74, P = 0.47). Similarly, its amplitude did not differ significantly across conditions (real time: 4.30 ± 0.97 μV/m²; delayed: 4.20 ± 0.83 μV/m²; t = 0.34, P = 0.74).

The N1 component was observed at a mean peak timing of 163 ms. The scalp topography of the ERP at this moment (in the real-time condition) is presented in Fig. 3A. As can be seen, the negativity associated with visual cursor presentation is clearly visible at bilateral parietal scalp sites. Concerning the timing of the N1 component (Fig. 3B), the repeated-measures ANOVA revealed a small but significant main effect of condition [F_(1,13) = 10.70, P = 0.01] with the N1 peaking later in the delayed condition (165 ± 4 ms) compared with the real-time condition (161 ± 4 ms). There was neither a significant main effect of hemisphere [left: 161 ± 4 ms; right: 165 ± 4 ms; F_(1,13) = 2.48, P = 0.14] nor an interaction across the two factors [F_(1,13) = 0.92, P = 0.36]. As for the amplitude of the N1 component (Fig. 3C), the repeated-measures ANOVA revealed a significant main effect of condition [F_(1,13) = 9.23, P = 0.01] with a larger N1 amplitude in the delayed condition (21.22 ± 1.78 μV/m²) compared with the real-time condition (17.65 ± 1.37 μV/m²). There was neither a significant main effect of hemisphere [left: 20.30 ± 2.08 μV/m²; right: 18.57 ± 1.83 μV/m²; F_(1,13) = 0.45, P = 0.51] nor an interaction across the two factors [F_(1,13) = 0.21, P = 0.66]. It should be noted that although ERP amplitude tends to increase with the number of trials due to increased signal-to-noise ratio, a larger N1 response was observed in the delayed condition despite the fact that it contained four times fewer trials, suggesting a true change in the magnitude of the neuronal response.

Fig. 3.

EEG results. A: scalp topography of current source density (CSD) activity at the peak timing of the N1 component in the real-time condition. The negativity associated with visual cursor presentation is clearly visible at parietal scalp sites. White circles identify electrodes selected for further analyses (P1, P3, and PO3 for left parietal and P2, P4, and PO4 for right parietal). B: event-related potentials in left and right parietal regions across conditions, time-locked to the onset of the cursor (t = 0 ms). C: average amplitude of the N1 component in left and right parietal regions across conditions. In both hemispheres, the N1 amplitude was attenuated in the real-time condition compared with the delayed condition. Error bars denote SE. *P < 0.05.

The P2 component had a mean peak timing of 223 ms. Concerning the timing of the P2 component, the repeated-measures ANOVA revealed no main effect of condition [real time: 222 ± 4 ms; delayed: 224 ± 4 ms; F_(1,13) = 0.61, P = 0.45], no main effect of hemisphere [left: 225 ± 4 ms; right: 221 ± 4 ms; F_(1,13) = 0.41, P = 0.53], and no interaction [F_(1,13) = 0.62, P = 0.45]. Similarly, for the amplitude of the P2 component, the repeated-measures ANOVA showed no main effect of condition [real time: 25.56 ± 1.85 μV/m²; delayed: 25.65 ± 2.00 μV/m²; F_(1,13) = 0.03, P = 0.87], no main effect of hemisphere [left: 25.54 ± 2.10 μV/m²; right: 25.67 ± 3.06 μV/m²; F_(1,13) = 0.001, P = 0.97], and no interaction [F_(1,13) = 0.70, P = 0.42].

Finally, the P300 component was measured at parieto-occipital scalp sites with a mean peak timing of 365 ms. Its timing did not differ significantly across conditions (real time: 362 ± 16 ms; delayed: 368 ± 13 ms; t = 0.54, P = 0.60). Similarly, its amplitude did not differ significantly across conditions [real time: 10.95 ± 2.17 μV/m²; delayed: 11.55 ± 1.87 μV/m²; t = 0.37, P = 0.71].

DISCUSSION

In this study we investigated whether the cortical processing of visual reafferent signals from the moving limb is modulated during voluntary movements. A reaching task was used in which the timing of hand visual feedback was manipulated while its retinotopic location was kept very similar across conditions. This made it possible to directly compare the amplitude of visual ERPs to accurately predicted (i.e., real time) or mispredicted (i.e., delayed) visual feedback. Results revealed that the N1 component was significantly attenuated when visual feedback was provided in real time compared with when it was delayed. These data suggest that akin to other sensory modalities, the visual consequences of an action are attenuated when they are accurately predicted, likely by a forward model.

The attenuation was restricted to the N1 component, which was most prominent at posterior parietal electrodes bilaterally. This location is consistent with evidence from Buzzell et al. (2013), who reported a similar N1 component at parietal scalp sites in response to motion stimuli. The timing of the present N1 component (peaking at 163 ms) is also in line with findings from Di Russo et al. (2005), who recorded an N1 component at parieto-occipital electrodes at 164 ms for visual stimuli provided in the lower visual field. Overall, given the topography and the timing, the present results support prior work pointing to the extrastriate visual areas and posterior parietal cortex (PPC) as neural generators of the N1 component (Clark and Hillyard 1996; Di Russo et al. 2005; Gomez Gonzalez et al. 1994).

It is well documented that the amplitude of the N1 component is modulated by visuospatial attention and by the retinotopic location of visual stimuli (Buzzell et al. 2013; Eimer and Schröger 1998; Gomez Gonzalez et al. 1994; Hillyard and Anllo-Vento 1998; Mangun 1995). In the present study participants maintained gaze on the target throughout the entire experiment, and the task was specifically designed so that the retinotopic features of the cursor would be comparable across conditions. Indeed, the location where the cursor was presented (crossing of a 5-cm radius) as well as the duration for which it was presented (150 ms) were identical across conditions. The kinematics of the movements were also tightly controlled, as confirmed by analyses revealing that the X position and the velocity of the hand at the 5-cm crossing did not differ significantly across conditions. For these reasons it is unlikely that the difference in N1 amplitude was related to differences in visuospatial attention or in the retinotopic location of the visual stimulus. On a related note, differences in temporal attention could have accounted for the present results. However, temporal attention is thought to increase the gain of visual ERPs (Correa et al. 2006; Miniussi et al. 1999); hence, had it influenced the present results, the amplitude of the N1 component would have been greater for the more probable real-time condition (80% of trials), which is contrary to what was observed.

The temporal delay also created a spatial mismatch between the visual and proprioceptive estimates of the hand, which could have influenced the amplitude of the N1 component given the implication of the PPC in multisensory integration. However, populations of multimodal cells in frontal and parietal cortices have been shown to respond more potently when vision and proprioception of the limb are congruent (Graziano 1999; Graziano et al. 2000; Simon-Dack et al. 2009). In support, Simon-Dack et al. (2009) observed that the amplitude of the visual N1 component generated by laser dots was enhanced when the laser appeared directly on the hand compared with when it appeared away from it. These findings are contrary to the present pattern of results, suggesting that the N1 modulation is unlikely to be due to a visuo-proprioceptive mismatch in the delayed condition. Similarly, the temporal delay in cursor presentation had the inevitable consequence that the visual feedback was provided at different phases of the ongoing movement. Hence, the visual ERP superimposed on slightly different underlying EEG activity in the two conditions. Nevertheless, the distribution and timing of visual ERPs were very similar across conditions. Furthermore, the fact that the modulation was present for the N1 component but not for components occurring earlier (P1) or later (P2, P300) argues for the specificity of the observed effect rather than to a nonspecific effect associated with the different phases of the movement.

In sum, the present findings point to a true change in the magnitude of the response within neural assemblies that give rise to the N1 component when the timing of visual feedback is manipulated. A possible substrate for the present attenuation is the PPC, given its involvement in the integration of the predicted and actual sensory consequences of the movement for the real time estimate of limb state (Desmurget et al. 1999; Mulliken et al. 2008). The PPC receives strong projections from the cerebellum (Clower et al. 2001; Prevosto et al. 2009), which has been suggested to be responsible for the reduced cortical response to self-generated reafference (Blakemore et al. 1998b). Indeed, Blakemore et al. (1999b) showed that for self-generated tactile stimuli, activity in the cerebellum significantly predicted the attenuation observed at the cortical level. One mechanism that could account for the present results is that visual predictions increased the baseline firing activity of spatially tuned neuronal ensembles of the PPC in anticipation of visual feedback (Waszak et al. 2012). This is supported by the fact that the BOLD response in the PPC is tuned to the upcoming movement direction in gaze-centered coordinates during reach planning (Bernier and Grafton 2010; Fabbri et al. 2010; Gallivan and Culham 2015), suggesting retinotopically organized preparatory activity. In the presence of such heightened baseline activity, and given that ERPs are corrected with respect to baseline (i.e., 0–50 ms in this experiment), such preparatory activity would give rise to lower ERP responses to correctly predicted sensory input (Waszak et al. 2012), as was observed in the present study. In sum, the present findings add to the current literature by showing that visual predictions act not only in a spatially constrained manner but also in a time-constrained manner, with delays as short as 150 ms significantly influencing the amplitude of the cortical response.

The present data are also consistent with recent theory suggesting that the cortical response to sensory input is proportional to the amount of information that is unaccounted for by a prediction (i.e., prediction error; Friston and Kiebel 2009). Indeed, Hsu et al. (2015) reported that mispredicted auditory feedback is associated with an increased amplitude of the auditory N1 component compared with correctly predicted feedback. In this context, an equally valid interpretation of the present results is that the amplitude of the N1 component was more ample in the delayed condition as a result of the error between the predicted and actual visual reafferent signals. In support of this, early PET work has reported increased activity in the PPC during reaching movements with shifted visual feedback (Clower et al. 1996). Similarly, Ogawa et al. (2007) showed that activity in the PPC is positively correlated with motor error during tracing with delayed visual feedback, possibly reflecting its role in evaluating visuomotor prediction error. Finally, when exposed to a constant delay in the relationship between an action and its visual effect (i.e., a flash), the amplitude of early visual ERPs gradually attenuates as a function of adaptation (Stekelenburg et al. 2011), which also supports the notion that the amplitude of the cortical response indexes prediction error.

Correct anticipation of sensory consequences has been shown to influence perceptual discrimination in somatosensory (Blakemore et al. 1999b), auditory (Sato 2008), and visual (Cardoso-Leite et al. 2010) modalities. Recent work directly linked the reduced ERP response for congruent action effects with reduced contrast discrimination (Roussel et al. 2014), leading to the suggestion that the neurophysiological markers of sensory attenuation reflect reduced perceptual processing of voluntary action effects. Although the present study was not designed to test whether the change in ERP amplitude was related to a change in perception, it appears that the sensitivity of the ERP to the temporal delay was greater than the perceptual threshold, because none of the participants consciously perceived the delay in the feedback. This is further supported by the lack of a difference in the P300 component, which has been suggested to index novelty detection (Correa et al. 2006; Miniussi et al. 1999; Patel and Azzam 2005). Still, the behavioral data revealed that participants did utilize visual feedback for limb state estimation in the delayed condition, because MT and movement amplitude were significantly larger in this condition compared with the real-time condition. In this light, given that the PPC performs Bayesian integration of multimodal sensory information, further work should assess how the relative reliance on hand visual feedback for the online control of the movement is influenced by temporal delays.

GRANTS

This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada and the Fonds de Recherche du Québec - Santé.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M.B. performed experiments; M.B., F.T., and P.-M.B. analyzed data; M.B. and P.-M.B. interpreted results of experiments; M.B. prepared figures; M.B. and P.-M.B. edited and revised manuscript; M.B., K.W., and P.-M.B. approved final version of manuscript; P.-M.B. conception and design of research; P.-M.B. drafted manuscript.