A total of 15 right-handed healthy subjects participated in the present study (n = 15; 10 women; 22.4 ± 1.63 years old). Handedness was determined via a shortened version of the Edinburgh Handedness inventory Oldfield (1971) and all subjects filled out a TMS safety questionnaire. None of the participants suffered from any neurological disorder or had a history of psychiatric illness, drug or alcohol abuse; neither was anybody undergoing a drug treatment that could influence their performance or their neural activity. All subjects were financially compensated for their participation and provided written informed consent. The protocol was approved by the Ethics Committee of the Université Catholique de Louvain.

Participants sat in front of a 21-inch monitor screen positioned about 60 cm in front of them with their arms semi-flexed and both hands resting palm-down on a response device developed in our laboratory (Quoilin et al., 2016). They performed an instructed-delay choice reaction time (RT) task, which required them to choose between abduction movements of the left or right index finger. The task was implemented with Matlab 7.5 (the Mathworks, Natick, Massachusetts, USAS) using the Psychophysics Toolbox extensions (Brainard, 1997; Pelli, 1997). The refresh rate of the monitor was set at 100 Hz.

The task consisted in a virtual “Rolling Ball” game previously used in another study (Quoilin et al., 2016; Figure 1A). In this game, participants were informed that the position of a preparatory cue (i.e., a ball separated from a goal by a gap) indicated the movement side for the forthcoming response: if the ball was on the left side of the screen, subjects had to prepare a left index finger response (to get ready to “shoot the ball into the goal”) and if the ball was on the right side, subjects had to prepare a right index finger response. Subjects were explicitly told to withhold their response until the onset of an imperative signal (i.e., a bridge). The latter appeared 1,000–1,200 ms after the ball and remained on the screen until a finger movement was detected or for a maximum duration of 500 ms. When the bridge was on the screen, subjects had to respond as fast as possible to allow the ball to roll on it and to quickly reach the goal. Subjects knew that they would get a score after each trial reflecting how fast and accurate they had been on the previous trial. Note that in each block, some catch trials (trials in which the bridge did not appear; 5% of all trials) were included. Subjects were required not to respond on these trials and were penalized if they did so. Hence, they had to avoid initiating their response prematurely, before the bridge onset. Trials were separated by the presentation of a blank screen lasting for a duration that varied between 2,050 and 2,300 ms (Figure 1B).

The home-made response device (Figure 1C) was composed of two pairs of metal edges fixed on a wooden platform (one for each hand) and each trial of the Rolling Ball game required participants to move one index finger from the outer to the inner metal edge (i.e., abduction of the index finger). The contact between the finger and the metal parts of the device was continuously monitored using a Makey Makey printed circuit board with an ATMega32u4 microcontroller running the Arduino Leonardo firmware, based on the principle of high resistance switching between two electrical contacts. This device provided us with a very precise measure of the RTs (precision = 1 ms) and allowed us to control for any anticipated movement. That is, the device permanently checked the initial position of each index finger (which had to be in contact with the outer metal edge) and any contact release before the onset of the imperative signal led to the cancellation of the trial and to a penalty.

Subjects received a feedback of their performance at the end of each trial. On correct trials, the feedback score (displayed in green) was inversely proportional to the reaction time (RT): the faster the subjects, the higher the score. The RT was defined as the time interval between the onset of the bridge and the time when the index finger left the outer metal edge of the response device. The score was determined based on the following equation, with ∝ = 0, 8 median RT measured at the end of the training session just before the main experiment:

Using this equation, scores on correct trials ranged from 1 to 100. Incorrect responses were penalized with negative scores displayed in red. They involved responses occurring too early, referred to as “anticipation errors” (penalized by 75 points), responses occurring too late, referred to as “time-out errors” (penalized by 50 points), responses provided with the incorrect hand (penalized by 20 points), referred to as “choice errors” and responses provided on catch trials (penalized by 12 points), referred as “catch errors.” Anticipation errors consisted in responses provided either before the bridge onset or after its onset but with a RT smaller than 100 ms. Time-out errors consisted in responses provided in more than 500 ms (after the bridge offset). Note that when subjects succeeded not to respond on a catch trial, they were rewarded by +12 points. The total score was always displayed at the end of each block.

TMS was delivered through one or two small figure-of-eight shaped coils (wing internal diameter 35 mm), each connected either to a Magstim 200² magnetic stimulator (Magstim, Whitland, Dyfed, UK) or a Magstim Bistim² magnetic stimulator. Both stimulators delivered monophasic pulses and their relationship to a specific hemisphere was counterbalanced between subjects. Each coil was placed tangentially over one primary motor cortex (M1) with the handle pointing backward and laterally at a 45° angle away from the midline, approximately perpendicular to the central sulcus (Figure 1D). Small coils were chosen because in most subjects, it is not possible to place two large coils over the two M1s at the same time. For each M1, the optimal scalp position to elicit a contralateral MEP in the first dorsal interosseous muscle (FDI) was identified and marked on a head cap placed on the subject's scalp to provide a reference mark throughout the experiment (Duque et al., 2014, 2016; Klein et al., 2014). Importantly, this was done by always checking for the fact that the two coils could be positioned simultaneously on the head without touching each other; to reduce electromagnetic interference it was sometimes necessary to adjust the orientation of the coils a little but these adaptations remained marginal and did not preclude us from obtaining the best MEP amplitudes.

The resting Motor Threshold (rMT) was determined at the hotspot for each M1 as the minimal TMS intensity required to evoke MEPs of about 50 μV peak-to-peak in the relaxed FDI muscle in at least 5 out of 10 consecutive trials. Across participants, the rMTs corresponded to 41.7 ± 5.05 and 40.8 ± 6.39% of the maximum stimulator output for the left and the right FDI, respectively. The intensity of TMS used throughout the experiment was always set at 115% of the individual rMT for each hemisphere.

The experiment started with two training blocks. The first one (20 trials) was conducted without TMS whereas the second one (40 trials) involved TMS, exactly as in the main experiment. Thereby, the subjects could first practice the task without being disturbed by the TMS pulse and then get used to the stimulations while performing the task in the second training block. The latter block also served to obtain the median RTs, used to individualize the scores on correct trials (see below). Then, during the main phase of the experiment, subjects performed 10 blocks of 40 trials (Figure 1E). Using these numbers, we obtained 20 MEPs in each condition.

The goal of the present experiment was to compare the amplitude of MEPs elicited during motor preparation using either single-coil or double-coil_{1 ms} TMS. In half of the trials, single-coil TMS was used, eliciting MEPs in a single hand (MEP_single), either in the left or the right FDI in a balanced proportion. In the other half, MEPs were elicited in both hands at once (MEP_double) using a double-coil_{1 ms} method where the two M1 are stimulated with a 1 ms inter-pulse interval (double-coil_{1 ms}; Grandjean et al., 2018). In all subjects, half of the double-coil_{1 ms} trials involved a pulse over left M1 first whereas the other half of the trials involved a pulse over the right M1 first. Therefore, for each hand, MEPs_double could either result from a first (MEP_double−1) or a second pulse (MEP_double−2). Importantly, the single- and double-coil_{1 ms} trials were always randomized within a block so that the subject could not anticipate the type of pulse (single or double) they would have next, an aspect that could bias MEPs, as suggested in a previous study (Wilhelm et al., 2016).

Single- and double-coil_{1 ms} TMS pulses were applied at three different timings during the Rolling Ball task (only one pulse per trial; Figure 1B). First, some TMS pulses occurred during the intertrial interval, at a random time falling 1,750–2,000 ms after the blank screen onset; these trials were used to compare MEP_single and MEP_double at baseline (rest) within the blocks (TMS_{baseline−in}, 20% of all trials). In the remaining trials, the TMS was delivered during the delay period either 900 ms (TMS_delay−900, 40% of all trials) or 950 ms (TMS_delay−950, 40% of all trials) after the occurrence of the preparatory cue. Based on previous studies (reviewed in Duque et al., 2017), we assumed that at these TMS_delay timings, inhibitory changes would be substantial whether MEPs are elicited in a selected condition (e.g., left MEPs elicited in a left hand trial) or a non-selected condition (e.g., left MEPs elicited in a right hand trial). Finally, we also recorded baseline MEPs outside the blocks (TMS_{baseline−out}), at six different times (before block 1 and after blocks 2, 4, 6, 8, and 10; 20 MEPs each). These MEPs provided us with a measure of CSE outside the context of the task, at complete rest. Moreover, the comparison of MEP_single and MEP_double at TMS_{baseline−out} allowed us to check whether we could replicate our previous observations (Grandjean et al., 2018).

EMG activity was recorded from surface electrodes (Neuroline, Medicotest, Oelstykke, Denmark) placed over the left and right FDI. MEPs recorded from these homonymous muscles offered a measure of CSE changes occurring in muscles that are involved in the task (whether selected or non-selected). Note that for all participants, stimulating the hotspot for the FDI also elicited reliable MEPs in the abductor digiti minimi (ADM), a pinkie abductor muscle which is irrelevant for the task. These MEPs were also considered in the present study. EMG data were collected for 1,000 ms on each trial, starting 300 ms before the TMS pulse. The EMG signals were amplified (x1,000), bandpass filtered online (10–500 Hz; NeuroloLog; Digitimer), and digitalized at 2,000 Hz for offline analysis.

Trials with background EMG activity (root mean square computed from −250 to −50 ms before the TMS pulse) exceeding 3 standard deviations (SD) around the mean were discarded for the following analyses. This was done to prevent contamination of the MEP measurements by significant fluctuations in background EMG (Duque et al., 2014, 2016; Klein et al., 2014). The remaining MEPs were classified according to the experimental condition within which they had been elicited. Trials in which subjects made an error were also removed from the data set; the task was easy so these trials remained rare and errors were not analyzed.

For each condition, we excluded trials with a peak-to peak MEP amplitude exceeding 3 SD around the mean. After screening the data for errors, background EMG activity and outliers, a total of 15.9 ± 2.7 trials per condition were left to evaluate CSE changes during action preparation. One subject had to be taken off the MEP analyses because we encountered a technical problem during the experiment (remaining n = 14 subjects).

Analyses were carried out with the RStudio software (version 1.0.153., RStudio, Inc., Boston, MA). The assumptions of normality and homogeneity of variance were tested before analyses. All data were systematically tested for the sphericity assumption using Maunchley's tests. The Greenhouse–Geisser correction was used for sphericity when necessary.

The RTs are shown on Figure 2 separately for the left and right hand trials. The ANOVA_RM revealed a significant influence of TMS_TIMING [F_(1,14) = 124.015 and p ≤ 0.001]: RTs were generally faster with TMS_delay (272.6 ± 36.4 ms) than with TMS_{baseline−in} (309.4 ± 38.8 ms), consistent with many previous reports showing that a TMS pulse applied close to the imperative signal can speed up the release of a motor response (Duque et al., 2012; Labruna et al., 2014; Greenhouse et al., 2015b). Furthermore, the MEP_CONDITION also influenced the RTs [F_{(2, 28)} = 6.007, p = 0.007]: Fisher LSD post-hoc tests revealed that RTs were significantly longer in the MEP_{single−Resp} condition than in the MEP_{single−NonResp} and MEP_double conditions (both p ≤ 0.004); the two latter were not different (p = 0.597). These results indicate that the RTs were slower in the presence of a single pulse eliciting a MEP in the responding hand compared to when the MEP was elicited in the non-responding hand or in both hands at once. Finally, the Mvt_SIDE × TMS_TIMING × MEP_CONDITION interaction was significant [F_{(2, 28)} = 5.125, p = 0.013]. As such, the slowing effect of MEP_{single−Resp} reported above was systematically observed with TMS_delay in both hands (all p ≤ 0.038). Yet, in trials with TMS_{baseline−in}, it was only present for right hand (both p ≤ 0.023) but not for left hand trials (both p ≥ 0.198).

The goal of the present study was to assess whether the MEPs acquired using double-coil_{1 ms} are equivalent to those obtained by means of a classical single-coil TMS method. To do so, we compared MEPs elicited by a first (MEP_double−1) or second (MEP_double−2) double-coil_{1 ms} TMS pulse to MEPs obtained using single-coil TMS (MEP_single). Both the amplitude and coefficient of variation (CV) of MEPs were considered. We compared these MEP variables in the context of a motor task requiring subjects to prepare and delay left or right index finger responses until the onset of an imperative signal. MEP_single are typically suppressed during the delay period (Bestmann and Duque, 2016; Duque et al., 2017). Here, we show that comparable inhibitory changes can be observed with MEP_double−1 and MEP_double−2. The MEPs exhibited comparable amplitudes and CVs, regardless of whether they had been elicited using a single- or double-coil_{1 ms} TMS approach.

The amplitude of MEPs was much smaller at TMS_delay compared to TMS_{baseline−in}, consistent with many previous reports (Duque and Ivry, 2009; Greenhouse et al., 2015b; Lebon et al., 2016; Quoilin et al., 2016; Wilhelm et al., 2016). This effect was observed regardless of whether the MEPs were recorded from a muscle that was selected or non-selected for the forthcoming response. This result may seem inconsistent with previous work (Greenhouse et al., 2015a,b; Klein et al., 2016). However, several recent studies have failed to observe a difference of inhibition between selected and non-selected conditions, suggesting that this effect is not consistent and does not systematically show up (Quoilin et al., 2016; Wilhelm et al., 2016, 2017). As suggested in Quoilin et al. (2016), it is likely to depend on the task details, including the use (or not) of response devices, the presence (or not) of catch trials, the time at which TMS is delivered and eventually, the presentation of a feedback (or not). Inhibition at TMS_delay was also observed for a muscle that was irrelevant in the task, corroborating the idea that withholding a prepared action is associated with widespread inhibitory influences suppressing CSE until the movement can be initiated (reviewed in Duque et al., 2017). The suppression of MEPs tended to be deeper in the left compared to the right hand, consistent with the view that inhibitory changes are often more pronounced on the non-dominant compared to the dominant side (Leocani et al., 2000; Duque et al., 2007; Quoilin et al., 2016; Wilhelm et al., 2016). Note that this tendency was not observed in a previous work (Klein et al., 2016). Yet, an important difference there is that Klein et al. (2016) registered MEP_left and MEP_right in separate blocks, reducing the signal to noise ratio when comparing these conditions. Furthermore, we found that MEPs were similarly decreased at TMS_delay−900 and TMS_delay−950, probably because preparatory inhibition had reached a plateau by the time TMS was applied, in accordance with recent observations (Lebon et al., 2016).

Most importantly, the strength of the inhibitory effect at TMS_delay was comparable across all MEP_TYPE. As such, bioequivalent analyses revealed that MEP_single, MEP_double−1, and MEP_double−2 displayed the exact same level of suppression during action preparation. This result may stand at odds with another study in which we observed differences between MEP_single and MEP_double at TMS_delay (Wilhelm et al., 2016). However, an important weakness in that work is that the single and double-coil_{1 ms} protocols were tested in separate blocks. Hence, the difference between MEP_single and MEP_double was likely due to the fact that subjects were more vigilant or alert when they expected two pulses to occur (increasing MEP amplitudes) compared to when only one pulse was anticipated (Labruna et al., 2011a; Klein et al., 2012, 2014). By intermingling all conditions within each block, the present study allowed to control for this bias: our data show that in its absence, all MEP_TYPE display a comparable degree of suppression during action preparation. Note however that, because MEPs are rather global readouts of CSE, these results do not allow to rule out completely the occurrence of some bilateral interactions following double-coil_{1 ms} TMS. Yet, even if present, these interactions do not alter MEP amplitudes in a systematic way and do not preclude from obtaining measures of preparatory inhibition that are comparable to those acquired with single-coil TMS.

In order to evaluate changes in the variability of CSE during action preparation, we measured the CV of MEPs elicited using single-coil or double-coil_{1 ms} TMS. Overall, we observed a decrease in the CV of MEPs at TMS_delay compared to TMS_{baseline−in}, even if this effect was not present in all conditions. Therefore, CSE tended to be less variable during action preparation compared to rest, consistent with a previous report (Klein-Flügge et al., 2013). Such a decrease in the variability of CSE during action preparation may reflect an optimization process of neuronal firing rates in the motor cortex (Churchland, 2006). Following this view, firing rates progressively become more consistent during action preparation, reaching a specific state to produce the desired movement (Rickert et al., 2009). Interestingly, this small decrease in the CV of MEPs at TMS_delay was not only observed for the FDI but also for the ADM. Hence, the variability of CSE decreased for both task-relevant and irrelevant muscles; the tuning of motor activity during action preparation may thus not be completely specific to the agonist effectors (Churchland et al., 2010; Klein-Flügge et al., 2013). Most importantly, changes in the CV from TMS_{baseline−in} to TMS_delay were equivalent for MEP_single, MEP_double−1 and MEP_double−2, suggesting that double-coil_{1 ms} TMS is as effective as single-coil TMS to capture changes in the variability of CSE during action preparation.

In the present study, we acquired two baseline measures of MEPs at rest. That is, MEPs were elicited during the intertrial interval (TMS_{baseline−in}) within the blocks, but also outside the blocks (TMS_{baseline−out}). At both timings, MEP amplitudes were generally comparable when elicited in the left or right hand, confirming that measures of CSE are highly comparable for both hemispheres at rest (Davidson and Tremblay, 2013). Yet, the CV of FDI MEPs was smaller in the right than in the left hand. Hence, neuronal firing rate may be steadier on the dominant side. Interestingly, MEP amplitudes were larger when acquired within the blocks compared to outside them and this effect was associated with a decrease in the CV of MEPs. Hence, CSE was larger and less variable when probed within the context of the motor task compared to when the subjects were at complete rest, outside the blocks. Such an effect on MEP amplitudes has been reported in a previous study comparing different baseline conditions (Labruna et al., 2011a). That is, Labruna et al. (2011a) showed that MEPs were larger when elicited in the context of a task requiring subjects to passively view hand or landscape pictures than when elicited outside the task, suggesting that the level of vigilance has a significant influence on CSE.

Importantly, our bioequivalence analyses revealed that MEP_single, MEP_double−1, and MEP_double−2 were comparable in all baseline conditions. The bioequivalence of MEPs at complete rest (TMS_{baseline−out}) had already been reported in a previous study (Grandjean et al., 2018). Here, we show that this equivalence persists when baseline MEPs are elicited in the context of a motor task (TMS_{baseline−in}).

First of all, RTs were generally faster with TMS_delay than with TMS_{baseline−in}, consistent with many previous reports showing that a TMS pulse applied close to the imperative signal can prime subjects to respond faster (Duque et al., 2012; Labruna et al., 2014; Greenhouse et al., 2015b; Quoilin et al., 2016) probably because the TMS sound triggers the release of the movement that is being prepared (Carlsen et al., 2007, 2011). Interestingly, we also found that RTs were longer in trials where a MEP_single occurred in the responding hand compared to when the MEP_single fell in the non-responding hand, or in both hands at once (MEP_double). This effect was present in all conditions at TMS_delay, indicating that the boosting effect of the TMS sound was slightly attenuated when the MEP fell specifically in the responding hand, compared to when it fell in the other hand or in both hands, consistent with other works (Duque et al., 2013; Wilhelm et al., 2016). Surprisingly this effect of the MEP condition was also observed with TMS_{baseline−in} in right hand trials but not left hand trials. This result was unexpected given that here, MEPs were elicited during the intertrial interval and should thus not affect behavior, an issue for future investigation.

The double-coil_{1 ms} protocol shows many advantages over the regular single-coil technique. First, the number of MEPs that can be collected in a given amount of time is doubled. This is a crucial aspect as it gives the opportunity to test more conditions within the same duration than could be done with a regular single-coil method. Second, CSE is probed bilaterally on the same trial meaning that both hands can be probed simultaneously. Hence, dominant and non-dominant hand MEPs are elicited in the exact same conditions during the task (Duque et al., 2013). This obviously increases the signal to noise ratio in a significant way. Third, the acquisition of MEPs in both hands allows researchers to make direct comparisons between bilateral MEPs on a single-trial basis and to develop new measures such as indexes reflecting the ratio between the CSE of the two hands. In fact, one may be interested in studying the impact of various task parameters (e.g., instruction, presence of reward, sensory evidence, level of urgency, effort required etc.) on the relationship between bilateral MEP amplitudes and CVs. Hence, the present technique opens new horizons in the study of how both hemispheres interact in various task settings (Verleger et al., 2009; Klein et al., 2016).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.