SEARCH

SEARCH BY CITATION

Keywords:

  • dorsal premotor cortex;
  • implicit motor learning;
  • offline motor learning;
  • online motor learning;
  • repetitive transcranial magnetic stimulation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Consolidation of motor memories associated with skilled practice can occur both online, concurrent with practice, and offline, after practice has ended. The current study investigated the role of dorsal premotor cortex (PMd) in early offline motor memory consolidation of implicit sequence-specific learning. Thirty-three participants were assigned to one of three groups of repetitive transcranial magnetic stimulation (rTMS) over left PMd (5 Hz, 1 Hz or control) immediately following practice of a novel continuous tracking task. There was no additional practice following rTMS. This procedure was repeated for 4 days. The continuous tracking task contained a repeated sequence that could be learned implicitly and random sequences that could not. On a separate fifth day, a retention test was performed to assess implicit sequence-specific motor learning of the task. Tracking error was decreased for the group who received 1 Hz rTMS over the PMd during the early consolidation period immediately following practice compared with control or 5 Hz rTMS. Enhanced sequence-specific learning with 1 Hz rTMS following practice was due to greater offline consolidation, not differences in online learning between the groups within practice days. A follow-up experiment revealed that stimulation of PMd following practice did not differentially change motor cortical excitability, suggesting that changes in offline consolidation can be largely attributed to stimulation-induced changes in PMd. These findings support a differential role for the PMd in support of online and offline sequence-specific learning of a visuomotor task and offer converging evidence for competing memory systems.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Skilled motor practice facilitates the formation of an internal model of movement, which may then be used to anticipate task-specific requirements at a later time (Shadmehr & Holcomb, 1997). Internal models are most susceptible to interference during and immediately following practice and become less susceptible to interference over time through persistent neural activity, a process called consolidation (Brashers-Krug et al., 1996; Robertson et al., 2004).

Motor memory consolidation can take place both explicitly, with conscious awareness, or implicitly, without conscious awareness of the skill being performed (Robertson et al., 2004). The neural processes of consolidation can take two forms: (i) online improvements that occur concurrent with practice or (ii) offline improvements that develop following the termination of practice (Brashers-Krug et al., 1996; Robertson et al., 2004). Importantly, these two processes are not completely independent or mutually exclusive.

Given its known role in the selection of movements (Kalaska & Crammond, 1995; Rushworth et al., 2003) and implicit motor learning (Ohbayashi et al., 2003; Meehan et al., 2011), the dorsal premotor cortex (PMd) is a logical candidate for involvement in motor memory consolidation. Our group reported improved implicit sequence-specific motor learning when 5 Hz repetitive transcranial magnetic stimulation (rTMS) was delivered over the PMd prior to skilled motor practice of a continuous tracking task (Boyd & Linsdell, 2009). Yet it is not clear whether improvements noted when PMd stimulation precedes motor practice result from only online or a combination of online and offline consolidation of sequence-specific elements.

The current work sought to directly assess the involvement of PMd in offline consolidation of skilled motor practice. In contrast to our previous work (Boyd & Linsdell, 2009), three groups received either 1 Hz, 5 Hz or control rTMS immediately following practice of a continuous visuomotor tracking task (Experiment 1). Based on our previous work, it was hypothesized that 5 Hz rTMS immediately following practice would enhance while 1 Hz rTMS would suppress motor learning compared with control stimulation. However, the effects of TMS are known to be ‘state dependent’ (Silvanto et al., 2008). State-dependence has been demonstrated during both perceptual and cognitive tasks where prior or concurrent neural activity (Silvanto et al., 2007b; Arai et al., 2011) and/or task-specific elements (Bestmann et al., 2008; Cohen & Robertson, 2011) influence expected outcomes. An alternative hypothesis is that 1 Hz rTMS, typically associated with inhibition, over PMd immediately following practice may enhance implicit sequence-specific motor learning through state-dependent mechanisms.

To confirm changes in offline consolidation could be attributed to PMd rather than altered primary motor cortex (M1) excitability, via PMd projections to M1 (Cavada & Goldman-Rakic, 1989; Hanakawa, 2011) motor-evoked potentials were elicited before and after practice preceded by either 1 Hz, 5 Hz or control rTMS (Experiment 2). We hypothesized that rTMS over the PMd immediately following practice would not alter M1 excitability and that any change in offline consolidation noted in Experiment 1 could be attributed to the PMd.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Experiment 1

Participants

Thirty-three healthy, right-handed participants (20 males and 13 females, age range 20–48 years) were enrolled in the study (Table 1). All participants provided informed consent, which complied with the Code of Ethics of the World Medical Association (Declaration of Helsinki), printed in the British Medical Journal (18 July 1964). Written informed consent of each subject was received. The University of British Columbia Clinical Research Ethics Board approved the protocol. Participants were excluded from the study if they showed any sign of neurological impairment or disease, or if they had any colour blindness that might impair response ability.

Table 1. Participant characteristics for Experiments 1 and 2
 Experiment 1Experiment 2
5 Hz1 HzControl5 Hz1 HzControl
Age (years, mean ± SD)24 ± 2.928 ± 5.623 ± 2.327 ± 4.827 ± 4.527 ± 4.5
Gender6M, 5F6M, 5F8M, 3F4M, 6F4M, 6F4M, 6F
Experimental design

The experiment took place over five testing sessions, on separate days, completed within 2 weeks. Prior to the start of the experiment participants were randomly assigned to one of three groups. The protocol was the same for each group, with the exception of the type of rTMS that followed practice of the continuous tracking (CT) task. One group received 1 Hz rTMS over the left PMd, the second group received 5 Hz rTMS over the left PMd, while the third group received sham stimulation over the left PMd as a control condition.

Each group completed four CT practice sessions; practice was immediately followed by rTMS according to group (days 1–4) (Fig. 1). To evaluate motor learning, a retention test was conducted on a separate day (day 5). In each practice session participants performed three blocks (30 trials) of the CT task. Practice sessions were scheduled to accommodate the participant but no more than 48 h elapsed between any of the sessions. On day 5, the retention test consisted of one block (10 trials) of continuous tracking without application of rTMS. The retention test was used to disentangle performance effects from more permanent changes in behaviour associated with motor learning (Salmoni et al., 1984).

image

Figure 1. Experimental protocol for each group. Note that the Control group followed the protocol for either the 5 Hz or the 1 Hz group. The protocol for the Control group was randomized across participants. Inset: one 30-s trial within a 5-min block of task practice. This 30-s trial was repeated 10 times within a 5-min block.

Download figure to PowerPoint

CT task

The CT task used in the current study was similar to that previously reported (Boyd & Linsdell, 2009). During the CT task participants were seated in front of a computer monitor. Holding a joystick in their right hand, participants tracked a target as it moved in a sine–cosine waveform. The target appeared as an open white circle and participant movements were shown as a red dot (Boyd & Linsdell, 2009). Joystick position sampling and all stimuli were presented at 40 Hz using custom software developed on the LabView platform (v. 8.6; National Instruments Co., Newbury, UK).

The pattern of the target movement was predefined according to a method modified from Wulf & Schmidt (1997). A unique 32-s trial was constructed from one 2-s baseline and three 10-s sine–cosine segments. One block consisted of ten 32-s trials. The middle third of each tracking pattern was repeated and identical across practice and retention (Boyd & Linsdell, 2009). The pattern was unknown to the participants and was constructed from the polynomial equation described by Wulf & Schmidt (1997) with the following general form:

  • display math

The middle (repeated) segment was constructed by using the same coefficients for every trial (b0 = 2.0, a1=−4.0, b1 = 3.0, b2=−3.6, a3 = 3.9, b3 = 4.5, a4 = 0.0, b4 = 1.0, a5=−3.8, b5=−0.5, a6 = 1.0 and b6 = 2.5). The first and third segments of the tracking pattern were generated randomly using coefficients ranging from −5.0 to 5.0. A different random sequence was used for both the first and third segments of every trial. There were 10 separate reversals in the direction in each third of the tracking pattern. The random and repeated patterns were equated for difficulty by ensuring that the overall excursion of each random sequence fell within a range of that required by the repeated sequence. Neither the trajectories of the target nor the participants' movements left a visible train on the screen and thus participants could not visualize the entire pattern. The same sets of trials were practised by all of the participants to ensure uniformity so that the random segments were the same for each participant.

Participants were not informed of the repeating sequence but were instructed daily to track the target as accurately as possible by controlling the position of the cursor with the joystick.

Transcranial magnetic stimulation

Transcranial magnetic stimulation was delivered with a Magstim Super Rapid2 stimulator using a 70-mm figure-of-eight air-cooled coil (Magstim Company Ltd., Whitland, UK). Participants were seated in a semi-reclined dental chair with their arms bent and supported by armrests. The TMS coil was orientated tangentially to the scalp with the handle at 45° to the midline in a posterior lateral orientation. Prior to the experiment, high-resolution anatomical magnetic resonance images (MRIs) were acquired for each participant (TR = 12.4 ms, TE = 5.4 ms, flip angle θ = 35°, FOV = 256 mm, 170 slices, 1-mm thickness) at the UBC MRI Research Centre on a Philips Achieva 3.0 T whole body MRI scanner (Phillips Healthcare, Andover, MD, USA) using a sensitivity encoding head coil (SENSE). These images were then imported into the BrainSight™ TMS neuronavigation software (BrainSight 2.0, Rogue Research Inc., Montreal, QC, Canada) to allow for stereotaxic registration of the TMS coil with the participants' anatomy for online control of coil positioning during each session and across days.

Surface electromyography over the participants' right flexor carpi radialis (FCR) was monitored using the evoked potential unit of the Super Rapid2 control unit (Magstim Company, Ltd) (Boyd & Linsdell, 2009). Initially, the FCR representation was marked on the participants' anatomical MRI as the medial edge of the left ‘hand knob’. This point acted as a starting point for determination of the motor ‘hot spot’ for FCR. Motor-evoked potentials (MEPs) were then used to determine the coil position that evoked the maximal response in the right FCR. The location and trajectory of the coil over left primary motor cortex (M1) were marked using the BrainSight™ stereotaxic software to minimize variability within and across days. Resting motor threshold (RMT) was determined for each participant as the percentage of stimulator output that elicited an MEP of ≥ 50 μV peak to peak on five out of 10 trials.

The site of stimulation for the left PMd was marked in Brainsight™ by moving one gyrus forward from the FCR ‘hot spot’ (Boyd & Linsdell, 2009). The PMd location was confirmed as the posterior aspect of the middle frontal gyrus (Munchau et al., 2002; Fridman et al., 2004). Isolation of this area from M1 was verified using single pulses to ensure that: (i) there was no electromyographic record of muscle activity recorded over the FCR, and (ii) there were no visually apparent muscle twitches in the forearm or hand. Once confirmed, the location and trajectory of the coil were recorded using BrainSight™ to ensure the consistent stimulation of the PMd across days (Boyd & Linsdell, 2009).

Five hertz stimulation consisted of 1200 pulses delivered in 10-s trains with an inter-train interval of 10 s. Intensity was set to 110% RMT. 1 Hz stimulation consisted of 1200 pulses delivered in 10-s trains with an inter-train interval of 1 s and an intensity of 110% RMT. Control stimulation was delivered using a custom sham coil that looks and sounds similar to the rTMS coil but does not induce any current in the underlying cortex (Magstim Company Ltd.). The parameters of the control stimulation were counterbalanced across participants such that six participants received control stimulation that mimicked 5 Hz stimulation and five participants received control stimulation that mimicked 1 Hz stimulation.

The rTMS parameters employed have been shown to induce an after-effect of approximately 15 min (Chen et al., 2003). To ensure that there was no interference with the effects of the rTMS protocols upon consolidation of motor practice, participants were required to remain quietly seated for 15 min following the end of stimulation.

Explicit awareness of the repeated sequence

Following the retention test on day 5, participants were shown ten 30-s trials of continuous target movements and asked to decide if they recognized any pattern that they performed during the practice sessions. Out of the 10 trials, three contained the ‘true’ middle sequence, i.e. the same as the repeated practice sequence, and seven were foils. Individuals were considered to have explicit awareness of the repeated sequence if they could both correctly recognize 2 of the 3 repeated sequences and properly label 5 of 7 of the foils as having not been seen before (Boyd & Linsdell, 2009).

Behavioural outcome measures

Motor performance was evaluated across practice and retention in two ways. Our primary analysis consisted of the overall root mean square error (RMSE). Overall RMSE reflects the average of the difference waveform derived by subtracting the instantaneous position of the target from the participant's location. This score was calculated separately for random and repeated sequences and averaged for all trials within a block (Wulf & Schmidt, 1997; Boyd & Winstein, 2004b; Vidoni & Boyd, 2009). The difference between overall RMSE during random and repeated sequence tracking reflects implicit learning and was used to evaluate reductions in tracking errors across practice and at retention. Random tracking performance was assessed using the second random sequence (Boyd & Winstein, 2004b; Boyd & Linsdell, 2009).

As overall RMSE reflects both spatial accuracy and temporal lag, improvement on each of these components of movement was also assessed (Boyd & Winstein, 2004a).Time lag of tracking is the time (in milliseconds) corresponding to the maximal cross-correlation coefficient and represents the temporal distance from the target. Spatial error is the residual RMSE score that remains following adjustment of the participant's cursor position to account for the time lag of tracking. Time lag scores in larger negative numbers indicate greater time lag of tracking, while a zero value represents no tracking time lag between participant and target. Lower RMSE scores indicate less overall error and show improved motor performance.

Statistical analysis

Statistical analyses were performed in three steps. First, improvement in performance during the acquisition phase (days 1–4) was assessed for overall RMSE, spatial error and time lag using separate 3 (Group: 1 Hz, 5 Hz, Control rTMS) × 12 (Block: 1–12) mixed-measures anovas for the random and repeated sequences. Group was treated as a between-subjects factor and Block was treated as a repeated measures factor. In all cases the dependent variables (overall RMSE, spatial error and time lag) were log transformed as Maulchy's sphericity test revealed that raw scores across blocks violated the sphericity assumption for each dependent variable and both sequences.

Second, implicit sequence-specific learning at retention was examined for overall RMSE, spatial error and time lag using three separate 3 (Group: 1 Hz, 5 Hz, Control rTMS) × 2 (Sequence: Random, Repeated) mixed-measures anovas. Group was treated as a between-subjects factor and Sequence was treated as a repeated measures factor. As implicit sequence-specific learning is defined as lower error/less lag during repeated compared with random sequence tracking, significant Group × Sequence interactions were investigated using contrasts comparing repeated vs. random sequence tracking performance within each group to determine if implicit sequence-specific learning was evident in each group. Bonferroni correction was applied with the corrected threshold of = 0.033 to correct for multiple comparisons.

Finally, we assessed the impact of the rTMS post-practice upon the retention of implicit learning from day to day. Offline sequence-specific motor learning was defined as the change in sequence-specific learning (random performance – repeated performance) from the previous day to the first block of the subsequent day (Robertson et al., 2004; Robertson & Cohen, 2006). Separate 3 (Group: 1 Hz, 5 Hz, Control rTMS) × 4 (Consolidation Period: Day 1, Day 2, Day 3 and Day 4) mixed-measures anovas were run to assess offline sequence-specific motor learning for RMSE, spatial error and time lag. Group was treated as a between-subjects factor and Consolidation Period was treated as a repeated measures factor.

To ensure that differences in offline learning could not be attributed to differences across the groups in online consolidation we also ran three separate 3 (Group: 1 Hz, 5 Hz, Control rTMS) × 4 (Day: Day 1, Day 2, Day 3 and Day 4) mixed-measures anovas to assess difference in online sequence-specific learning for RMSE, spatial error and time lag. Group was treated as a between-subjects factor and Consolidation Period was treated as a repeated measure. Online sequence-specific learning was defined as the change in sequence-specific performance from Block 1 to Block 3 within each day. Statistical analyses were performed in spss v.20. For all analyses Group was treated as a between-subjects factor. All other variables were treated as a repeated measures factor. Greenhouse-Geisser epsilon corrections and Bonferonni corrections were applied where appropriate.

Experiment 2

The aim of experiment 2 was to determine whether motor practice followed by stimulation over the left PMd had an effect on the excitability of M1.

Participants

Thirty healthy, right-handed participants (12 males and 18 females, age range 20–33 years) were enrolled in the study (Table 1). All participants provided informed consent; the University of British Columbia Clinical Research Ethics Board approved the protocol. Participants were excluded from the study if they showed any sign of neurological impairment or disease, or if they had any colour blindness that might impair response ability.

Procedure

The experiment consisted of a single session. Prior to the start of the experiment participants were randomly assigned to one of three groups. For each group, RMT and M1 excitability (indexed by the amplitude of MEPs) were assessed before and after each participant completed three blocks of continuous tracking practice paired with rTMS. The testing protocol was the same for each group; only the type of rTMS that followed task practice differed. As in Experiment 1, one group received 1 Hz rTMS over the left PMd, the second received 5 Hz rTMS over the left PMd, while the third group received sham stimulation over the left PMd.

Continuous tracking

The CT task was the same as that described for Experiment 1. Only one practice session containing three blocks of CT task practice was completed.

Transcranial magnetic stimulation

The procedures for delivering rTMS were the same as those outlined in Experiment 1. However, to gauge the effect of rTMS over the PMd on M1 two additional measures were collected: (i) RMT was established before and after CT task practice plus rTMS, and (ii) MEPs were elicited by 10 single pulses at 110% RMT before and after practice plus rTMS.

Analyses

The effect of stimulation over the PMd on RMT and MEP amplitude was assessed using separate 3 (Group: 1 Hz, 5 Hz, Control rTMS) × 2 (Time: Pre, Post) mixed-measures anovas. Group was treated as a between-subjects factor. Time was treated as a repeated measures factor. Linear contrasts corrected for multiple comparisons using the Bonferonni correction were applied where appropriate.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Experiment 1

Practice

The Group by Block mixed-measures anova considering practice performance with RMSE as the dependent measure revealed a main effect of Block for both the random (F11,330 = 19.66, < 0.001) and repeated (F11,330 = 14.70, < 0.001) sequences. The main effect of Block can be attributed to a decrease in RMSE across blocks with practice for both repeated and random sequences (Fig. 2A and B). Group by Block mixed-measures anovas upon spatial error and lag revealed that the improvement in RMSE across practice block can be attributed to both reduced spatial error (Random: F11,330 = 13.33, < 0.001; Repeated: F11,330 = 9.41, < 0.001) (Fig. 2C and D) and time lag (Random: F11,330 = 19.66, < 0.001; Repeated: F11,330 = 12.17, < 0.001) (Fig. 2E and F).

image

Figure 2. Tracking performance across the 4 days of practice (three blocks per day) and retention testing (one block labelled ‘Ret’) for: (A) overall tracking error for the repeated sequence, (B) overall tracking error for random sequences, (C) spatial tracking error for the repeated sequence, (D) spatial tracking error for random sequences, (E) time lag of tracking for the repeated sequence, and (F) time lag of tracking for the random sequences. For overall and spatial tracking error lower scores indicate improved performance. For time lag of tracking scores closer to zero represent improved performance. Data are mean; error bars are SEM.

Download figure to PowerPoint

Implicit sequence-specific motor learning

The mixed-measures Group by Sequence anova on Overall RMSE at retention (Day 5) revealed a significant interaction (F2,30 = 3.81; = 0.033), as well as a trend for a main effect of Sequence (F2,30 = 3.27, = 0.081). Inspection of the data (Fig. 3A) shows that the interaction can be attributed to lower Overall RMSE (i.e. improved performance) during repeated compared with random sequence tracking at retention in individuals who received 1 Hz rTMS during the consolidation period immediately following practice (contrast, = 0.007). Reduced error during repeated compared with during random sequence tracking is indicative of implicit sequence-specific learning in this group. In contrast, overall RMSE during repeated compared with random sequence tracking at the retention test was not different for the groups that received 5 Hz rTMS or control stimulation (= 0.96 and 0.89, respectively).

image

Figure 3. Change in RMSE for random and repeated sequences from Early Practice to Retention for each group for (A) Overall tracking error and (B) Spatial tracking error. Sequence-specific learning is reflected by greater change in repeated sequence tracking performance from Early Practice to Retention relative to random sequence tracking performance. Data are mean; error bars are SEM.

Download figure to PowerPoint

The corresponding Group by Sequence anova using spatial RMSE as the dependent measure revealed a main effect of Sequence (F1,30 = 3.84, = 0.06). Post-hoc t-tests comparing repeated vs. random sequence spatial RMSE suggest that the trend for a main effect can be attributed to reduced spatial RMSE during repeated compared with random sequence tracking at Retention (= 0.014; Fig. 3B) in the 1 Hz group. There were no differences in spatial RMSE for individuals who received 5 Hz rTMS or control stimulation.

The Group by Sequence anova for time lag of tracking failed to reveal any significant effects.

Offline learning

Mixed measures Group by Sequence anova on change in RMSE reflecting implicit motor learning from the previous day to the first block of the subsequent day revealed main effects of Group (F2,30 = 6.60, = 0.004) and Consolidation Period (F3,90 = 4.23, = 0.017). Scheffe's post-hoc tests revealed that the main effect of Group can be attributed to significantly greater sequence-specific offline learning in the 1 Hz group compared with the Control and 5 Hz rTMS groups (= 0.030 and 0.003, respectively) (Fig. 4A – dark grey bars). The main effect of Sequence can be attributed to greater consolidation of implicit motor learning from Day 4 to the retention test compared with consolidation between Day 2 to Day 3 and Day 3 to Day 4 (< 0.001 and = 0.024, respectively) (Fig. 4B – dark grey bars). The Group by Sequence anova on spatial error revealed main effects of Group (F2,30 = 5.10, < 0.012) and Consolidation Period (F3,90 = 4.09, < 0.014). The main effects of Group (Fig. 4A – light grey bars) and Consolidation Period (Fig. 4B – light grey bars) reveal that the changes in RMSE can be attributed to consolidation of spatial accuracy.

image

Figure 4. Main effects for offline learning sequence-specific learning. (A) Main effect of Group – mean change in offline retention of sequence-specific learning for each group, collapsed across retention period. Means are shown for both overall RMSE (dark grey) and spatial error (light grey). (B) Main effect of Day – mean change in offline retention of sequence-specific learning for each retention period, collapsed across group. Means are shown for overall RMSE (dark grey) and spatial error (light grey). ‘Ret’ = retention day. Bars represent SEM.

Download figure to PowerPoint

The mixed-measures Group by Sequence anova with time lag as the dependent measure failed to reveal any effects.

Online learning

None of the analyses on RMSE, spatial accuracy or lag revealed any effects associated with change in implicit performance from Block 1 to Block 3 on each day of practice. Online learning within each practice day was consistent for all groups.

Explicit awareness of the repeating sequence

Three of the 11 individuals in the 5 Hz rTMS group acquired sufficient explicit awareness of the repeating sequence to be able to recognize it at the recognition test. This was also the case for two individuals in the 1 Hz rTMS group and one individual in the Control group.

Experiment 2

The mixed-measures Group by Time anovas performed on RMT and MEP amplitude failed to reveal any significant effects of the varied forms of rTMS following continuous tracking on excitability in M1 (Table 2).

Table 2. Mean change score (pre- to post-stimulation) for peak-to-peak MEP amplitude and resting motor threshold across groups (Experiment 2); values are mean ± SEM
 5 Hz1 HzControl
MEP amplitude (μV)79 ± 4627 ± 2861 ± 106
RMT (% stimulator output)−0.3 ± 0.5−0.6 ± 0.40.1 ± 0.2

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

The present study is the first to demonstrate the cumulative impact of rTMS over PMd immediately following practice upon consolidation of implicit sequence-specific motor learning. While all three experimental groups (1 Hz rTMS, 5 Hz rTMS and sham stimulation) demonstrated improvement in performance over time, only the group receiving 1 Hz rTMS over the PMd immediately following task practice enhanced offline learning of an implicit motor skill (Experiment 1). Enhanced implicit sequence-specific learning with 1 Hz rTMS following practice was largely explained by improved spatial rather than temporal accuracy of movements (Experiment 1). Furthermore, enhanced motor learning associated with 1 Hz rTMS over the PMd during early consolidation does not appear to be attributable to spread of stimulation to M1 or to PMd to M1 connections, as M1 excitability was not changed by rTMS over PMd (Experiment 2).

The enhancement of motor learning following application of 1 Hz rTMS over PMd immediately after practice of the continuous visuomotor tracking task differs from our previous results (Boyd & Linsdell, 2009). In our past work 5 Hz rTMS delivered prior to practice of the same visuomotor task as was employed here enhanced implicit sequence-specific motor learning across multiple days of practice. The difference between our current and our previous studies suggests state-dependency in the form of a task-dependent role for PMd during online performance and offline consolidation of implicit sequence-specific learning of a visuomotor task.

Given its anatomical location and functional connectivity, the PMd is a likely convergence point for cognition and motor control. PMd is generally associated with explicit declarative aspects of motor learning. While the PMd has been implicated in facilitating the transition between implicitly learned movements that constitute a sequence (Mushiake et al., 1991), activity in the PMd is reduced when explicit awareness of the implicit motor sequences is gained (Hazeltine et al., 1997; Honda et al., 1998). During online learning it is likely that the PMd serves to enhance implicit sequence-specific learning by linking specific movements which are dependent upon sensory cues (Nowak et al., 2009; Taubert et al., 2010). This role may be particularly important during interleaved practice and may explain why anodal transcranial direct current stimulation over the PMd during constant repetitive practice does not result in improved consolidation of performance gains (Nitsche et al., 2003; Kantak et al., 2012). In contrast, early offline consolidation of information relating to sequencing of action selection may interfere with early consolidation of more procedural elements relating to individual movements, which are most likely represented in M1 (Muellbacher et al., 2002; Wilkinson et al., 2010). This may result from early offline consolidation of information being more reliant on a declarative memory and thus more explicit. This assumption is consistent with observations of differential rates of consolidation for explicit declarative memories relative to procedural memory (Brown & Robertson, 2007a; Ghilardi et al., 2009; Galea et al., 2010) and competition between procedural and declarative memory systems (Brown & Robertson, 2007a,b; Galea et al., 2010). Interference may occur even when explicit instruction is not given and participants have not autogenously acquired declarative knowledge of a sequence through practice (Vidoni & Boyd, 2007). Therefore, reducing the cortical excitability of PMd through 1 Hz rTMS during early offline consolidation may relieve suppression of procedural representations in M1 during this critical period and facilitate an early boost in procedural learning not seen at later stages of offline consolidation (Hotermans et al., 2008).

Another interesting result was the lack of dissociation between implicit motor sequence learning for the 5 Hz and sham stimulation groups. Relative to the sham control group, one might expect the 5 Hz group to show the opposite effect to that induced by 1 Hz rTMS. The fact that 5 Hz rTMS immediately following practice had no impact upon offline consolidation of implicit sequence-specific learning suggests that increasing the cortical excitability of the PMd during early offline learning was not sufficient to enhance declarative consolidation beyond that already occurring during the period of early offline consolidation probed here (~5–30 min post-practice). The lack of sequence-specific learning despite the same amount of practice as the 1 Hz group suggests a state-dependent element where current activity in PMd, the activity producing the interference effect, is not enhanced by stimulating PMd. The net result is that offline consolidation and implicit sequence-specific motor learning are similar to those seen in the control group in the absence of stimulation, where any learning is associated with gains in sensorimotor efficiency rather than sequence-specific elements. This further supports a competitive model of declarative/procedural consolidation where competition is biased towards the developing declarative memories.

Interestingly, the enhancement associated with cumulative 1 Hz rTMS over the PMd appeared to reflect retained improvement in spatial accuracy rather than a reduction in response lag. While these two variables are not completely independent of each other our results suggest that consolidation of spatial aspects of a motor sequence may be mediated by PMd and M1 networks but that procedural elements of these representations are stored in M1 (Muellbacher et al., 2002). The relative insensitivity of temporal aspects to 1 Hz rTMS during early offline consolidation highlights the importance of other cortical areas for implicit sequence-specific learning, such as the supplementary motor area (Mushiake et al., 1991) and cerebellum (Boyd & Winstein, 2004a). In particular, the changes in spatial tracking error may relate to the role of the PMd in preparing aspects of spatial working memory during externally guided movements (Mushiake et al., 1991).

Traditionally, 1 Hz rTMS has been associated with inhibitory effects that persist beyond cessation of stimulation (Wassermann et al., 1996; Chen et al., 2003; Vidoni et al., 2010). Our interpretation of our results is based upon this assumption, but an alternative explanation may be that enhanced implicit sequence-specific learning observed following 1 Hz rTMS post-practice is linked to state-dependent effects present during application of the 1 Hz rTMS. Silvanto et al. (2007a,b) and Silvanto & Pascual-Leone (2008) demonstrated similar state-dependent effects in the visual cortex using adaptation paradigms. Therefore, it cannot be ruled out that resonant activity within the PMd, tied to online learning that persisted into the early period of offline consolidation, may have caused 1 Hz rTMS to enhance the PMd contributions to early offline consolidation. Under this scenario, 1 Hz rTMS may serve to reduce neural noise of weak non-sequence-related neural activity within the PMd to a greater extent than the elevated residual sequence-related neural activity tied to online practice. The net result would be akin to the phenomena of surround inhibition reported in the motor cortex that enhances motor ability (Hallett, 2004; Beck & Hallett, 2010), the visual cortex that enhances visual perception (Angelucci et al., 2002) and the somatosensory cortex that enhances tactile acuity (Drevets et al., 1995). State dependency would also explain the lack of effect elicited by 5 Hz rTMS where both the sequence-related and non-sequence-related neural activity would be facilitated. However, given the already elevated excitability in the neurons involved with the repeated sequence representation, the effects of the rTMS would be more pronounced in the less active neural pathways representing the random sequence compared with the already excited neural pathways representing the repeated sequence (Bienenstock et al., 1982; Kuo et al., 2008). The net result would be a reduction in the difference between the signal (repeated sequence neural activity) and the noise (random sequence neural activity). One limitation to the current work is that we are unable to directly assess changes in cortical excitability of the PMd itself. Future work is needed to determine whether rTMS following practice of interleaved random and repeated sequences can elicit state dependency during the period of early offline consolidation.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Our data highlight the potential differential roles for the PMd in implicit motor learning and early offline motor memory consolidation of a novel motor task. The results confirm past work demonstrating that with practice participants can implicitly learn a repeated sequence (Brashers-Krug et al., 1996; Shadmehr & Holcomb, 1997; Meehan et al., 2011) and that sequence-specific learning can be altered via rTMS (Boyd & Linsdell, 2009). Importantly, we found that 1 Hz rTMS over the PMd during early consolidation improved sequence-specific implicit motor learning, probably by reducing competition between consolidation of motor parameters and action selection following interleaved practice. Applying rTMS during early consolidation may be an adjunctive mechanism to enhance gains associated with practice through consolidation of specific elements of motor memory.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Support was provided to S.K.M. by the Canadian Institutes of Health Research and the Michael Smith Foundation for Health Research and to L.A.B. by the Canada Research Chairs and the Michael Smith Foundation for Health Research. This work was also supported by awards from the Natural Sciences and Engineering Research Council of Canada (Award #401890) and the Vancouver Coastal Health Research Institute to L.A.B.

Abbreviations
CT

continuous tracking

FCR

flexor carpi radialis

MEP

motor-evoked potential

MRI

magnetic resonance image/imaging

PMd

dorsal premotor cortex

RMSE

root mean square error

RMT

resting motor threshold

rTMS

repetitive transcranial magnetic stimulation

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References