Hebbian priming of human motor learning

Across organisms, motor learning is governed by experience-dependent plasticity in neural circuits and depends on precise timing of pre- and postsynaptic activity in line with Hebbian mechanisms. Non-invasive neuromodulation can modulate these intrinsic learning processes. Paired corticospinal-motoneuronal stimulation (PCMS) pairs TMS-evoked descending corticospinal volleys with antidromic volleys from peripheral nerve stimulation (PNS) at the level of corticomotoneuronal (CM) synapses to induce spike-timing-dependent plasticity (STDP). Prior work shows PCMS can bidirectionally modulate CM transmission and transiently improve motor function after spinal cord injury, but it is unknown whether PCMS can promote motor learning and how it interacts with practice-dependent plasticity. The present study investigated whether PCMS, timed to target CM synapses relevant for a ballistic index finger movement task, could prime subsequent motor learning. The rationale was that ballistic performance requires high voluntary drive and motor unit firing, and enhanced CM synaptic efficiency could yield greater motoneuron input, increasing peak acceleration and potentially contributing to early improvements in ballistic motor learning.

Background literature establishes that motor learning reflects timing-dependent synaptic plasticity consistent with Hebbian/STDP rules. PCMS in humans can induce bidirectional changes in CM transmission by precisely timing cortical and peripheral volleys. Previous studies reported PCMS-induced increases in CM transmission and transient motor function gains in spinal cord injury, and modulation of corticospinal excitability with specific inter-arrival intervals (e.g., ~2 ms facilitating CM transmission). However, effects on motor learning per se and interactions with experience-dependent plasticity were unclear. Related paradigms like paired associative stimulation (PAS) target cortical sensorimotor circuits with defined interstimulus intervals and can modulate MEPs. Evidence also suggests ballistic training may enhance spinal-level transmission (e.g., CMEPs) with minimal changes in F-waves, implicating CM synapses. Collectively, prior work motivates testing whether PCMS directed at CM synapses can prime learning, and whether priming depends on spike-timing order and temporal proximity.

Design: Four experiments with healthy young adults examined effects of PCMS on ballistic motor learning and corticospinal excitability (MEPs).

• Participants: Total N=66 (32 males), 20–30 years, right-handed (except one), screened healthy. Ethics approved (H-17019671).
• Task: Ballistic right index finger flexions in a constrained plane using a handle-mounted accelerometer. Trials: 1 s window every 4 s. Practice: 3 blocks of 50 trials (150 total) with 2-min breaks. Feedback (score normalized to post-PCMS) after each practice trial; encouragement at least every 10th trial. Familiarization: 5 trials on first day.
• Outcomes: Peak acceleration (10 trials at baseline, post-stimulation, post-practice). EMG (FDI) during task to compute EMG RMS (onset to +70 ms) and rate of EMG rise (onset to +30 ms), both normalized to Mmax. Corticospinal excitability via TMS-evoked MEPs (20 pulses/time-point), normalized to Mmax. Peripheral excitability via Mmax and F-waves. Time points: baseline, post-stimulation, post-practice; in Experiment IV extended to 0, 15, 30, 45, 60 min post.
• Stimulation/Recording: EMG from right FDI (Ag-AgCl electrodes). TMS: Magstim 200, D70 figure-of-eight coil over left M1 FDI hotspot, handle at 45° PA current. rMT = intensity eliciting ≥0.05 mV MEPs in 5/10. Test MEPs at 120% rMT. Coil position stabilized with neuronavigation. Central conduction time measured via MEP latency during 10% MVC for PCMS timing. PNS: Ulnar nerve at wrist (DS7A, 200 µs) to elicit Mmax and F-waves; latencies used to compute peripheral conduction time and individualize interstimulus intervals.
• PCMS protocols: 100 paired stimuli at 0.1 Hz. TMS at 150% rMT; PNS at 130% Mmax. Inter-arrival interval (IAI) at CM synapse individualized using peripheral and central conduction times. Protocols: PCMS+ (TMS volley arrives 2 ms before antidromic PNS volley at CM synapse; IAI −2 ms), PCMS− (IAI +15 ms; designed to depress CM transmission), PCMS_coupled-control (PNS precedes TMS by 100 ms; IAI +100 ms, outside spinal interaction window). Sham: PNS just above perceptual threshold; TMS coil inverted.
• Unpaired protocols (Experiment IV): rTMS alone or rPNS alone using identical intensities/frequencies as paired protocols (0.1 Hz), to test if pairing is required for excitability changes.
• Experimental setups: • Experiment I (between-groups, N=26): PCMS+ then practice vs Rest (no paired stimuli) then practice. • Experiment II (double-blind, sham-controlled, N=20): PCMS+ then practice vs Sham then practice; retention test 7 days later (50 trials). • Experiment III (within-subject crossover, N=18): three sessions (PCMS+, PCMS−, PCMS_coupled-control), each followed by identical practice; sessions 1 week apart; order randomized/counterbalanced. • Experiment IV: Part A (within-subject, N=10): rTMS-only vs rPNS-only; no practice. Part B (within-subject, N=8): PCMS+, PCMS−, PCMS_coupled-control without practice; plus separate sessions of motor practice only vs rest (no PCMS). All sessions spaced 1 week; randomized order.
• Data processing: Trials with pre-stimulus EMG activity excluded. Outliers removed at mean±2 SD. MEPs normalized to Mmax. F-waves high-pass filtered (200 Hz), persistence computed as % waves >0.045 mV. Acceleration normalized to session baseline. EMG measures normalized to Mmax.
• Statistics: Linear mixed-effects models (lme4) with factors GROUP or PROTOCOL and TIME, random intercepts per subject; DAY included additively where relevant to control order effects. Post-hoc pairwise comparisons Holm–Sidak adjusted. Additional analyses of B1-start to B3-end changes; Pearson correlations between EMG changes and acceleration changes. Alpha p<0.05.

Experiment I (N=26): PCMS+ priming improved ballistic learning vs Rest. GROUP×TIME interaction on peak acceleration: F(3,4038)=52, p<0.001. PCMS+ superior at B2 (139.8%±1.8 vs 115.7%±1.8, p=0.01) and B3 (148.2%±1.8 vs 120.5%±1.8, p<0.01). Improvement from B1 to B3 larger with PCMS+ (22.28%±1.0) vs Rest (10.93%±1.0), p<0.001. Baseline performance similar. MEPs: GROUP×TIME F(2,1766)=10.9, p<0.001; larger increase baseline→post-stimulation with PCMS+ (103.5%±11.3) vs Rest (26.4%±12.4), p<0.01; no baseline difference.

Experiment II (N=20, double-blind): Replicated behavioral benefits. GROUP×TIME F(4,3882)=14.6, p<0.001. Baseline similar. PCMS+ better at B3 (128.6%±1.4) vs Sham (119.3%±1.4), p<0.01; B1→B3 change greater (17.4%±0.8 vs 11.8%±0.8, p<0.001). Retention at 7 days: PCMS+ 131.4%±4.4 vs Sham 119.8%±4.4, p<0.01. MEPs: GROUP×TIME F(2,1457)=5.9, p<0.01; baseline MEPs larger in PCMS+ (9%±5) vs Sham (5%±2), p=0.03; relative increase baseline→post-stimulation larger PCMS+ (58%±8) vs Sham (25%±8), p<0.01. EMG pooled (N=46): EMG amplitude and rate of rise increased with practice (main TIME); GROUP×TIME interactions indicated larger EMG increases with PCMS+. Positive correlation between acceleration change and rate of EMG rise change.

Experiment III (N=18, crossover): PROTOCOL×TIME on acceleration F(6,8482)=12.2, p<0.001. PCMS+ > PCMS_coupled-control at B2 and B3 (p<0.05). PCMS− had lower performance in B1 vs PCMS+ and PCMS_coupled-control (both p<0.01). B1→B3 improvements: PCMS+ 15.9%±0.8 > PCMS− 10.8%±0.8 (p<0.001) and > PCMS_coupled-control 8.3%±0.8 (p<0.001). MEPs: PROTOCOL×TIME F(4,4143)=10.95, p<0.001. Relative increases baseline→post-stimulation: PCMS+ +51.3%±5.9 > PCMS− +35.3%±5.9 (p=0.03) but < PCMS_coupled-control +75.1%±6.0 (p<0.01). At post-practice, PCMS+ and PCMS_coupled-control > PCMS− (p<0.01). PCMS− showed no significant within-protocol change baseline→post-practice (p=0.08), unlike PCMS+ and PCMS_coupled-control (both p<0.001).

Experiment IV: Unpaired stimulation (N=10): rTMS-only and rPNS-only did not change MEPs (TIME F(3,1331)=1.01, p=0.38). Paired PCMS without practice (N=8): All protocols increased MEPs; PROTOCOL×TIME F(8,1694)=2.05, p=0.03. PCMS+ > PCMS− (mean +85%±10.9, p<0.001) and > PCMS_coupled-control (+49.5%±11.0, p<0.001) across post time-course; PCMS− < PCMS_coupled-control (−36.3%±11.0, p<0.01). Motor practice vs rest (N=8): Practice increased MEPs vs rest (Group F(1,1056)=64.7, p<0.001).

Overall: PCMS+ primed greater and lasting improvements in ballistic motor learning, with timing- and order-dependence approximating Hebbian rules. Corticospinal excitability increases required paired stimulation; effects showed timing specificity but not clear bidirectionality. Learning gains were associated with greater FDI activation (EMG) during practice.

The study demonstrates that exogenously induced, timing-specific plasticity via PCMS interacts with subsequent practice-dependent plasticity to enhance ballistic motor learning. Facilitatory PCMS targeted to ensure presynaptic activation slightly precedes postsynaptic activation at CM synapses (−2 ms IAI) improved learning beyond rest and sham, with benefits persisting at least seven days, indicating robust behavioral relevance. The within-subject timing/order manipulations show that priming adheres to Hebbian-like rules: temporal proximity and correct order of pre-before-post at the CM synapse facilitated learning, while reversing timing (PCMS−) transiently impaired early learning. Importantly, increased corticospinal excitability alone did not guarantee enhanced learning: PCMS_coupled-control increased MEPs but did not improve learning to the level of PCMS+, highlighting network specificity—plasticity must be targeted to the circuitry underpinning the behavior. EMG analyses linked performance gains to increased muscle activation and rate of EMG rise in the targeted FDI, supporting a spinal contribution via more effective CM activation. Across experiments, paired stimulation was necessary for corticospinal excitability changes; unpaired rTMS or rPNS did not alter MEPs. While behavioral effects followed Hebbian principles, corticospinal excitability changes were timing-specific but not strictly bidirectional, likely reflecting multilevel (spinal and cortical) contributions to MEPs and differences from CMEP-based measures of spinal transmission. These findings provide a mechanistic rationale for using individualized, timing-specific PCMS to augment the effects of motor practice.

Non-invasive paired corticomotoneuronal stimulation timed to facilitate CM synapses primes and enhances subsequent ballistic motor learning in healthy adults, with effects persisting after one week. The priming effect depends on precise spike-timing and order at the targeted synapses, approximating Hebbian learning rules. Paired stimulation is necessary to modulate corticospinal excitability, and timing specificity is critical. The results suggest PCMS as a promising add-on to sensorimotor training and neurorehabilitation to strengthen practice-induced plasticity. Future work should test clinical populations, examine cumulative effects of repeated priming sessions, directly assess spinal transmission changes (e.g., CMEPs), and refine individualized timing to maximize behavioral gains.

Experiment I lacked blinding and sham control, allowing potential expectation effects early in practice (addressed in Experiment II). Baseline MEP differences occurred in some protocols/groups (e.g., Experiment II and III), complicating interpretation of absolute excitability levels; analyses focused on relative changes mitigate this. The within-subject crossover in Experiment III may be susceptible to carry-over effects despite week-long spacing. MEPs index the overall corticomotor pathway and cannot isolate spinal CM synaptic changes; CMEPs were not recorded, limiting precise localization of excitability changes. PCMS− did not reduce MEP amplitudes as previously reported with CMEPs, possibly due to cortical contributions or responder variability, creating ambiguity about bidirectionality at the spinal level. Samples comprised young healthy adults, limiting generalizability to patient populations. Session numbers were limited; longer-term, multi-session effects remain unknown.