Limited Contribution of Primary Motor Cortex in Eye-Hand Coordination: A TMS Study

Data acquisition.

The experimental setup, adapted from Landelle et al. (2016), is illustrated in Figure 1. Subjects were comfortably seated in a dark room facing a screen positioned on the frontal plane 57 cm away from the subject's eye. To minimize measurement errors, subjects' head movements were restrained by a chin rest and a padded forehead rest so that the eyes in primary position were directed toward the center of the screen. A mask was positioned under the participants' chin to block vision of their hands. In some of the experimental conditions (see below), participants were required to grasp with the right hand a force sensor (ELPM-T1M-25N, Entran) between the index finger and the thumb. The right forearm was resting over the subject's laps. The output of force sensor (i.e., GF) was sent to a multichannel signal conditioner (MSC12, Entran), then recorded at 1000 Hz with a resolution of 0.02N. Electromyographic activity was recorded from the first dorsal interosseous (FDI) at a 1000 Hz sampling frequency (Delsys Bagnoli). Electrodes were positioned over the belly tendon montage, and a Dermatrode self-adhering electrode was positioned on right olecranon elbow for ground.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

A, Schematic drawing of the experimental setup. All parts of this figure were drawn by the authors. B, Photography relative to the EXT-TACTILE condition in which passive grip force modulation was generated via the experimenter (who is holding the participant's hand).

The target was a red disk laser beam (0.5° in diameter) projected on the screen. The laser was moved by an optical scanner (GSI M2 series) servo-controlled by a PC. The delay in the servo-command was <2 ms. The optical scanner motion was restricted to one dimension so that the target moved only along the horizontal axis. Eye movements were recorded using an infrared video-based eye tracker (Eyelink Desktop-mounted system; SR Research). Horizontal and vertical positions of the right eye were recorded at a sampling rate of 1000 Hz. Before each block of trials, we calibrated the output from the eye tracker by recording the raw eye positions as subjects fixated a grid composed of 9 known locations. The mean values during 1000 ms fixation intervals at each location were then used for converting off line raw eye tracker values to horizontal and vertical eye position in degrees of visual angles.

TMS hot spot search.

A 70 mm figure-of-eight double coil connected to a Magstim Bistim2 magnetic stimulator (Magstim) was positioned tangentially to the scalp, oriented perpendicular to the central sulcus and 45° angle to the interhemispheric fissure (Brasil-Neto et al., 1992; Mills et al., 1992). With the coil handle pointing laterally and caudally, a posterior-anterior brain current through the precentral gyrus was induced (Wassermann et al., 2008). While the participant held a constant GF of 3N, the site at which the largest GF pulse elicited was determined. When this site was found we determined the active motor threshold (AMT) as the lowest intensity that could elicit noticeable changes in GF in 5 of 10 trials while maintaining a GF of 3N. The mean group AMT was 40.8 ± 7.3% of maximum stimulator output (range: 33%–56%). Subsequently, the intensity of TMS was raised so as to induce reliable 1N GF pulses (while maintaining 3N). This procedure allowed normalizing the behavioral and visual effects of TMS across subjects. The corresponding mean group TMS intensity was 45.7 ± 7.8%, which corresponds to ∼110% of AMT. Before each TMS session, we ensured that the coil was still adequately positioned (i.e., giving rise to a 1N increment in GF).

Experimental design.

In all trials, subjects were instructed to track with their eyes a moving target on the screen as accurately as possible. Depending on the experimental condition, the motion of the target was driven either by the subjects themselves (SELF) or by an external agent (EXTERNAL). We propose to present first the SELF protocol in which subjects drove the target motion with the right hand through GF modulations. The mapping was chosen such that, when GF was 3N, the target was located at the center of the screen. When GF increased to 5N, the target moved 15° to the right (+), conversely when GF decreased to 1N the target moved 15° to the left (−). Overall, the gain between GF and target position was 1N for 7.5°. Subjects were encouraged to use the whole extent of the screen (±15°) while making sure that the target did not fall outside of the screen boundaries. Subjects were asked to perform random oscillatory target movements (for a similar procedure, see Steinbach and Held, 1968; Angel and Garland, 1972; Landelle et al., 2016). The underlying motivation was to make target motion as unpredictable as possible when subsequently played back in the EXTERNAL conditions. To facilitate the production of random movements, a template was given during demonstration trials. During the experimental trials, we ensured that GF modulation led to a mean absolute target velocity close to 30°/s: mean target velocity was computed online, and the experimenters provided some verbal feedback to the subject, such as “please move faster” or “please slow down,” when necessary. This procedure ensured minimal changes in mean target velocity across subjects (SD = 0.95°/s), experimental conditions (SD = 1.16°/s), and trials (SD = 0.42°/s).

A total of four conditions with a self-moved target were tested. These were the following. In the first one called SELF, the participants simply moved the target randomly with the hand and tracked it with the eyes. In the SELF-TMS condition, the task was similar but occasionally (4 times per trial) TMS pulses were triggered. Despite fast target jumps inherent to the effect of TMS on GF, subjects were encouraged to keep on tracking the target. The third condition (SELF-MIMIC) was similar to SELF, except that we occasionally (4 times per trial) induced some transient target jumps that mimicked those observed during SELF-TMS. These visual perturbations were accompanied by a TMS click from a coil positioned 10 cm above the scalp (SHAM TMS). This mimicking was possible thanks to a polynomial function (fifth order) derived from pilot studies (N = 5) that fitted TMS induced GF pulses between 30 and 230 ms after TMS. The underlying motivation for mimicking the effect of TMS was to investigate the effect of the visual perturbation and TMS click in the absence of TMS. As exposed in Figure 2A, this procedure was rather successful in that mean target trajectory in SELF-TMS and SELF-MIMIC were very similar. However, visual perturbations in SELF-MIMIC were less variable because we used the same polynomial function for all the participants. In the last condition (SELF-TMS-MASK), the procedure was similar to SELF-TMS, except that we manipulated visual feedback so as to mask the target jumps induced by TMS. To achieve this goal, we developed an algorithm to extrapolate the “intended” target trajectory for 200 ms based on its position and velocity just before the behavioral effect of TMS. Practically within the next 200 ms, target velocity was decreased by 1/250 of its initial value every millisecond. As shown in Figure 2B, this procedure was rather effective in that the resulting target trajectory was smooth (no more target jumps) and was very similar to the one observed during SELF for comparable time periods. This condition was used to assess the effect of TMS independently of its visual consequences.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Mean group target trajectories under the SELF conditions during rightward target motion. A, Comparison between target trajectories during SELF-TMS and SELF-MIMIC. Shaded area around each trajectory represents the SE across subjects. These two trajectories largely overlap, thereby demonstrating that the visual perturbation (target jump) induced by TMS was successfully imitated during our SELF-MIMIC condition. B, Same as A but for SELF and SELF-TMS-MASK. Our extrapolation procedure was rather effective in suppressing the target jump normally associated with TMS.

Participants also had to perform trials under which target motion was externally driven (EXTERNAL). A total of four EXTERNAL conditions were tested. In the first one (EXT), target trajectories collected during SELF were played back in the same order. Each participant was presented their own trials (Angel and Garland, 1972; Landelle et al., 2016). This procedure allowed within-subject comparisons. During this condition, the right hand of the subject was relaxed while resting on the subject's lap. In the second condition (EXT-TMS), the same procedure was used, except that occasionally (4 times per trial) TMS was applied. In the third one (EXT-TMS-GRIP), the procedure was similar to EXT-TMS, but the hand was no longer relaxed and the subject had to generate a constant GF of 3N. The underlying motivation was to assess whether activity in M1, independently of eye-hand coordination, could influence the effect of TMS on eye motion. In both TMS conditions, the order of trial playback was identical as in EXT. Finally, in a fourth condition (EXT-TACT), the experimenter moved the target by placing his hand over the participant's relaxed hand, which was holding the force sensor (Fig. 1B). The rationale was to assess the possible contribution of tactile information under SELF eye-tracking performance. This procedure did not allow us to present the same target trajectories as in the other EXTERNAL conditions, but we ensured comparable mean target velocities.

Overall, we explored a total of 8 experimental conditions. Each participant performed 1 block of 10 trials (30 s each) in each of these experimental conditions. The order of the blocks was randomized, except the SELF block, which had to be completed first before participants could perform the EXT, EXT-TMS, and EXT-TMS-GRIP blocks.

After the completion of these 8 blocks, half of the subjects completed an extra block of 10 trials. During this last block, the self-moved target was unexpectedly blanked 4 times per trial. In one-half of these blanks, TMS was triggered slightly after target extinction (SELF-TMS-BLANK). In the other half, blanks were not accompanied by TMS (SELF-BLANK). We reasoned that the removal of visual feedback would likely increase the saliency of hand efferent signals, and thus possibly augment the perturbing effect of TMS.

Timeline of TMS and target occlusion.

The initiation of TMS was fully determined by the ongoing kinematics of the target. We intended to trigger TMS around the center of oscillation (Gagnon et al., 2006; Nuding et al., 2009), namely, when absolute target velocity was high (>40°/s) and absolute acceleration was low (<2°/s²). Within each trial, TMS was applied four times: twice during a rightward target motion and twice during a leftward target motion, with an order that was randomized across trials. We also imposed that TMS could not be triggered during the first 4 s, and that consecutive TMS pulses were separated by at least 4.5 s. During the SELF conditions, this experimental design implied that TMS could be triggered either during an increasing or a decreasing contraction of the fingers. In agreement with earlier observations (Cros et al., 2007; Gruet et al., 2013), we found that MEPs and GF pulses were substantially greater during an increasing contraction (MEP = 3.78 ± 1.82 mV; GF pulses = 1.68 ± 0.49N) than a decreasing one (MEP = 2.05 ± 1.53 mV; GF pulses = 0.73 ± 0.13N) (for more details, see Data analysis). The rationale to trigger TMS during increasing and decreasing contractions was to investigate whether the effect of TMS on eye motion would scale with MEP magnitude. However, as will be exposed later, TMS had no short-latency effect on SP during both increasing and decreasing contractions; we thus have decided to pool these two types of contraction in Results. Because a similar conclusion was obtained for TMS during rightward and leftward motion of the target during EXTERNAL trials, those observations were also pooled.

Regarding transient target occlusions during SELF-BLANK, the duration of occlusion (400 ms) was chosen to be in the range of values used by others (Mehta and Schaal, 2002; Bennett and Barnes, 2003; Orban de Xivry et al., 2008; Landelle et al., 2016). The following constraints were used to trigger target occlusions. First, occlusions could only be initiated during rightward movement (increasing contraction), namely, when target velocity was close to 0 (<0.1°/s), and target acceleration was high (>300°/s²). Second, we imposed that target occlusions could not be introduced during the first 4 s, and that they had to be separated by at least 4.5 s. Third, four occlusions were triggered per trial. In two of them, TMS was triggered 150 ms after the initiation of the blank. This procedure allowed to trigger TMS around the center of oscillation (as in SELF-TMS). Before this block, participants were informed about the possible occurrence of TMS and target occlusions, but they were instructed to keep their eyes on the target as if it was still present on the screen.

Data analysis.

Because the stimuli were moving exclusively along the horizontal meridian, we focused our analyses on the horizontal component of eye movements. We then performed a sequence of analyses to separate periods of SP, saccades and blinks from the raw eye position signals. The identification of the blinks was performed based on the pupil diameter (that was also recorded). This procedure led to the removal of <1% of eye recordings. Eye position time series were then low-pass filtered with a Butterworth (fourth order) using a cutoff frequency of 25 Hz. The resultant eye position signals were differentiated to obtain the velocity traces. The eye velocity signals were low-pass filtered with a cutoff frequency of 25 Hz to remove the noise from the numerical differentiation. The resultant eye velocity signals were then differentiated to provide the acceleration traces that we also low-pass filtered at 25 Hz to remove the noise. A dedicated MATLAB script was used to identify the beginning and end of each saccade. This identification was based on the acceleration and deceleration peaks of the eye (>1500°/s²). Further visual inspection allowed to identify smaller saccades (<1°) that could not be identified automatically by our program. Based on these computations, epochs of pursuit and of saccades were extracted.

To assess the participants' ability to predict the dynamics of the target or hand-target during visually guided tracking, we computed several dependent variables. First, we computed the mean absolute position error (PE) by averaging the absolute difference in position between the target and the eye over the whole trial. Second, we computed the mean absolute velocity error (VE; i.e., the average absolute difference between the eye and target velocity). Although PE was evaluated over the whole trial (i.e., including both periods of saccades and SP), VE was computed only during SP periods. Third, to evaluate the temporal relationship between the eye and target movements, we computed the lag between eye and target velocity signals using a cross-correlation (a positive lag corresponding to the eye lagging behind the target). This lag computation was based on the eye signal excluding saccades. Finally, we computed the SP gain by averaging the ratio between instantaneous eye and target velocities during phases of SP (to avoid numerical instabilities, only situations where absolute target velocity was >10°/s were considered). For all these analyses, the first second of each trial was discarded.

Regarding the effect of TMS, we tried first to characterize its effect on hand action. The latency and magnitude of changes in GF and MEP in first dorsal interosseous were computed. The latency of MEP was calculated with the following procedure. Mean and SD of rectified EMG was determined in the 100 ms window preceding TMS. Then in the subsequent data points, we identified the moment at which EMG went above a threshold value that was set to mean EMG + 2 SD (Danion et al., 2003; Volz et al., 2015). A similar procedure was adopted to determine the latency of GF pulses induced by TMS. To assess the magnitude of TMS-induced changes, we measured the peak-to-peak amplitude of MEP and GF pulses within the next 100 ms that followed the individual latencies.

Concerning the effect of TMS on SP, another set of analyses was conducted to investigate possible alteration in SP velocity after TMS (up to 150 ms). To achieve this goal, we used a linear interpolation routine to bridge the gaps produced by removal of saccades from the eye velocity trajectory (Bennett and Barnes, 2003). Possible alterations in SP velocity were investigated both at the level of its mean value and its fluctuation across trials. To address this issue, for each subject, and in each condition, we computed the mean and standard error (SE) of SP velocity across trials for each of the 150 time points following TMS. As TMS was triggered based on target velocity, we did not explicitly control SP velocity at that moment, and observed some disparities across experimental conditions with group mean velocities ranging from 13.3 to 18.6°/s. We also noticed that, in some cases, eye velocity could be really low (<5°/s), plus blinks could also occur in the vicinity of TMS. Because the effect of TMS has been shown to depend substantially on SP velocity (Gagnon et al., 2006; Nuding et al., 2009), we felt it was critical to exclude the earlier cases and equalize SP velocity at the moment of TMS. This was achieved by selecting observations in which velocity was closest from 20°/s, an eye velocity previously studied and reported sensitive to TMS over frontal eye field (Gagnon et al., 2006). Overall, for each condition and each participant, we kept 20 observations (10 rightward + 10 leftward) out of the 40 available.

Statistical analysis.

We decided not to use a statistical design that included all the experimental conditions. Instead, our statistical analyses were designed to address some specific issues. The first one was to confirm that eye tracking was more accurate when subjects moved the target themselves compared with when it was moved by an external agent. To address this point, we focused on the comparison between eye tracking performance in SELF, EXT, and EXT-TACT. Specifically, we compared PE, VE, eye-target lag, and SP gain over whole trials. A second issue was to assess the possible effect of TMS in the absence of eye-hand coordination (externally moved target). To achieve this goal, we focused on SP eye movements in EXT, EXT-TMS, and EXT-TMS-GRIP. Specifically, we looked for possible differences in terms of PE and VE over the 200 ms that followed TMS. To investigate possibly more subtle effects of TMS, we also inspected SP velocity profiles associated with each of the three external conditions, along with statistical analyses of mean SP velocity within the 25–125 ms epoch following TMS (Gagnon et al., 2006). Data alignment was based on target trajectory, which was similar in all these three conditions. Subsequently, we assessed the possible effect of TMS during eye-hand coordination (self-moved target). This time, we focused on the comparison between SELF, SELF-TMS, SELF-MIMIC, and SELF-TMS-MASK. To separate the possible contribution of TMS and the occurrence of a target jump, two-way repeated-measures ANOVA was performed with TMS (with/without) and T-JUMP (with/without) as within-subject factors. Again, we searched for possible differences in terms of PE, VE, and SP velocity immediately after TMS. Segments of eye motion in SELF in which target trajectory fulfilled our criteria for the initiation of TMS were used for baseline performance. Finally, we assessed the possible effect of TMS during eye-hand coordination in the absence of visual information by using data collected in SELF-BLANK and SELF-TMS-BLANK. Here, the comparison was based on velocity profiles obtained in the presence/absence of TMS during target occlusion, along with comparable data from the SELF condition.

For all these issues, we used repeated-measure ANOVA to assess the effect of experimental conditions. Tukey corrections were used for post hoc t tests to correct for multiple comparisons. A conventional 0.05 significance threshold was used for all analyses.