The cross education of strength and skill following unilateral strength training in the upper and lower limbs

Abstract

Cross education is the strength gain or skill improvement transferred to the contralateral limb following unilateral training or practice. The present study examined the transfer of both strength and skill following a strength training program. Forty participants (20M, 20F) completed a 6-wk unilateral training program of dominant wrist flexion or dorsiflexion. Strength, force variability, and muscle activity were assessed pretraining, posttraining, and following 6 wk of detraining (retention). Analyses of covariance compared the experimental limb (trained or untrained) to the control (dominant or nondominant). There were no sex differences in the training response. Cross education of strength at posttraining was 6% (P < 0.01) in the untrained arm and 13% (P < 0.01) in the untrained leg. Contralateral strength continued to increase following detraining to 15% in the arm (P < 0.01) and 14% in the leg (P < 0.01). There was no difference in strength gains between upper and lower limbs (P > 0.05). Cross education of skill (force variability) demonstrated greater improvements in the untrained limbs compared with the control limbs during contractions performed without concurrent feedback. Significant increases in V-wave amplitude (P = 0.02) and central activation (P < 0.01) were highly correlated with contralateral strength gains. There was no change in agonist amplitude or motor unit firing rates in the untrained limbs (P > 0.05). The neuromuscular mechanisms mirrored the force increases at posttraining and retention supporting central drive adaptations of cross education. The continued strength increases at retention identified the presence of motor learning in cross education, as confirmed by force variability.

NEW & NOTEWORTHY: We examined cross education of strength and skill following 6 wk of unilateral training and 6 wk of detraining. A novel finding was the continued increase in contralateral strength following both training and detraining. Neuromuscular adaptations were highly correlated with strength gains in the trained and contralateral limbs. Motor learning was evident in the trained and contralateral limbs during contractions performed without concurrent feedback.

INTRODUCTION

The cross education of strength and skill learning was first discovered in 1894 by Scripture et al., who determined that muscular strength and task steadiness could be improved in the contralateral limb following unilateral training. Cross education of strength refers to the strength gain that is transferred to the contralateral limb following a unilateral training program in the ipsilateral limb. The mechanisms behind the cross education of strength include cortical and spinal adaptations, which alter the neural drive to the contralateral, untrained limb (Lee et al. 2010; Ruddy and Carson 2013). The same is true for the cross education of skill following unilateral motor task practice, which is typically examined following one to two sessions of unilateral motor task practice.

Leung and colleagues (2015) found no difference between transcranial magnetic stimulation (TMS) responses to a single session of unilateral metronome-paced elbow flexion (strength training) and visuomotor tracking of the upper limb (skill practice), indicating the similarity of mechanisms contributing to the cross education of strength and skill. However, most studies in this field examine the transfer of strength and skill separately following unique training paradigms (Keogh et al. 2010; Ruddy and Carson 2013). Since all resistive exercise (i.e., strength training) includes a skill component, and therefore the potential for motor learning (Green et al. 2014; McGuire et al. 2014a, 2014b), it is likely that a strength-training program results in the cross education of both strength and skill. Therefore, the present study investigated the presence of skill transfer (cross education of skill) following unilateral strength training.

Early meta-analyses of cross education estimated that the contralateral strength gain was ~7.8% (Carroll et al. 2006; Munn et al. 2004). However, these analyses were limited to few studies and have been expanded by Manca and colleagues (2017) resulting in an estimated contralateral strength gain of 11.9%. Furthermore, a comprehensive review of the literature including 96 studies, undertaken by the authors, estimates the contralateral strength gain to be an estimated 18% (Green and Gabriel In Press). The potential of cross education for clinical purposes is evident; however, the improvement of motor skills is an important clinical aspect that has been widely overlooked following strength training. Keogh and colleagues (2010) examined motor output variability after unilateral task-specific coordination training (joint angle trajectory matching) vs. general strength training (unilateral bicep curls and wrist flexion training). Although only the coordination training group had improved variability in the contralateral limb, both groups improved bilateral coactivation, indicating that the “task to assess variability” (postural tremor during pointing) may have been too dissimilar from the strength training movements.

While a change in coactivation is a neural adaptation associated with coordination, changes in task variability demonstrate a motor learning dimension that represent adaptation of the internal model associated with the task (Kawato 1999; Slifkin and Newell 1999). Previous cross-education studies examining coactivation in the contralateral limb have demonstrated improvements in coactivation (Carolan and Cafarelli 1992; Farthing et al. 2007; Fimland et al. 2009; Hortobágyi et al. 1997; Lagerquist et al. 2006). However, to our knowledge the variability of a voluntary contraction has yet to be examined in the contralateral limb following unilateral strength training. In addition, no unilateral training study has examined the changes in coactivation following a period of detraining to determine the persistence of motor learning.

There are discrepant findings in cross-education literature with respect to changes in neural drive to the homologous muscle. Despite the evidence that unilateral training increases contralateral neural drive and corticospinal excitability (Frazer et al. 2017; Hendy and Lamon 2017; Kidgell et al. 2011; Ruddy and Carson 2013; Ruddy et al. 2017), previous studies examining the cross education of strength have found equivocal results in measures of central drive. The amplitude of contralateral muscle activity following unilateral training has been found to increase ~28% (range from 3 to 59%) in the majority of studies; however, eight studies have found no significant change in agonist activity (Boyes et al. 2017; Colomer-Poveda et al. 2017; Dragert and Zehr 2013; Farthing et al. 2005, 2007; Fimland et al. 2009; Hortobágyi et al. 1997; Kidgell et al. 2011; Moritani and DeVries 1979; Narici et al. 1989; Rich and Cafarelli 2000; Shima et al. 2002; Tillin et al. 2012).

Studies assessing the “completeness of activation” as assessed by twitch interpolation have shown a significant increase (Lee et al. 2009; Shima et al. 2002), or no change (Tillin et al. 2011). Similarly, V-wave amplitude has been examined in two cross-education studies with Colomer-Poveda et al. (2017) demonstrating no change and Fimland and colleagues (2009) demonstrating a nonsignificant 29% increase. In the latter case, the authors concluded that the high variability of the measure was the reason for the nonsignificant finding. Lastly, an increase in motor unit firing rates (MUFRs) would be expected to accompany an increase in neural drive to the muscle (Kent-Braun and Le Blanc 1996; Knight and Kamen 2008). However, in two studies examining unilateral training, there was no increase in MUFRs in the contralateral limb at posttraining (Patten et al. 2001; Rich and Cafarelli 2000). This result is surprising and deserves additional examination.

Thus there are a number of unresolved issues that motivate a reexamination of cross education to assess both strength and skill following a unilateral strength training program.

Therefore, 20 men and 20 women completed a 6-wk unilateral training program in the upper and lower limbs, with assessment at pre- and posttraining, as well as following a 6-wk detraining period to evaluate retention. It was hypothesized that unilateral strength training would increase contralateral strength and central drive, as demonstrated with increased central activation and V-wave amplitude. Furthermore, it was hypothesized that strength training would result in the cross education of skill, as demonstrated by decreased force variability in the contralateral limb.

MATERIALS AND METHODS

Experimental approach to the problem.

The measurement scheduled was designed to explore both the strength and motor learning aspects of strength training. There were two pretraining sessions (familiarization and baseline), separated by a minimum of 48 h and a maximum of 7 days. The first pretest session was used to subtract initial gains in strength due solely to task familiarity, so that all comparison started from baseline. Training adaptations and motor learning were assessed immediately following 6 wk of unilateral strength training and with a retention test after 6 wk of detraining. We have previously shown that decreases in antagonist cocontraction and force variability for maximal effort contractions decreased following practice and were retained after 2 wk and 3 mo of detraining (McGuire et al. 2014a).

All four limbs of each participant were tested at each session, regardless of group assignment. Unilateral training effects of strength training were evaluated in the following way. The dominant arm of the leg-training group, served as the control for the dominant (trained) arm of the arm-training group (see Fig. 1A, comparison 1). Conversely, the dominant leg of the arm-training group was the control for the dominant (trained) leg of the leg-training group (see Fig. 1A, comparison 2). Cross education was evaluated by comparing the nondominant limbs. The nondominant arm from leg-training group served as the control for the nondominant (untrained) arm of the arm-training group (see Fig. 1B, comparison 3). Similarly, the nondominant leg from the arm-training group served as the control for the nondominant (untrained) leg of the leg-training group (see Fig. 1B, comparison 4). Surface EMG (sEMG) was monitored in both dominant and nondominant limb while testing regardless of the limb performing the movements to ensure that contralateral adaptations were due to cross education rather than force irradiation (i.e., postural stability during testing). Instructions for performing maximal voluntary contractions (MVCs) and ramp contractions were kept consistent across participants and sessions to minimize experimenter influence.

Fig. 1.

Statistical comparisons for the training effects of the trained arm (1) and leg (2) and cross-education effects of the untrained arm (3) and leg (4) compared with the relevant dominant and nondominant controls. The experimental (A: ipsilateral; B: contralateral) limbs of the training group are noted in black.

Participants.

Forty participants (20 men, 20 women) reviewed and signed informed consent documents as approved by the Brock University Research Ethics Board and conducted according to the principles expressed in the Declaration of Helsinki. All participants were young adults (24 ± 3 yr), recreationally, moderately active and free of self-reported neurological or orthopedic abnormalities as ruled out by the Physical Activity Readiness Questionnaire (PAR-Q+) from the Canadian Society for Exercise Physiology. Handedness and footedness were determined using an adapted version of the Lateral Preference Inventory (Coren 1993). Participants were randomly assigned to either the arm-training group (ATG; wrist flexion training) or the leg-training group (LTG; dorsiflexion training) with an equal number of men and women within each group. The dorsiflexors and wrist flexors were selected due to their involvement in activities of daily living and gait and the importance of cross education to poststroke rehabilitation. The strength of the dorsiflexor muscles were selected due to their importance in gait and posture and the common occurrence of poststroke weakness (i.e., drop foot) (Patten et al. 2004; Zehr and Loadman 2012; Dragert and Zehr 2013).

Experimental setup.

All testing took place inside a Faraday cage in the Electromyographic Kinesiology Laboratory at Brock University. Participants were seated in a jig to isolate the intended muscles and minimize extraneous movement. A computer screen was placed in front of the participant to display their force. For dorsiflexion contractions participants were seated with the hip, knee, and ankle at 90° angles. Isometric contractions were performed by contracting against a metal bar, which was lowered over the top of the foot (dorsiflexion), and pressing down upon the footplate (plantar flexion). For wrist flexion contractions, participants were seated at a table with their arm resting on the surface at 160° of elbow extension. The forearm was in a neutral (half-supinated) position, and the hand was fixed between metal bars to contract isometrically against the palmar (wrist flexion) or dorsal (wrist extension) bar.

Surface electromyography.

The tibialis anterior (TA), soleus, flexor carpi radialis (FCR), and extensor carpi radialis on both the right and left sides were prepared for sEMG. The skin was shaved, cleansed with isopropyl alcohol, and lightly abraded. The motor point of each muscle was found using low-level electrical stimulation passed over the surface of the skin at a rate of 1.5 pulses/s. The motor point was determined as the point at which a barely visible contraction was elicited with the least amount of stimulation. Once found, the motor point was marked with indelible ink. Pediatric sized electrodes (5-mm electrode diameter, F-E9 20 mm; GRASS Technologies, Astro-Med, Warwick, RI) were placed in a bipolar configuration with one electrode directly over the motor point and the second electrode placed immediately adjacent, in line with the muscle fibers, for an interelectrode distance of 1 cm. This electrode configuration and placement has been previously demonstrated to minimize cross talk during voluntary contractions while maintaining the compound muscle action potential shape during evoked contractions, with a high intersession reliability [intraclass correlation (ICC) = 0.85] (Green et al. 2015). By electrically identifying the motor point, we could maintain consistent electrode placement across sessions. Ground electrodes were placed on the medial malleolus and patella for the lower leg recordings and on the back of the hand and olecranon process for the forearm recordings. sEMG was amplified 500–2,000 times (Grass P511; Astro-Med) to maximize the resolution of the 16-bit analogue-to-digital converter (DI-720; DATAQ Instruments, Akron, OH). The signals were band-pass filtered (3–1,000 Hz) before digitization at 2,000 Hz (WinDaq Acquisition; DATAQ Instruments). Skin-electrode impedance was recorded pre- and posttesting and confirmed to be <10 kΩ before testing began. Skin temperature was and recorded pre- and posttesting for each limb.

Surface decomposition EMG was collected using a five-pin sensor (dEMG System; Delsys, Boston, MA) with an interelectrode distance of 5 mm, which results in four differential channels of data. The dEMG sensor was secured to the skin over the FCR and TA muscles directly adjacent to the sEMG electrodes and affixed with tape and a strap around the limb to maintain slight pressure on the skin. An additional ground electrode was placed on the olecranon process (forearm testing) or the patella (lower leg testing). The dEMG signals were amplified 1,000 times and filtered between 20 and 450 Hz to maximize the resolution of the 16-bit analogue-to-digital converter before digitization at 20,000 Hz (Bagnoli-16 and EMGworks 4.2; Delsys). The force signals from the load cells (MB-100 and SSMH; Interface, Scottsdale, AZ) were sampled without amplification or filtering through the Bagnoli-16 A/D board at 20,000 Hz as well as the DI-720 A/D board at 2,000 Hz.

Testing protocol.

The order of testing (upper vs. lower, dominant vs. nondominant) was balanced across participants and groups and kept consistent across sessions for each participant. Testing for each limb was performed as follows (see Fig. 2 for protocol).

Fig. 2.

Testing protocol including evoked potentials, maximal voluntary contractions (MVCs) of the wrist flexors and dorsiflexors with an interpolated twitch, MVCs of the wrist extensors and plantar flexors, and 60% trapezoidal contractions of the wrist flexors and dorsiflexors.

To begin, a series of three maximal M waves were elicited from the median (for FCR stimulation) or common fibular nerve (for TA stimulation) while the participant was at rest with a 1-ms square-wave pulse using a handheld stimulation probe (Grass S88 Stimulator and SIU8T; Astro-Med). Participants then performed three isometric MVCs of the agonist muscles (wrist flexion and dorsiflexion) each lasting 4 s in duration with 2-min intertrial rest periods. Participants were instructed to contract “hard-and-fast” and hold their maximum steady, which was assisted by a force trace presented to participants on an oscilloscope. An interpolated twitch was evoked in the middle of each contraction at a supramaximal stimulation level corresponding to ~110% of the M-wave stimulation level.

Participants performed three isometric MVCs in the “opposite” direction (wrist extension and plantar flexion contractions) lasting 4 s in duration with 2-min intertrial rest periods. Participants then performed a 6-s ramp contraction (wrist flexion or dorsiflexion) to 20% MVC force, which was completed as a signal quality check for the dEMG electrodes. Three isometric ramp contractions to 60% MVC force were performed with 2-min intertrial rest periods. The ramps increased at 10% MVC per second for a rise of 6 s, held a steady plateau at 60% MVC for 6 s, and decreased force at 10% MVC per second for a drop of 6 s. A computer screen placed in front of the participants displayed a “target” trapezoidal ramp contraction, which participants were asked to follow with their force trace resulting in concurrent knowledge of results.

During the retention test only (12-wk test), the ramp contractions were performed without feedback to serve as a retention test. Three blinded ramps (screen turned off) were performed immediately after the three contractions with feedback. Participants were reminded of their goal to increase force to what they believed to be 60% of maximal strength over 6 s, hold it steady for 6 s, and release consistently for 6 s. The timing was counted out for the participants as follows “ramp up-5-4-3-2-1, hold steady-5-4-3-2-1, ramp down-5-4-3-2-1.” No feedback from the investigator was given between trials or between limbs.

Training.

Training was performed over a period of 6 wk outside the laboratory and included dynamic contractions performed using a pulley cable (for dorsiflexion) or a dumbbell (for wrist flexion) set at 80% of the participants MVC force. Participants trained 4 times per week, completing 3 sets of 10–12 reps/session. At week 3, participants returned to the laboratory to complete three sets of three MVCs on the trained limb to increase the prescribed weight to reflect the participant’s current strength (new 80% MVC weight). To minimize mirror contractions, participants were instructed to focus on the dominant limb while resting the nondominant limb with their knee flexed and foot resting on the floor (for dorsiflexion contractions) or with their elbow bent and their hand resting in their lap (for wrist flexion contractions). Participants were instructed not to specifically train the contralateral (nondominant) side or the alternate muscle group (i.e., dorsiflexors if in ATG and wrist flexors if in LTG) or begin any new training regimen over the course of their study involvement. During detraining, participants were instructed to resume normal activity, with the exception of specifically training the wrist flexors or dorsiflexors.

Data reduction.

Force and root-mean-square (RMS) amplitude of the sEMG signal were calculated from a 1-s window at the center of the MVC, terminating before the interpolated twitch where applicable. Agonist RMS sEMG was then normalized to the peak-to-peak amplitude of the corresponding M-waves evoked at rest (RMS/Mmax). sEMG activity of the limb at rest was obtained from the same data window to monitor postural stabilizing activity. The amount of cocontraction during ramp contractions was calculated as the antagonist activity during normalized to its RMS amplitude when the muscle was maximally contracting as an agonist (Carolan and Cafarelli 1992). The amount of cross talk between muscles was measured from the M-wave peak-to-peak amplitude present in the antagonist and by using the cross correlation between agonist and antagonist sEMG during an MVC.

Wrist flexion and dorsiflexion MVCs included twitch interpolation for the purpose of identifying V-waves and calculating central activation. V-waves were identified from the sEMG signal as a distinct wave occurring approximately 20–50 ms following the M-wave, as evoked by the twitch interpolation. If no V-wave could be identified on a minimum of two of the three contractions, that limb was removed from the V-wave analysis. The V-wave was expressed as a ratio to the immediately preceding Mmax (V/M ratio) and averaged for the identified contractions. The central activation ratio (CAR) was calculated as the force increase obtained by the interpolated twitch compared with the maximal voluntary force using the formula [CAR = 1 – (interpolated twitch force/maximal voluntary force) × 100)] (Dowling et al. 1994; Kent-Braun and Le Blanc 1996), where maximal voluntary force was the mean force taken from 1 s immediately before the interpolated twitch, and interpolated twitch force was force increase resulting from the stimulation.

Motor units from the 60% ramp contractions were decomposed using the Precision Decomposition Algorithm III in the dEMG Analysis software (version 1.1; Delsys), which identifies each motor unit and the firing instances (Chang et al. 2008; De Luca et al. 2006; Nawab et al. 2010). The decomposition results were then validated using a built-in validation software employing Decompose-Synthesize-Decompose-Compare technique (Nawab et al. 2010; De Luca and Contessa 2012). The MUFRs were calculated as the inverse of the interpulse interval and smoothed using a Hanning window of 0.95 s. The firing rate for each motor unit was then calculated as the mean firing rate occurring in a 1-s window at the center of the plateau portion (60% maximal force) of the ramp contraction. The coefficient of variation (σ/μ × 100) was calculated for each MUFR. Motor unit firing rates were included in the analysis if they met the following criteria: 1) the motor unit firing instances met a minimum decomposition accuracy of 90%; 2) the trial had a minimum of five motor units with >90% decomposition accuracy; and 3) the motor unit firing rate had less than a 20% coefficient of variation.

The variance ratio was calculated for each set of three ramp contractions, which determines the reproducibility of the forces traces, on a point-by-point basis (Kadaba et al. 1989). By interpolating the force traces to the identical number of points, the variance ratio measures how consistently the shape of each ramp is performed (Green et al. 2014). This measures the variability of the motor pattern between trials, with a lower score indicating less variability, thus a comparison of identical trials would produce a variability ratio score of zero. The root-mean-square-error (RMSE) was calculated as the standard deviation during the plateau portion of each contraction. It was normalized to the average force level to account for strength differences, effectively resulting in the coefficient of variation of the ramp plateau.

Statistics.

All statistical analyses were performed in SAS (SAS 9.4; SAS Institute, Cary, NC). The data were first screened to determine if they met the statistical assumptions before performing analysis of covariance (ANCOVA) procedures. When normality was violated, appropriate transformations were performed. Preliminary analyses were also performed to determine if there were any significance differences between men and women. There were no significant sex × session interaction terms, so the data were collapsed for further analyses. Significant differences were evaluated using a two-way repeated-measures ANCOVA with one between factors (experimental vs. control) and one within factor (baseline, posttraining, retention). Baseline scores were used as a covariate to subtract out initial discrepancies in strength. Furthermore, for measures calculated during the 60% ramp contractions, there were four sessions included in the analysis to examine the ramp contractions performed with “no feedback” during the retention: baseline, posttraining, retention (feedback), retention (no feedback).

To assess the training adaptations the following ANCOVAs were performed: 1) trained arm of the ATG vs. dominant control arm of the LTG; and 2) trained leg of the LTG vs. dominant control leg of the ATG. Similarly, to assess cross education in the untrained limb the following ANCOVAs were performed: 1) untrained arm of the ATG vs. nondominant control arm of the LTG; and 2) untrained leg of the LTG vs. nondominant leg of the ATG. Contrasts were performed using least squares means with a Bonferroni adjustment to compare the difference between sessions (baseline to posttraining, and baseline to retention), as well as the magnitude of training adaptations between the experimental limbs (trained vs. untrained) at each session. Effect sizes were calculated according to Cohen (1988) for each limb separately across time points (baseline to posttraining, baseline to retention, and baseline to no feedback where applicable). As suggested by Cohen (1988), d = 0.2 is considered a small effect, 0.5 a medium effect, and 0.8 a large effect.

For measures calculated during the no feedback contractions (force variability and antagonist cocontraction), a separate ANOVA was performed. Since it was expected that the no feedback contractions would result in large changes from baseline for all groups, the ANOVA was run to compare the percent change score from baseline to no feedback between groups (experimental vs. control).

To determine the correlation between neuromuscular adaptations and strength improvements due to training and cross education, correlations were calculated using the repeated-measures ANOVA technique (Bakdash and Marusich 2017) for force of the trained or untrained limb and the neuromuscular measures (RMS/M, MUFRs, V/M ratio, and CAR).

The contralateral transfer of strength in the experimental limbs (E: trained or untrained) is presented as percent strength gain accounting for the control limbs (C: dominant or nondominant) according to the following formula (Carroll et al. 2006):

$$\frac{EPost-EBaseline}{EBaseline}-\frac{CPost-CBaseline}{CBaseline}$$

The strength gain magnitude was compared between upper and lower limbs and between trained and untrained limbs using change scores:

$$Post\hspace{0.17em}Gain=\frac{Post-Baseline}{Baseline}Retention\hspace{0.17em}Gain=\frac{Retention-Baseline}{Baseline}$$

The strength gain in the upper and lower limbs was compared using an ANOVA, with post hoc testing (least square means) to examine the limbs at posttraining and at retention. The strength gain in the trained and untrained limbs was examined using a regression (X: trained by Y: untrained) separated by upper and lower limbs and by posttraining and retention testing.

RESULTS

Methodological controls.

Participant demographics are detailed in Table 1. Resting limb activity was <3% of maximal activity recorded during MVCs. Cross talk, as measured from the M-waves, was ~9% of the agonist activity being recorded in the antagonist signal. Cross talk, as examined in the flexion MVCs using the cross-correlation function, was <1%. Skin-electrode impedance and skin temperature stayed consistent within and between sessions.

Table 1.

Participant demographics, baseline strength, and training compliance

	Arm Training	Leg Training
Age, yr	25 ± 3	23 ± 2
Height, cm	174.3 ± 10.1	172.6 ± 10.7
Weight. kg	72.1 ± 9.3	69.3 ± 10.2
Strength,* N
Dominant arm	112.0 ± 30.2	123.4 ± 40.4†
Nondominant arm	115.9 ± 35.2	121.3 ± 33.0†
Dominant leg	280.3 ± 97.7	243.1 ± 74.2
Nondominant leg	277.6 ± 102.4	236.2 ± 91.3
Compliance, %	96 ± 6	93 ± 7

Values are means ± SD; n = 20.

No significant differences were found between groups (P > 0.05).

^*Strength is average force from baseline maximal voluntary contractions.

^†Arm strength for the leg-training group is calculated from a subset of participants (n = 12) due to a slight change in equipment configuration.

Across limbs the average change in wrist flexion and dorsiflexion strength from familiarization to baseline testing was a 6% increase, while the agonist activity increased by 3%. The wrist extension and plantar flexion MVCs increased by 10% from familiarization to baseline. The V/M ratio increased 8%, but central activation increased less than 1% due to the high initial levels of central activation achieved in the familiarization session (97 ± 3.5%). The initial motor learning was evident in the large decreases in variance ratio (up to 100%) and RMSE (up to 41%).

The following results are presented for the trained dominant limb (“trained”), the contralateral nondominant limb (“untrained”), the dominant limb of the alternate group (“dominant control”), and the nondominant limb of the alternate group (“nondominant control”). The F-ratios of each ANCOVA are presented in text for the main effect of sessions (baseline, posttraining, retention) for each group. To further examine significant main effects, contrasts examining baseline to posttraining and baseline to retention were performed and are presented in the tables and figures. All values presented are calculated from the least squares means. Percent change scores and effect sizes are presented in Table 2 (upper limb) and Table 3 (lower limb).

Table 2.

Change scores calculated from least squares means of the upper limb from baseline to posttraining and baseline to retention testing

	Trained Arm		Dominant Control Arm		Untrained Arm		Nondominant Control Arm
Variable	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %
Strength, N	27.6 (27%)*	19.7 (19%)*	3.5 (3%)	3.3 (3%)	11.2 (11%)*	18.5 (18%)*	5.3 (5%)	3.1 (3%)
	d = 0.84	d = 0.67	d = 0.07	d = 0.06	d = 0.32	d = 0.45	d = 0.10	d = 0.07
Agonist EMG (RMS/Mmax)	0.017 (23%)	0.011 (15%)	−0.002 (−3%)	0.000 (0%)	0.005 (6%)	0.009 (13%)	−0.004 (−6%)	0.007 (9%)
	d = 0.68	d = 0.39	d = −0.02	d = −0.01	d = 0.16	d = 0.34	d = −0.04	d = 0.14
Central activation,%	0.90 (0.9%)*	0.95 (1.0%)*	0.21 (0.2%)	0.03 (0.0%)	1.12 (1.2%)*	1.65 (1.7%)*	0.54 (0.6%)	0.16 (0.2%)
	d = 0.64	d = 0.65	d = 0.10	d = 0.02	d = 0.51	d = 0.70	d = 0.17	d = 0.06
MUFRs, pps	0.05 (0%)	−0.11 (− 1%)	0.78 (5%)	0.11 (1%)	−0.14 (−1%)	−0.51 (−3%)	−0.72 (−4%)	−0.66 (−4%)
	d = 0.03	d = −0.06	d = 0.44	d = 0.06	d = −0.06	d = −0.22	d = −0.35	d = −0.29
Variance ratio	−0.002 (−12%)	0.001 (6%)	−0.003 (−17%)	−0.006 (−29%)	−0.006 (−25%)	−0.008 (−36%)	0.001 (4%)	−0.005 (−21%)
	d = −0.30	d = 0.10	d = −0.30	d = −0.37	d = −0.09	d = −0.19	d = 0.06	d = −0.01
ΔNo Feedback	0.062 (385%) d = 2.52		0.040 (198%) d = 1.72		0.040 (170%) d = 1.99		0.044 (179%) d = 2.37
RMSE, %	−0.20 (−7%)	0.03 (1%)	0.18 (6%)	−0.11 (−4%)	−0.43 (−13%)	−0.29 (−9%)	0.43 (16%)	0.35 (13%)
	d = −0.15	d = 0.03	d = 0.23	d = −0.02	d = −0.28	d = −0.19	d = 0.32	d = 0.25
ΔNo Feedback	1.59 (55%) d = 1.27		2.00 (71%) d = 1.63		0.96 (29%)† d = 0.66		2.64 (97%)† d = 2.14

Cohen’s d is given for effect size of the change from baseline; n = 20. Strength, electromyography (EMG) amplitude, and central activation were calculated from maximal voluntary contractions. Motor unit firing rates (MUFRs), variance ratio, and root-mean-square error of the force trace (RMSE) were calculated from the ramp contractions to 60% maximal force. Variability is also reported for the “no feedback” ramps performed during the retention testing. ΔPost, change from baseline to posttraining; ΔRetention, change from baseline to retention (detraining); ΔNo feedback, change from baseline to retention (ramps performed without feedback); RMS/M ratio: the root-mean-square amplitude normalized to the M-wave peak-to-peak amplitude.

^*Significant change from baseline (P < 0.05).

^†Significant difference between experimental and control limb (for no feedback only).

Table 3.

Change scores calculated from least squares means of the lower limb from baseline to posttraining and baseline to retention testing

	Trained Leg		Dominant Control Leg		Untrained Leg		Nondominant Control Leg
Variable	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %	ΔPost-raw, %	ΔRetention raw, %
Strength, N	52.4 (20%)*	58.4 (22%)*	7.6 (3%)	15.0 (6%)	37.6 (15%)*	55.0 (22%)*	5.7 (2%)	19.3 (7%)
	d = 0.69	d = 0.70	d = 0.08	d = 0.15	d = 0.40	d = 0.58	d = 0.06	d = 0.19
Agonist EMG (RMS/Mmax)	0.026 (33%)*	0.032 (40%)*	0.007 (9%)	0.009 (12%)	0.011 (15%)	0.012 (15%)*	0.003 (4%)	0.009 (11%)
	d = 0.78	d = 0.89	d = 0.30	d = 0.31	d = 0.43	d = 0.46	d = 0.15	d = 0.29
V/M ratio‡	0.18 (59%)*	0.13 (44%)*	0.09 (29%)*	0.07 (22%)	0.14 (41%)	0.11 (34%)*	0.06 (16%)	0.001 (0%)
	d = 1.11	d = 1.08	d = 0.47	d = 0.33	d = 0.65	d = 0.57	d = 0.24	d = 0.02
MUFRs, pps	−0.65 (−4%)	−0.05 (0%)	0.67 (5%)	0.68 (5%)	0.67 (5%)	0.09 (1%)	0.46 (3%)	0.04 (0%)
	d = −0.39	d = −0.02	d = 0.32	d = 0.30	d = 0.36	d = 0.05	d = 0.20	d = 0.02
Variance ratio	−0.007 (−21%)	−0.009 (−28%)*	−0.006 (−22%)	−0.006 (−20%)	−0.010 (−30%)*	−0.003 (−9%)	0.001 (2%)	0.006 (21%)
	d = −0.25	d = −0.61	d = −0.13	d = −0.31	d = −0.81	d = −0.47	d = 0.05	d = 0.34
ΔNo feedback	0.028 (83%) d = 1.11		0.042 (151%) d = 1.50		0.014 (42%)† d = 0.81		0.041 (136%)† d = 1.68
RMSE, N	−0.19 (−5%)	−0.35 (−10%)	−0.37 (−12%)	−0.32 (−10%)	−0.33 (−9%)	−0.28 (−8%)	−0.46 (−13%)	−0.36 (−10%)
	d = −0.01	d = −0.11	d = −0.28	d = −0.28	d = −0.16	d = −0.12	d = −0.28	d = −0.19
ΔNo feedback	0.76 (22%) d = 0.59		0.68 (22%) d = 0.68		0.62 (18%) d = 0.45		0.87 (25%) d = 0.51

Cohen’s d is given for effect size of the change from baseline; n = 20. Strength, electromyography (EMG) amplitude, and V-wave amplitude normalized to M-wave amplitude (V/M ratio) were calculated from maximal voluntary contractions. Motor unit firing rates (MUFRs, variance ratio, and root-mean-square error of the force trace (RMSE) were calculated from the ramp contractions to 60% maximal force. Variability is also reported for the “no feedback” ramps performed during the retention testing. ΔPost, change from baseline to posttraining; ΔRetention, change from baseline to retention (detraining); ΔNo feedback, change from baseline to retention (ramps performed without feedback); RMS/M ratio: the root-mean-square amplitude normalized to the M-wave peak-to-peak amplitude.

^*Significant change from baseline (P < 0.05).

^†Significant difference between experimental and control limb (for no feedback only).

^‡V/M ratio: n = 16 for trained and untrained; n = 19 for control legs.

Strength.

Maximal strength significantly increased in the trained arm [F_(2,75) = 25.27, P < 0.01], trained leg [F_(2,76) = 12.45, P < 0.01], untrained arm [F_(2,76) = 10.54, P < 0.01], and untrained leg [F_(2,76) = 8.91, P < 0.01]. The training stimulus resulted in strength increases of 24% (d = 0.84) at posttraining and 16% (d = 0.67) at retention in the trained arm and 17% (d = 0.69) at posttraining and 17% (d = 0.70) at retention in the trained leg. Cross education resulted in strength increases of 6% (d = 0.32) at posttraining and 15% (d = 0.45) at retention in the untrained arm and 13% (d = 0.40) at posttraining and 14% (d = 0.58) at retention in the untrained leg as demonstrated in Fig. 3. There was no increase in any of the control limbs (P > 0.05).

Fig. 3.

Least squares means (error bars: SE) of maximal voluntary contraction force at baseline, posttraining, and retention. *Significance between experimental and control limb (P < 0.05); n = 20.

Strength was compared between the trained and untrained limbs using contrasts at each session. At baseline there was no difference between trained and untrained strength in the arms (P = 0.95) or legs (P = 0.98). At posttraining the trained and untrained arms were significantly different (P < 0.01) demonstrating that the trained arm increased strength significantly more than the untrained arm. At posttraining the trained and untrained legs were not significantly different (P = 0.35), indicating that the magnitude of cross education was equal to the magnitude of training. At retention testing, the trained and untrained arm strength had equalized (P = 0.84), as did the trained and untrained leg strength (P = 0.79).

The gain scores between upper and lower limb strength (as reported above) were compared for the trained and untrained limbs. There was no significant difference in the limb (upper vs. lower) × session (postgain and retention gain) interaction for the trained [F_(1,38) = 1.72, P = 0.20] or the untrained [F_(1,38) = 0.41, P = 0.53] limbs.

The strength gain in the trained (X) vs. untrained (Y) limb are presented in Fig. 4. The regression analysis demonstrated a significant correlation between trained and untrained strength for the lower limb at posttraining (R = 0.73, P < 0.01) and retention (R = 0.89, P < 0.01). However, the correlation was not significant in the upper limb at posttraining (R = 0.17, P = 0.47) or retention (R = 0.29, P = 0.21).

Fig. 4.

Correlation between the percenter strength gained in the trained (X) and untrained (Y) limbs for the upper (gray) and lower (black) limbs. A: posttraining; B: retention. **Significance of the correlation model (P < 0.01).

Extensor muscle strength (wrist extension and plantar flexion) was measured to assess the specificity cross education. Wrist extension strength increased for the trained arm [F_(2,76) = 4.63, P = 0.01] and untrained arm [F_(2,76) = 7.88, P < 0.01] but also increased for the dominant control arm [F_(2,76) = 3.26, P = 0.04] and the nondominant control arm [F_(2,76) = 5.43, P < 0.01]. Plantar flexion strength also increased for the trained leg [F_(2,75) = 14.88, P < 0.01] and untrained leg [F_(2,76) = 6.02, P < 0.01] but also increased for the dominant control leg [F_(2,75) = 4.95, P < 0.01] and nondominant control leg [F_(2,76) = 4.47, P = 0.01].

Agonist EMG amplitude and MURFs.

The agonist RMS/Mmax ratio, demonstrated in Fig. 5, significantly increased in the trained leg [F_(2,76) = 16.19, P < 0.01], with an increase at posttraining (33%, d = 0.78) and retention (40%, d = 0.89). There was no significant session main effect in the trained arm [F_(2,76) = 1.77, P = 0.20] despite increases of 23% (d = 0.68) at posttraining and 15% (d = 0.39) at retention testing. The session main effect was not significant for the untrained arm [F_(2,76) = 0.81, P = 0.45) with small increases at posttraining (6%, P = 0.55, d = 0.16) and retention (13%, P = 0.21, d = 0.34]. Although not significant, the effect sizes in the untrained arm are larger than in the nondominant control arm (d = −0.04–0.14). The session main effect approached significance in the untrained leg [F_(2,76) = 2.74, P = 0.07], with small increases in TA activity at posttraining (15%, P = 0.051, d = 0.43) and significantly at retention (15%, P = 0.04, d = 0.46). There was no increase in any of the control limbs (P > 0.05).

Fig. 5.

Least squares means (error bars: SE) of the agonist root-mean-square activity of the flexor carpi radialis (upper; A) and tibialis anterior (lower; B) calculated from the maximal voluntary contractions at baseline, posttraining, and retention. *Significance between experimental and control limb (P < 0.05); n = 20.

There was no significant change in the MUFRs of the trained arm [F_(2,76) = 0.06, P = 0.94], trained leg [F_(2,76) = 1.03, P = 0.36], untrained arm [F_(2,76) = 0.46, P = 0.64], or untrained leg [F_(2,76) = 0.97, P = 0.38]. The average MUFR was 16.30 (SD 2.04) in the FCR and 15.09 (SD 2.06) in the TA, with an average change over sessions of 3.6%.

Evoked contractions.

Interclass correlation analysis revealed that the V/M ratio was unreliable for the upper limb (ICC < 0.34) and therefore was only calculated for the lower limb (ICC > 0.80). V-waves could be identified on 16 participants in the trained and untrained limbs and 19 participants in the dominant and nondominant control limbs. There was a significant increase in V/M ratio in the trained leg [F_(2,66) = 8.70, P < 0.01] and the untrained leg [F_(2,66) = 3.97, P = 0.02] as illustrated in Fig. 6. At posttraining the V/M ratio increased 59% (d = 1.11) in the trained leg and 41% (d = 0.65) in the untrained leg. This increase was primarily retained following detraining in the trained (44%, d = 1.08) and untrained legs (34%, d = 0.57). Neither of the control legs had a significant session main effect (P > 0.05); however, the V/M ratio in the dominant control leg was significantly higher at posttraining than baseline [F_(1,66) = 4.29, P = 0.04]. Although significant, the V/M ratio increase in the dominant control limb (29%, d = 0.47) was smaller than that of the trained or untrained limb and returned to baseline values by retention testing.

Fig. 6.

Least squares means (error bars: SE) of the V-wave to M-wave (V/M) ratio at baseline, posttraining, and retention. *Significance between experimental and control limb (P < 0.05); for the trained and untrained limbs, n = 16; for the control limbs, n = 19.

Central activation in the lower limb was, on average, 99.7% at baseline. Therefore, the room for improvement was extremely limited (< 0.3%) so this variable was excluded. The upper limb demonstrated a slight dominance effect at baseline, with an average CAR of 97% in the dominant arm compared with 95% in the nondominant arm as shown in Fig. 7. There was a significant increase in the trained arm [F_(2,75) = 5.22, P < 0.01] from 97.5% CAR at baseline to 98.5% CAR at posttraining and retention and in the untrained arm [F_(2,75) = 4.92, P < 0.01] from 95.0% CAR at baseline to 96.1 and 96.6% at posttraining and retention, respectively. There was no increase in any of the control limbs (P > 0.05) with an average change over sessions of 0.15%.

Fig. 7.

Least squares means (error bars: SE) of the central activation ratio calculated from the interpolated twitch during maximal voluntary contractions at baseline, posttraining, and retention. *Significance between experimental and control limb (P < 0.05); n = 20.

Correlations.

To determine the relationship between force gains and neuromuscular adaptations in the trained and untrained limbs, repeated-measures correlations (r_rm) were calculated from a repeated measures ANOVA. The correlation between force and RMS amplitude normalized to M-wave amplitude was poor in the trained (r_rm = 0.17, P = 0.28) and untrained (r_rm = 0.13, P = 0.41) arms. The relationship was slightly better in the lower limb with a significant correlation in the trained (r_rm = 0.60, P < 0.01) and untrained (r_rm = 0.32, P = 0.04) legs. The V/M ratio was significantly correlated with force gains in the trained (r_rm = 0.66, P < 0.01) and untrained (r_rm = 0.59, P < 0.01) legs. The CAR was also significantly correlated with force gains in the trained (r_rm = 0.45, P < 0.01) and untrained (r_rm = 0.68, P < 0.01) arms. The correlation between strength gains and changes in MUFRs was poor (r_rm = 0.04 – 0.19) and nonsignificant (P > 0.05) for all limbs.

Variability and cocontraction of ramp contractions.

There was no consistent pattern of change in the antagonist cocontraction and no significant changes in the untrained limb at any session, as assessed by ANCOVAs (P > 0.05). As a result, the motor skill findings will focus on the variance ratio and RMSE. The only significant changes in the variance ratio at posttraining or retention occurred in the trained leg at retention (−28%, P = 0.03) and untrained leg at posttraining (−30%, P = 0.03). The RMSE demonstrated no significant changes in any limb at posttraining or retention testing (P > 0.05). As expected, there was an increase in variability during the no feedback contractions for all limbs due to the removal of concurrent feedback; therefore, the percent changes (baseline to no feedback) were statistically compared (ANOVA) between experimental (trained or untrained) and control (dominant or nondominant) limbs. The force variability measures calculated during retention contractions without feedback (no feedback) are presented in Fig. 8 as a percent change from baseline. A smaller increase relative to baseline indicates a greater amount of learning.

Fig. 8.

Percent change from baseline to “no feedback” calculated from least squares means of the variance ratio (left) and root-mean-square error (RMSE; right) calculated from the force trace of the trapezoidal contractions; n = 20. A: upper limb; B: lower limb. Note: a lower percent change indicates better performance (i.e., less variability) during ‘no feedback’ contractions.

Overall, the variance ratio and RMSE had greater increases in the upper limb than the lower limb, indicating that the wrist flexion ramp was a more difficult task without feedback than the dorsiflexion and/or participants found it more difficult to estimate 60% of their maximum in wrist flexion. In the upper limb, the variance ratio did not indicate that there was any training or cross-education adaptations contributing to motor skill improvement. However, the RMSE demonstrated a significantly smaller increase from baseline in the untrained (29%, d = 0.66) compared with the nondominant control (97%, d = 2.14) arm and a smaller, but nonsignificant, increase in the trained arm (55%, d = 1.27) compared with the dominant control (71%, d = 1.63) arm. In the lower limb, the variance ratio demonstrated a motor learning adaptation with a significantly smaller increase from baseline in the untrained (42%, d = 0.81) compared with the nondominant control (136%, d = 1.68) leg, and a nonsignificantly smaller increase in the trained (83%, d = 1.11) compared with the dominant control (151%, d = 1.50) leg. The RMSE in the lower legs demonstrated a very consistent increase across all legs (18–25%, d = 0.45–0.68).

DISCUSSION

The purpose of this study was to evaluate the neuromuscular mechanisms contributing to strength and skill improvements following a 6-wk unilateral, strength training program. In addition, the persistence of neuromuscular adaptations was examined following 6 wk of detraining. The unilateral training study successfully induced cross-education strength gains of magnitudes in line with previous work (Carroll et al. 2006; Manca et al. 2017; Munn et al. 2004). The experimental design of the present study ensured that the contralateral strength gain was due to cross education of unilateral training. Carroll and colleagues (2006) showed that the lack of a familiarization session overestimated the cross education strength gain by up to 4%. In agreement, the present study found a strength increase from familiarization to baseline of 4% in the contralateral arm and 7% in the contralateral leg. Therefore, the exclusion of a familiarization session would have greatly overestimated the cross-education effect. Furthermore, the presence of a control group ensured that the strength gain was due to the cross education of unilateral training rather than a training effect due to repeated testing or increased familiarity with the testing equipment. The effect sizes of cross education at posttraining and retention (d = 0.32–0.58) indicate a small to moderate effect, which has potential for clinically relevant gains.

Few studies have experimentally compared cross education between men and women and between the upper and lower limb. Therefore, 20 men and 20 women were randomly allocated to an arm-training (wrist flexion) or leg-training (dorsiflexion) group. No sex difference were observed in the patterns of change, confirming the results of Hubal and colleagues (2005). To date, the authors are not aware of any unilateral training study that statistically compared the magnitude cross education between upper and lower limbs following matched training protocols. Two meta-analyses have pooled upper and lower limb data to determine the effect of limb on cross education. Munn and colleagues (2004) estimated the cross-education strength gain to be 3.8% in the upper limb compared with 10.4% in the lower limb; however, this difference was nonsignificant (P = 0.16). Alternatively, Manca and colleagues (2017) found a significant (P = 0.006) difference between the estimated 9.4% strength gain in the upper limbs and the 16.4% strength gain in the lower limbs. The present study demonstrated a slightly smaller contralateral strength gain at posttraining in the upper limb (11%) compared with the lower limb (15%), but there was no significant difference between the magnitude of cross education in the upper and lower limbs. However, inspection of Fig. 4 demonstrates that the magnitude of cross transfer (strength gain from trained to untrained) in the lower limb is somewhat larger than in the upper limb.

Cross education has been demonstrated to be highly specific to the training contraction type (Beyer et al. 2016; Hortobágyi et al. 1997; Hubal et al. 2005) and to the homologous muscle (Hortobágyi 2005; Hortobágyi et al. 1997, 1999; Lee and Carroll 2007). Therefore, the magnitude of cross education demonstrated in the present study may have been underestimated due to the different contraction types used for training (dynamic) vs. testing (isometric). The use of dynamic contractions was selected to be most applicable to everyday movement (i.e., activities of daily living) and logistical for rehabilitation programs, while isometric testing was selected for the purpose of recording motor unit firing rates. The specificity of training also applies to the task, flexion vs. extension flexion contractions. An increase in extension force at posttraining in all limbs was not likely a feature of training or cross education. Rather, the fewer number of total contractions for extension resulted in continued familiarization in the extension direction.

The correlations between contralateral strength gains and the neuromuscular mechanisms provide linkage between central adaptations and cross education. The discrepant findings in the correlation of RMS/Mmax and force between the trained and untrained limbs, as well as between the upper and lower limbs, follows the conflicting results of previous research. Interestingly, there was no significant correlation between FCR activity and wrist flexion force, but there was a significant correlation between TA activity and dorsiflexion force in both the trained and untrained limbs. Assessment of FCR activity scatter plots reveals that participants with high RMS/M-wave amplitude at baseline demonstrated decreases at posttraining and retention, indicating an inflated value at baseline. However, these participants were not removed as their values did not meet the threshold criteria for outliers. There was no significant correlation between motor unit firing rates and force for any limb due to the lack of change in motor unit firing rates. Alternatively, both the V/M ratio and CAR demonstrated strong correlations with contralateral force, confirming the increase in central drive to the contralateral limb following unilateral training (Aagaard et al. 2002; Fimland et al. 2009). It is not surprising that the neuromuscular adaptations mirrored the training and cross-education force increases at posttraining and retention.

There were small increases (6–15%) in the contralateral limb’s agonist activity at posttraining; however, these were not accompanied by any change in MUFRs. Previous cross-education studies examining muscle activity have found ambiguous results, which are highly variable depending on the training paradigm. Similar to the present study, two studies investigating the change in motor unit firing rates following unilateral training found no significant change in either the trained or untrained limbs despite ipsilateral and contralateral strength gains (Patten et al. 2001; Rich and Cafarelli 2000). This lack of change in MUFRs may indicate a methodological limitation of recording firing rates during the ramp contractions (at 60% maximal force) rather than during the MVCs. Previous work by Kamen and Knight (2004) found a significant increase in maximal contraction MUFRs with familiarization despite no significant change in 10 and 50% force contractions. It is likely that small adaptations may have been present if contractions were performed at the same absolute force as baseline (i.e., 60% baseline force) rather than a percentage of the increased strength.

Despite a lack of adaptation in the MUFRs, the increase in agonist activity was accompanied by increases in central drive, as assessed by the V-wave (Aagaard et al. 2002; Fimland et al. 2009; Upton et al. 1971). Two previous cross-education studies examining V-wave amplitude found no significant increase in the contralateral limb following training (Fimland et al. 2009; Colomer-Poveda et al. 2017). The present study demonstrated moderate to large increases in the V/M ratio in the untrained limb after 6 wk of training (41%, d = 0.65) and persisting after 6 wk of detraining (34%, d = 0.57). Similarly, the CAR was used to evaluate the “completeness of skeletal muscle activation” (Behm et al. 2001; Kent-Braun and Le Blanc 1996; Shield and Zhou 2004), and quantify the amount of efferent drive to the muscle (Shima et al. 2002). Previous work has demonstrated equivocal results in contralateral voluntary activation following unilateral training (Lee et al. 2009; Shima et al. 2002; Tillin et al. 2011). In the present study, the CAR of the lower limb was already >99% at baseline. However, the upper limb demonstrated moderate to large increases in the CAR of the contralateral arm at posttraining (1.2%, d = 0.51) and a continued increase to retention testing (1.7%, d = 0.70).

The persistence of strength and neuromuscular mechanisms following detraining in the present study is a novel finding in the field of cross education. Previous unilateral training studies (range 3–10 wk training), which examined cross education following a period of detraining (range 2–12 wk), all reported a loss in cross-education strength following detraining (Housh et al. 1996a, 1996b; Houston et al. 1983; Narici et al. 1989; Shaver 1975; Shima et al. 2002; Slater-Hammel 1950; Weir et al. 1997). Interestingly, the present study demonstrated a continued increase in contralateral strength following detraining. This continued increase resulted in the untrained limb strength being equal to the trained limb strength at retention. This was accompanied by the maintenance or continued increase of the neuromuscular adaptations (RMS/Mmax, V/M ratio, and CAR) in the contralateral limbs. This novel finding indicates the contribution of motor learning to cross education.

The permanency of motor learning was assessed during retention testing using force variability calculated during no feedback contractions. The variance ratio was calculated to evaluate the stability of motor performance by measuring the participant’s ability to reproduce a consistent motor pattern (McGuire et al. 2014a). The RMSE was calculated to evaluate task performance by measuring the ability to match a target and calculating force steadiness across a targeted plateau (McGuire et al. 2014a). Surprisingly, the variance ratio and RMSE did not significantly change in the trained and untrained limbs at posttraining and retention. However, some evidence of motor learning in the untrained limb was present in the no feedback contractions demonstrated in the RMSE for the untrained arm and in the variance ratio for the untrained leg, indicating that the contralateral limbs were able to reproduce the target task more consistently and steadily than the control limbs. The lack of improvement in variability at posttraining and retention is likely due to the different contraction types used in training (dynamic self-paced to 80% force) vs. testing (isometric ramps to 60% force). However, the presence of motor learning during the no feedback contractions suggest that the cross education of skill is robust in the face of specificity.

The presence of motor learning following strength training demonstrates the skill component of strength training and the cross education of skill without a specific coordination task. These results suggest that the theories of cross education involving the bihemispheric storage and interhemispheric access of motor engrams (Lee et al. 2010; Parlow and Kinsbourne 1989; Ruddy and Carson 2013; Taylor and Heilman 1980) may be generalizable to tasks performed by the homologous muscle beyond the practiced exercise (i.e., contraction type). Additionally, the changes in strength and neuromuscular adaptations were similar between the trained and untrained limbs, providing support for the hypothesis that unilateral training induces adaptations of a motor center (such as the premotor cortex) that provides common drive to both hemispheres (Ruddy and Carson 2013).

Conclusion.

In the present study we experimentally evaluated the contributions of neuromuscular adaptations and motor learning to cross education with the following conclusions. There was no difference between sexes or upper and lower limbs in the magnitude of training or cross education. Cross education of strength totaled 15% in the untrained wrist flexors and 14% in the untrained dorsiflexors. Agonist RMS amplitude, V-wave amplitude, and CAR confirmed the neuromuscular adaptations associated with cross education. However, there was no change in motor unit firing rates at 60% of maximal force. The continued increase in contralateral strength at retention demonstrated the persistence of cross education following 6 wk of detraining. Force variability measures demonstrated the presence of motor learning in the contralateral limbs.

GRANTS

This work was funded by an operating grant to D. A. Gabriel from the Natural Sciences and Engineering Research Council of Canada.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

L.A.G. and D.A.G. conceived and designed research; L.A.G. performed experiments; L.A.G. and D.A.G. analyzed data; L.A.G. and D.A.G. interpreted results of experiments; L.A.G. and D.A.G. prepared figures; L.A.G. and D.A.G. drafted manuscript; L.A.G. and D.A.G. edited and revised manuscript; L.A.G. and D.A.G. approved final version of manuscript.