Sixteen healthy volunteers (11 female and 5 male) with a mean age of 24 years (range 20–30) participated in the experiment. All were right-handed (mean Oldfield [36
] handedness score of 93, range 60–100) and had normal or corrected-to-normal visual acuity. Subjects were screened for any history of neurologic illness or neurosurgery and for any metal or electronic implants. All subjects gave written informed consent prior to the experiment. The experimental procedures were in accordance with the declaration of Helsinki and approved by the local ethics committee of the Radboud University Nijmegen.
Procedure and task
Subjects were seated in a comfortable chair, in front of a 15 inch computer screen (distance ~75 cm). The subject's right forearm, wrist, and fingers 2–5 were immobilised in a tight U-shaped cast with the elbow flexed 90 degrees and the forearm semi-pronated. The thumb was left entirely free to move (see Figure ). In this setup, the axes of thumb abduction/adduction and of thumb flexion/extension were close to the horizontal and vertical space axes, respectively.
Figure 8 Experimental setup. The right arm was fixed with Velcro straps. Thumb movements were measured by two miniature uni-axial accelerometers that were mounted on the proximal phalanx of the thumb, in orthogonal planes. Electromyographic activity from the thumb (more ...)
The experimental tasks required subjects to move the thumb of their right hand in response to a series of visual signals, as quickly and as accurately as possible. Figure gives a schematic representation of the time course of a typical trial. In each trial, three signals (with a visual extent of ~0.5°) were presented in the centre of a circle (diameter ~8° of visual angle) that remained on the computer screen throughout the experiment. First, a neutral warning signal (blue square) marked the beginning of a new trial. After a 1200 ms delay this warning signal was replaced by a response signal instructing the subject to move the thumb (go trial; green square; 80% probability) or to withhold a response (no-go trial; red square; 20% probability). The required direction of each thumb movement was precued during the interval between warning and response signal, by a blue arrow that was briefly flashed (100 ms) at the position of the warning signal. The five possible directions (90°, 135°, 180°, 225°, and 270°) were always tick-marked on the circle. One of these five movement directions was precued in each trial. There were two task variants, which only differed in the timing of the directional precue. In the early-precue task the direction was specified 600 ms before the response signal, whereas in the late-precue task it was specified 100 ms before the response signal. Subjects were instructed to completely relax their arm muscles during the period preceding the response signal and to respond in go trials only. After go trials, a marker was plotted on the circle to show the direction of the movement just executed. After no-go trials, a centrally presented green or red square informed the subject whether the response was correctly withheld or not.
Before performing the experimental tasks subjects were trained extensively. The aim of the training was to familiarise the subjects with the required stimulus-response associations and to practice complete muscle relaxation whenever movement was not appropriate. Furthermore, subjects who had no prior experience with TMS were familiarised with the technique. A large part of the training was completed in a separate session one or two days before the test session. In the first session the subjects initially performed a short warm-up task. This task was identical to the late-precue task except that the precued directions on subsequent trials were arranged orderly, in an anti-clockwise fashion (90°, 135°, 180°, 225°, 270°, 90°, etc. etc.). Subjects completed at least two blocks of 30 trials. If movements were inaccurate, additional blocks were presented until the experimenters deemed performance to be adequate. Next, the subjects performed three blocks (40 trials each) of the early-precue task and three blocks of the late-precue task (counterbalanced across subjects). Auditory feedback of electric muscle activity was provided continuously during all three tasks. Finally, subjects unfamiliar with TMS were introduced to the technique. In the second session the subjects initially performed the warm-up task while auditory feedback of their muscle activity was provided. Again, at least two blocks of 30 trials were completed. If movements were inaccurate or excessive muscle activity was present prior to movement initiation, additional blocks were presented until performance was deemed adequate. After their movement threshold was determined (see below), the subjects practiced 20 trials of the subsequent early-precue task, which included the application of TMS pulses during task performance. Subjects were asked to ignore any possible interference of TMS and perform the task to their best ability. The experimental tasks were carried out in the final part of the second session. The subjects completed nine blocks of the early-precue task followed by two blocks of the late-precue task. According to our hypothesis, reaction times in the late-precue task were expected to be longer than in the early-precue task. The late-precue task was always performed after the early-precue task. In this way, if a training effect would occur, it would only lead to an underestimation of the precue effect, as the effects would have opposite directions. In a random 280 (~78%) out of the 360 early-precue trials a single TMS pulse was applied at either 900 ms (40 trials), 300 ms (80 trials), or 100 ms (80 trials) before, or 250 ms (80 trials) after the response signal. No TMS was applied in the late-precue task. As TMS frequently evoked involuntary thumb movements, feedback was omitted after TMS trials. Each block consisted of 40 trials, resulting in 360 early-precue trials, (280 with and 80 without TMS) and 80 late-precue trials.
Transcranial magnetic stimulation
TMS was delivered using a figure-of-eight shaped coil (diameter of each wing 70 mm) connected to a Magstim 2002 stimulator (Magstim Company, Whitland, United Kingdom). The coil was positioned tangentially on the left hemiscalp with its handle pointing backward at an angle of about 45 degrees from the midsagittal axis. First, the optimal position for evoking isolated thumb movements was identified. At this position the movement threshold was determined. Movement threshold was defined as the lowest stimulator output evoking a thumb acceleration of ≥ 0.9 ms-2 in at least five out of eight successive stimulations. Stimulation intensity was set slightly above this threshold, on average 49 (± 10) % of maximum stimulator output. Coil position was monitored continuously (BrainSight TMS, Rogue Research, Montreal, Canada) and adjusted whenever its distance to the optimal stimulation position exceeded 5 mm.
Thumb movements were recorded by two miniature uni-axial accelerometers (Model 256–100, sensitivity 10 mV/ms-2; Endevco Corp., San Juan Capistrano, CA) that were mounted on the proximal phalanx of the thumb in orthogonal planes to detect acceleration in the abduction/adduction and extension/flexion axes. Accelerometer signals were conditioned with a gain of 10 (Model 4416B signal conditioner, Endevco Corp.). EMG from the thumb muscle (electrodes over the abductor pollicis brevis 'APB'), index finger muscle (electrodes over first dorsal interosseus 'FDI'), and the wrist muscles (electrodes over flexor carpi radialis 'FCR') of the right hand was recorded using adhesive Ag/AgCl surface electrodes (Kendall-LTP, Chicopee, MA). Electrodes were placed in "belly-tendon" arrangements, following standard skin preparation. EMG signals were amplified with a gain of 250 using an Ekida amplifier (Ekida GmbH, Helmstadt, Germany). Accelerometer and EMG signals were anti-aliasing filtered (1 kHz cut-off), then digitised at a rate of 5 kHz (acceleration resolution 0.15 ms-2/bit, voltage resolution 0.61 μV/bit) using Spike2 software and a Power 1401 A/D converter (Cambridge Electronic Design, Cambridge, United Kingdom). Figure shows some example traces of TMS responses in EMG and accelerometer signals.
Figure 9 Example EMG and accelerometer traces. Example traces of responses to TMS, from two of the subjects. Of each of the two subjects, six arbitrary TMS-trials were selected and the EMG and accelerometer signals recorded in those trials are plotted. The upper (more ...)
Data processing and analysis
Data were processed off-line using MATLAB (MathWorks, Natick, MA). Acceleration and EMG data were digitally filtered (low-pass 100 Hz, band-pass 10–500 Hz, respectively) and segmented into epochs running from 1200 ms before to 800 ms after each response signal. The two accelerometer signals were converted into polar coordinate (magnitude-angle) time series. A peak detection algorithm was applied to the magnitude of this signal to determine onsets and corresponding directions of voluntary and of TMS-evoked movements.
Reaction time (RT) was defined as the latency between the response signal and the first peak of acceleration. Because TMS may influence RT [23
] only trials without TMS were used for the analyses of voluntary movements (with the exception of the analysis that assessed the effect of TMS on RT; see below). Per subject, trials with an RT of more then 2.5 standard deviations from the mean RT were discarded from all analyses. The RT of the early- and late-precue conditions was compared with a two-tailed paired-samples t
-test. The effect of the precue on the direction of the subsequent voluntary thumb movement was analysed with a one-way repeated-measures ANOVA with the within-subjects factor precue (90°, 135°, 180°, 225°, 270°) and the first-peak acceleration angle as dependent variable. To assess the contribution of the three muscles to the different movements, we calculated the average root mean square (RMS) amplitude during the first 150 ms of the EMG bursts associated with the voluntary movements.
We performed an additional analysis on the RTs to assess the effect of a TMS perturbation on motor preparation. For each trial in which TMS was applied during the preparatory interval (at -900 ms, -300 ms, or -100 ms) the RT was determined. Trials with an RT of more than 2.5 standard deviations from the mean RT were discarded. The remaining RTs were compared to the RT from early-precue trials without TMS using a one-way ANOVA with the within-subjects factor time (no TMS, -900 ms, -300 ms, -100 ms).
The required direction of the upcoming movement was precued not before 600 ms prior to the response signal. Therefore the period between 1200 and 600 ms prior to the response signal was termed the baseline interval. Responses to TMS given in this interval (i.e. stimulation time -900 ms) were considered as an individual baseline for analyses of TMS-evoked movements, MEPs, and pretrigger RMS amplitudes.
TMS-evoked movements should have more-or-less constant latencies, because these depend mechanistically on the conduction time of the nervous pathway. Consequently, trials where the latency of the first-peak acceleration of TMS-evoked movements deviated more than 10 ms from the mode across all latencies were discarded from all analyses. To assess whether the precue influenced the direction of subsequent TMS-evoked movements, we defined a "baseline zone". This was a window of ± 30° centred on the average direction of TMS-evoked movements at baseline. We assessed whether there was a temporal modulation of the thumb movement representation by calculating the proportion of TMS-evoked movements that fell outside this baseline zone, at each stimulation time. These proportions were then submitted to a one-way repeated-measures ANOVA with time as a within-subjects factor (-900 ms, -600 ms, -100 ms, 250 ms). A further analysis was conducted to elucidate whether any changes in the thumb movement representation reflected the direction that was precued. Therefore, we determined the proportion of TMS-evoked movements that fell within a ± 30° window centred on the direction that had been precued (i.e. "precued target zone"). Analogous to the previous analysis, the proportion of TMS-evoked movements within the precued target zone was analysed with a one-way repeated-measures ANOVA with the within-subjects factor time (-900 ms, -600 ms, -100 ms, 250 ms).
Corticospinal excitability was assessed by the peak-to-peak MEP amplitude between 10 and 50 ms after the TMS trigger. To make sure the target muscles were at rest during the critical period of each trial, a trial was discarded from all analyses if voluntary EMG during the 200 ms preceding the TMS pulse or preceding the response signal exceeded 50 μV. In addition, the EMG RMS amplitudes 100 ms prior to TMS were calculated. To reduce between-subject variability, the MEP and pretrigger RMS amplitudes were normalised to the average MEP or average pretrigger RMS amplitude (respectively) across the three muscles measured at baseline (a value of 1 was assigned and all other values expressed relative to this value). The normalised MEP amplitudes were initially submitted to a three-way repeated-measures ANOVA with within-subjects factors muscle (APB, FDI, FCR), precue (90°, 135°, 180°, 225°, 270°) and time (-900 ms -600 ms, -100 ms, 250 ms). Significant interactions were further specified with separate two- and one-way ANOVAs. To assess whether preliminary muscle activation could explain any temporal modulation of TMS-evoked movements or MEP amplitudes, one-way ANOVAs with the within-subjects factor time were also conducted on the normalised pretrigger RMS amplitudes of each muscle.
Degrees of freedom were adjusted with the Greenhouse-Geisser epsilon if the sphericity assumption was not met, but for statistical interpretation uncorrected degrees of freedom are reported. Statistical significance was set at the 0.05 level. Significant effects in the omnibus tests were taken as justification for further specification by post-hoc Fisher's LSD tests. Unless stated otherwise, data are presented as mean ± standard error of mean (SE).