Brain Topogr (2007) 20:77–88 DOI 10.1007/s10548-007-0033-2
ORIGINAL PAPER
Neuro-Physiological Adaptations Associated with Cross-Education of Strength Jonathan P. Farthing Æ Ron Borowsky Æ Philip D. Chilibeck Æ Gord Binsted Æ Gordon E. Sarty
Accepted: 8 August 2007 / Published online: 12 October 2007 Springer Science+Business Media, LLC 2007
Abstract Cross-education of strength is the increase in strength of the untrained contralateral limb after unilateral training of the opposite homologous limb. We investigated central and peripheral neural adaptations associated with cross-education of strength. Twenty-three right-handed females were randomized into a unilateral training group or an imagery group. A sub-sample of eight subjects (four training, four imagery) was assessed with functional magnetic resonance imaging (fMRI) for patterns of cortical activation during exercise. Strength training was 6 weeks of maximal isometric ulnar deviation of the right arm, four times per week. Peak torque, muscle thickness (ultrasound), agonist–antagonist electromyography (EMG), and fMRI were assessed before and after training. Strength training was highly effective for increasing strength in trained (45.3%; P \ 0.01) and untrained (47.1%; P \ 0.01) limbs. The imagery group showed no increase in strength for either arm. Muscle thickness increased only in the trained arm of the training group (8.4%; P \ 0.001). After training, there was an enlarged region of activation in contralateral sensorimotor cortex and left temporal lobe during muscle contractions with the untrained left arm (P \ 0.001). Training was associated with a significantly greater change in agonist muscle EMG pooled over both limbs, compared to the imagery group (P \ 0.05). These results suggest that cross-education of strength may be J. P. Farthing (&) P. D. Chilibeck G. Binsted College of Kinesiology, University of Saskatchewan, 87 Campus Drive, Saskatoon, SK, Canada S7N 5B2 e-mail:
[email protected] R. Borowsky G. E. Sarty Cognition and Neuroscience Programs, Department of Psychology, College of Arts and Science, University of Saskatchewan, Saskatoon, SK, Canada
partly controlled by adaptations within sensorimotor cortex, consistent with previous studies of motor learning. However, this research demonstrates the involvement of temporal lobe regions that subserve semantic memory for movement, which has not been previously studied in this context. We argue that temporal lobe regions might play a significant role in the cross-education of strength. Keywords Unilateral training Motor learning Interlateral transfer fMRI
Introduction Cross-education is the performance improvement (i.e., strength, skill execution, endurance) of the untrained limb after a period of unilateral practice (i.e., strength training, skill learning, endurance training) of the homologous contralateral limb. Cross-education of strength has been attributed to neural mechanisms since it is unaccompanied by muscle hypertrophy [11, 12, 33]. Despite extensive scrutiny, the neuro-physiological mechanisms of crosseducation of strength have remained elusive [21]. There is general consensus that cross-education is caused by changes in neural activity or neural circuits, but whether the sites of neural adaptation reside in cortical, sub-cortical, or spinal pathways is uncertain [6, 29]. Peripheral neural adaptations may occur in both limbs in response to unilateral training including increased agonist activation [22, 49] and decreased antagonist co-activation [5]. These adaptations are not always confirmed [9], and unfortunately, they cannot provide much insight into how the adaptations are transferred to the untrained limb. Adaptations at the level of the spinal cord, such as increased spinal reflex excitability, do not appear to accompany
123
78
cross-education of strength suggesting a prominent role of supraspinal mechanisms [28]. The theory that cross-education is controlled centrally through communication between cortical hemispheres can be dated as far back as Hellebrandt [19], yet no study to date has examined cortical adaptation with cross-education of strength. Crosseducation of strength has been shown with imagery training [55], which suggests key influence from the brain for cross-education since imagery training should eliminate any influence from muscles. However, studies are mixed as to the effectiveness of imagery training for increasing strength [20, 43, 58]. Neuro-imaging techniques provide the opportunity to gain insight into the neural mechanisms associated with cross-education of strength. Functional imaging has not previously been used to examine cortical activation before and after unilateral resistance training. Cross-education of strength could involve some form of motor learning [29], which may be associated with neural plasticity in the brain, in regions including motor cortex (M1) [26, 40, 41, 46], premotor cortex (PMC) and supplementary motor area (SMA) [17, 18], and cerebellum [25, 45]. Bilateral activations in M1, SMA, and PMC are commonly associated with unilateral limb movement [2, 8, 14, 27, 30, 37, 51, 53]. Particularly in M1, plasticity plays a role in the early consolidation of learning motor skills [34, 50, 57]. Increased or decreased activation in M1 after training could reflect, respectively, increasing activity of neurons associated with descending signal to agonists, and decreasing activity of neurons projecting to antagonists (decreasing co-activation). Decreased co-activation has been documented using electromyography (EMG) in conjunction with cross-education of strength [5]. However, comparisons between changes in cortical and muscle activation with cross-education of strength have never been made. The objective of this investigation was to determine if cross-education of strength is accompanied by changes in brain activation assessed by functional magnetic resonance imaging (fMRI), supporting the hypothesis that the effect is mediated by supraspinal mechanisms [12]. Specifically, we hypothesized that for both limbs, central neural adaptation would involve changes in activation with training and peripheral adaptation would involve changes in agonist– antagonist coordination beneficial for the desired direction of movement.
Materials and Methods Subjects Ethical approval for the experiment was given by the University of Saskatchewan’s biomedical review board for
123
Brain Topogr (2007) 20:77–88
research in human subjects, and all participating subjects gave their written informed consent. Twenty-three righthanded females with little prior resistance training experience in the previous year volunteered for the study. A sub-sample of eight subjects participated in the MRI experiment (described below). The subject pool was restricted to one gender for purposes of reducing betweensubject variation, and to replicate previous studies in our lab [12]. Prior to any physical assessment, all subjects reported prior resistance training experience and completed the Waterloo Handedness Questionnaire (WHQ) to determine the degree of handedness [4]. Subjects were excluded from the study if they did not obtain a positive score on the WHQ (indicating right handedness). Right-handed subjects were chosen in order to replicate the substantial amount of cross-education (39%) found previously after dominant hand training in right-handers [12].
Design Prior to the testing protocol, all subjects were familiarized with the testing environment/protocol and baseline measures were taken. To alleviate some of the learning effect that might be attributed to the testing protocol, each subject performed two or three low intensity practice repetitions of the strength task to become familiar with the motor requirements. Subjects later returned for baseline assessment of right and left arm muscle thickness of the posterior medial forearm, followed by right and left limb maximal isometric ulnar deviation peak torque measurement with simultaneous measurement of muscle activation by EMG. Once baseline was completed, subjects were randomized to a physical training (n = 12) or imagery group (n = 11). Subjects had a mean (±SE) age of 21.0 ± 0.5 years, mass of 62.4 ± 1.6 kg, height of 166.5 ± 1.5 cm, training experience of 3.5 ± 0.7 months, and handedness score of 17.1 ± 0.7. There were no baseline differences between groups. Once training was complete, all participants were again assessed for muscle thickness, isometric peak torque, and EMG in both the trained and untrained limbs.
Muscle Thickness, Strength, and Electromyography Muscle thickness, isometric ulnar deviation peak torque, and maximal isometric muscle activation were measured identically to recent studies in our lab, and detailed methods and reproducibility for each procedure has been reported [12]. Muscle thickness was measured using B-Mode ultrasound (Aloka SSD-500, Tokyo, Japan) and was completed on both the right and left posterior medial forearm (flexor
Brain Topogr (2007) 20:77–88
carpi ulnaris and flexor digitorum superficialis) before and after training. A stringent and reliable method for accurate muscle thickness measurement was adhered to [12]. Muscle thickness was assessed prior to any strength assessment. Isometric ulnar deviation peak torque was assessed on a Biodex isokinetic dynamometer (Biodex System 3, Biodex Medical Systems, Shirley, NY, USA) before and after training for both the right and left arm, and was done in the supine position to resemble the MRI setup. All participants were instructed that optimal performance of the ulnar deviation strength task was achieved by combining ulnar deviation of the wrist and a handgrip contraction (a weak grasp of the handle would result in weak wrist deviation torque). The highest peak torque achieved in four attempts was used for comparison. Maximal isometric muscle activation as assessed by EMG (Bagnoli-2, Delsys Inc., Boston, MA, USA) was collected for the right and left arm during each of the four repetitions on the isokinetic dynamometer for both the primary agonist (flexor carpi ulnaris) and the primary antagonist (extensor carpi radialis). Raw data (volts) were later converted to mean absolute value (MAV) using the accompanying software (EMGworks, version 2). A reference electrode (common ground) was applied to the kneecap with a single conductor. The EMG main amplifier unit included single differential electrodes with a bandwidth of 20 ± 5 Hz to 450 ± 50 Hz, a 12 dB/octave cut-off slope, and a maximum output voltage frequency range ±5 V. The overall amplification or gain per channel was 1,000. The system noise was \1.2 lV (rms) for the specified bandwidth. The electrodes were two silver bars (10 mm · 1 mm diameter) spaced 10 mm apart, with a Common Mode Rejection Ratio (CMRR) of 92 dB. A stringent method for land-marking the electrodes was utilized to ensure accurate placement of the electrodes for each testing occasion [12].
Strength Training Program Training consisted of 6 weeks of unilateral maximal isometric ulnar deviation of the right hand four times per week for a total of 24 training sessions in order to replicate the high magnitude of cross-education we reported previously [12]. Individuals were not allowed to complete post testing until they completed at least 21 of 24 training sessions. The training program was progressive in nature beginning with two sets of eight repetitions and progressing up to six sets of eight repetitions by the second week of training. A taper down to three sets was used for the last two training sessions to facilitate recovery from any overtraining [16]. Training was completed on an isokinetic dynamometer in
79
identical fashion as described for testing. Each isometric contraction was 2-s long and was separated by 2-s of rest (cadenced by a metronome). The rest interval between sets was 30 s. The primary researcher was present during the training sessions and provided verbal encouragement and reminded subjects to relax their non-training arm. All subjects were allowed to view their real time torque production during training to provide a measure of performance feedback. In order to avoid muscle activation in the non-training arm during training, participants were encouraged to relax the non-training arm and rest it across their waist area. Previous studies in our lab have demonstrated that ulnar deviation can be performed without significant activation of the non-training limb [12].
Imagery Group To account for the potential learning effect associated with becoming familiar with the lab environment and the strength training apparatus (dynamometer), and any possible effect of imagery or mental training [20, 43, 54, 57], the training group was compared to a group who participated in 6 weeks of motor imagery for a total of 24 sessions. Consequently, the change in strength, muscle activation, or brain activation in the imagery group could be compared to the changes shown in the training group to determine the effect of physical training alone. All imagery sessions were completed in the lab using the same dynamometer set up (right hand) as the physical training group. Subjects were positioned in the chair the same as for the testing procedures and grasped the handle. Each motor imagery session involved 2-s long repetitions, separated by 2-s of rest, and was cadenced (muted metronome) by the primary researcher using verbal prompts (i.e., ‘‘contract’’ and ‘‘relax’’). All imagery group subjects were encouraged to focus on a single point somewhere on the ceiling until the training set was completed. Prior to each imagery session, subjects were instructed to focus on the internal aspects of imagery (i.e., feelings and sensations of the particular muscle contraction) and avoid actual contraction of the muscles at all times.
MRI Experiment Functional magnetic resonance imaging was used to assess cortical adaptation before and after unilateral training of the right hand. Subjects participating in the MRI experiment (n = 8) were recruited from each group (training: n = 4; imagery: n = 4). The MRI scans were completed prior to all other measures (i.e., strength, muscle thickness, and EMG). A special familiarization session was completed
123
80
during the week prior to the initial MRI scan. Each subject practiced the requirements of the MRI sequence and the strength task using a specially constructed model MRI contained near the MRI suite. The MRI itself was not available for familiarization due to clinical use during the week, but a brief orientation was done immediately prior to scanning. The model MRI provided subjects with a good simulation of being inside an MRI with the head coil attachment. This was done to reduce potential discomfort or anxiety experienced by subjects during actual image acquisition and reduce the chance of head movement. The model MRI was equipped with the same computer software (EPrime) as the MRI itself to expose subjects to the prompts and instructions they would be viewing during true imaging. The MRI experiment consisted of two conditions: left hand contractions and right hand contractions. The order of conditions was random for each subject and kept consistent for pre- and post-training MRI scans. Participants completed five sets of eight repetitions for each condition. During each set, muscle contraction was alternated with rest to simulate a typical training set. When the word ‘contract’ appeared on screen, participants engaged in the contraction until it disappeared. After eight repetitions, a 30-s rest interval was given, where participants viewed a fixation cross (+). This cycle was repeated until five sets were completed for the condition. Limiting the amount of motion artefact in the data is vitally important for producing valid fMRI maps for movement paradigms. Maximal contractions could not be used in the MRI due to motion artefact caused by head and body movement from intense exertion. So to avoid motion artefact with our strength task, participants were instructed to use a contraction intensity of 60% of maximum. Pilot testing revealed that this was appropriate to produce very intense motor cortex response without introducing motion artefact. To further minimize head motion during contraction, participants were instructed to avoid sudden ‘jerking’ movements when contraction was initiated. This was facilitated by using ‘ramp’ contractions, with a smooth transition between contractions. For each condition, the uninvolved limb was rested across the participants’ waist
Fig. 1 Illustration of the Biodex handle setup for isometric ulnar deviation (A) and the specialized wooden apparatus handle constructed for the same task in the MRI environment (B)
123
Brain Topogr (2007) 20:77–88
with the intention of restricting movement during contractions of the involved limb. The duration of contraction in the MRI (1.8 s) was slightly shorter than for training and testing (2 s). Pilot testing revealed that 2-s acquisition intervals for contractions in the MRI was too close to the natural breathing rhythm and participants found it difficult to avoid synchronized breathing with the task (e.g., always exhaling during the contraction and inhaling during rest). A previous experiment in our lab demonstrated that breathing in synchrony with a motor task can be associated with activation artefact [13]. Therefore, subjects were instructed to avoid breath-hold at any time and avoid synchronizing their breathing with the task. This was facilitated by a slightly shorter acquisition time for contractions in the MRI (1.8 s). Isometric ulnar deviation of the right or left hand was made possible in the MRI with the construction of a specialized wooden apparatus with two handles fastened to a 2 · 3 foot section of half-inch plywood. The apparatus was constructed so the position of the arm relative to the body was very similar to the dynamometer set-up (along side and in plane with the body), with the exception that the apparatus was made with handles for both hands. It was necessary that the apparatus could accommodate contractions with either hand without repositioning the subject. The isokinetic dynamometer handle attachment was mimicked by a wooden dowel of equal diameter pierced through a stack of wood blocks (for reinforcement) and glued down into a flat wood plank (see Fig. 1). The wood plank apparatus could be maneuvered up and down the MRI bed to accommodate different arm lengths, and was positioned so the participant was in a neutral position, comfortable, and not forced into shoulder depression or elevation. All imaging procedures were completed with a 1.5-T Siemens (Erlangen, Germany) Symphony Magnetic Resonance Imager. For each condition, 168 volumes of 12 slice axial single-shot echo planar images (EPI) were obtained (90 flip angle, TR = 1,800 ms, TE = 55 ms, 64 · 64 acquisition matrix, 128 · 128 reconstruction matrix). The first eight volumes were used for image stabilization and discarded prior to analysis, and the remaining 160 volumes
Brain Topogr (2007) 20:77–88
81
consisted of five blocks of 32 volumes each. Each block consisted of 16 volumes of task (ulnar deviation contractions) and 16 volumes of rest. Slice thickness was 8 mm with 2 mm separation between slices. Prompting for the experiment was done using a data projector and computer running EPrime software (PST Inc., Pittsburgh, PA, USA) and a projection screen visible to the subject by a mirror attached to the head coil. EPrime software was used to trigger image acquisition in synchrony with the presentation of visual prompts. Prior to completing the experimental conditions, T1 weighted spin-warp spin-echo anatomical images (TR = 400 ms, TE = 12 ms, 256 · 256 acquisition matrix) were acquired in the axial, sagittal, and coronal planes, and used for overlaying functional activation maps. To obtain the full volume of cortex for each participant, the third or fourth inferior-most slice was centered on the posterior commissure, depending on the distance from the posterior commissure to the top of the brain. Slice thickness was 8 mm with 2 mm spacing for all anatomical images. The blocked design experiment was analyzed using the BOLDfold technique [13, 47]. BOLDfold measures the average BOLD (Blood-Oxygen-Level-Dependent) response per block and requires a sufficient rest interval between stimulus volumes for each response to return to baseline. After correction for baseline drift, the BOLDfold method consists of determining the mean BOLD function for each voxel, averaged over the design blocks, and then repeating and concatenating those mean functions n times, where n is the number of blocks, and correlating that function to the actual data as measure of consistency across repetitions. Voxels with a correlation greater than g = 0.60 were considered activated. The corresponding P-value (Type I error rate) for that correlation is P \ 0.05 (twotailed) when using a conservative Bonferroni-correction for 100,000 comparisons. The activation intensity for an activated voxel was defined as the maximum of the mean BOLD function. Following the BOLDfold analysis, two maps were computed for each condition (right-hand and left-hand contractions), a threshold map gCondition(p) of g goodnessof-fit values and a visibility or intensity map VCondition(p) representing BOLD amplitude, where p is a voxel coordinate [3]. The activation map for each condition was defined as MCondition(p) = KCondition,g(p) VCondition(p) where KCondition,g(p) = 1 if gCondition(p) [ 0.60 and zero otherwise. Following these computations, intersection (shared) maps (Mint) and unique maps (Muni) were computed for paired conditions pre and post for each subject according to: Mint ðpÞ ¼ Kpre;g ðpÞ Kpost;g ðpÞ ðVpre ðpÞ þ Vpost ðpÞÞ=2
ð1Þ
Muni ðpÞ ¼ ½Kpre;g ðpÞ Vpre ðpÞ Kpost;g ðpÞ Vpost ðpÞ ½1 Kpre;g ðpÞ Kpost;g ðpÞ
ð2Þ
The intersection map shows only activated voxels that are common between conditions pre and post, and the unique map represents a difference and shows task subtraction for activations that are not common to conditions pre and post. The intersection and unique maps were averaged across participants within each group (training, imagery) to compute the final maps. The final maps included intersection and unique maps overlaid together on a standardized image underlay to produce a single map for each stimulus condition (right and left contractions) displaying both shared and unique activations. Using Analysis of Functional NeuroImages (AFNI) software [7], the maps were generated in accordance with the methods of Borowsky et al. [3]. Voxels separated by 1.1 mm distance were clustered and clusters of volume less than 100 ll were clipped out. The data were then blurred using an isotropic Gaussian blur with a full width at half maximum (FWHM) of 3.91 mm. Averaging of images within groups was done following Talairach transformation of individual data to a standardized brain atlas [52]. Mean activation maps in Talairach coordinates were determined for each condition along with the corresponding one sample t statistic for each voxel. The maps show regions of activations that exceed both the activation g threshold and pass a one-tailed t-test of the activation amplitude against zero for the group. The final maps, constructed from the average maps, are binary (1 = on, 0 = off) and are presented without color scaling to show varying intensity [3].
Statistics Strength and muscle thickness were analyzed using a Group (training, imagery) · Time (pre, post) · Arm (trained, untrained) multivariate analysis (SPSS version 11.5) appropriate for multiple dependent variables with repeated measures on the last two factors. Percent change for strength and muscle thickness was determined by subtracting the pre-training from the post-training score, dividing by the pre-training score and multiplying by 100[(post – pre)/pre · 100]. This was analyzed using a Group (training, imagery) · Arm (trained, untrained) multivariate analysis to help simplify the results and to allow comparison of the relative magnitude of cross-education as it is commonly reported in the literature [35]. Univariate ANOVA and relevant one-way ANOVA and paired t-tests followed if multivariate significance was found. EMG activation was analyzed using a Group
123
82
Brain Topogr (2007) 20:77–88
(training, imagery) · Time (pre, post) · Arm (untrained, trained) · Muscle (agonist, antagonist) ANOVA with repeated measures on the last three factors. EMG was not included in the multivariate analysis because of an additional factor in the design (agonist, antagonist muscles). A separate Group (training, imagery) · Time (pre, post) · Arm (trained, untrained) ANOVA was used to compare the change in strength over time for the MRI sub-sample. Significance was set at a = 0.05.
Results Strength and Muscle Thickness As expected with strongly right-handed participants, the left arm was significantly weaker than the right arm for all subjects before and after training (Table 1, P \ 0.01). There was a significant Group · Time · Arm multivariate interaction for strength and muscle thickness (P \ 0.01). Univariate ANOVA for strength revealed a significant Group · Time interaction (P \ 0.05). There was a significant increase in strength over time for the trained arm (P \ 0.01), and the untrained arm (cross-education) of the training group (P \ 0.01). No significant changes over time for strength were observed for the imagery group. Univariate ANOVA for muscle thickness revealed a significant Group · Time · Arm interaction (P \ 0.01). Only the muscle thickness of the training arm of the training group increased significantly over time (Table 1, P \ 0.01). The analysis of percent change revealed a significant Group · Arm multivariate interaction for strength and muscle thickness (P \ 0.01). Univariate ANOVA for strength revealed a significant Group effect (P \ 0.01). There was no significant difference for relative strength change between the trained and untrained arm pooled across groups. The relative change in strength pooled across arms for the training group (46.5 ± 4.8%) was significantly greater than the imagery group (0.1 ± 5.0%; P \ 0.05). The relative magnitude of strength change for
Fig. 2 Relative magnitude of strength change for both arms in each group. *Percent change for each arm of the training group significantly greater than the corresponding arm from the imagery group (P \ 0.05)
both the trained and untrained arms is presented in Fig. 2. Univariate ANOVA for muscle thickness revealed a significant Group · Arm interaction, (P \ 0.01). The relative increase in muscle thickness for the trained arm of the training group (8.4 ± 1.6%) was significantly greater than all other arms (P \ 0.05). The separate analysis of strength for the eight subjects in the MRI sub-sample revealed a significant Group · Time interaction (P = 0.01). Both the right and left arms of the training group sub-sample increased significantly over time (P \ 0.05) with no significant changes for the imagery group sub-sample. Strength data over time for the MRI sub-sample is presented in Fig. 3.
Electromyography The Group · Time · Arm · Muscle ANOVA revealed a significant three-way interaction (Group · Time · Muscle) for MAV EMG with training (P \ 0.05). There was a significant Arm main effect (P \ 0.01), indicating greater overall activation during contraction with the trained arm pooled across groups. To better interpret the
Table 1 Maximum isometric strength and muscle thickness as a function of test (pre and post) and arm (right and left) for each group (Mean ± SE)
Strength (Nm) Muscle thickness (cm)
Training group (n = 11)
Imagery group (n = 12)
Pre
Post
Pre
Post
Untrained
15.3 ± 1.1
22.3 ± 1.5*
15.5 ± 1.6
15.8 ± 2.1
Trained
19.9 ± 1.3
28.8 ± 2.0*
18.8 ± 1.3
18.8 ± 1.6
Untrained Trained
3.10 ± 0.05 3.11 ± 0.06
3.10 ± 0.06 3.37 ± 0.07*
3.02 ± 0.08 3.04 ± 0.04
3.01 ± 0.06 3.03 ± 0.04
*Significantly different than corresponding pre-value (P \ 0.01)
123
Brain Topogr (2007) 20:77–88
83
Fig. 3 Strength change over time for the trained and untrained arm for the MRI subjects. MRI subjects from the training and imagery group are presented on separate plots. Solid lines represent the trained arm and dotted lines represent the untrained arm. The trend from the MRI sub-sample resembles the group data presented in Fig. 2 and Table 1. *Significant increase from pre to post (P \ 0.05)
three-way interaction, change scores were computed for each muscle and pooled across arms. One-way ANOVA (Group) was computed for the agonist change scores and was significant (P \ 0.05). The agonist change score for the training group (0.086 ± 0.037 mV) was significantly different from the imagery group (–0.041 ± 0.031 mV; P \ 0.05). Pre- and post-training MAV muscle activation values are presented in Table 2. fMRI Figure 4 displays average data for the right and left hand contraction conditions for the MRI subjects from each
group (i.e., training: n = 4; imagery: n = 4). The maps show three axial slices for each subject, displaying activation of contralateral pre- and post-central sulcus (M1 and S1) before and after the training period. These maps are intended to provide a representation of the data from which the proceeding unique and shared activation maps were computed. The average maps depict strong contralateral M1 activation for each condition. For the training group’s untrained left arm post-training (cross-education), contralateral M1 activation spreads further medial/dorsal and anterior into pre-central sulcus. Shared and unique average activation maps for pre- and post-training scans are both overlaid on the standardized brain image in Fig. 5. The activation maps are displayed on a 90% underlay (outer 10% is stripped off the brain) for a more revealing display of 3D rendered activation patterns by showing depth of activation into cortex [3]. Using the 90% underlay is more revealing than the 100% underlay, as activations are sometimes not visible on the surface of the cortex. Separate maps were generated for each condition (right and left contractions) and group (training, imagery). The 3D brain is displayed in two views: right and left hemisphere. Figure 5 displays shared and unique activations for preand post-training for both the right (trained) and left (untrained) contraction conditions for the training group (t(3) = 3.537, P \ 0.001). The left condition showed shared activation in right M1 and S1, and left ventral M1. Regions of activation unique to post-training included right M1 and S1, left ventral M1 and S1, left anterior and posterior middle temporal gyrus, left inferior temporal gyrus, medial occipital cortex and posterior medial and left lateral cerebellum. Regions of activation unique to pre-training included only dorsal M1/S1 and superior frontal gyrus. The right condition exhibited shared activation in left M1 and S1, and right superior occipital cortex. Regions of unique activation for post-training included left M1 and S1, left dorsal PMC, superior frontal gyrus, medial and left occipital cortex, and anterior cerebellum. Regions of activation unique to pre-training included right ventral S1, right PMC, right inferior occipital cortex, right posterior
Table 2 Maximal isometric muscle activation (mean absolute value from EMG) as a function of test (pre and post), arm (trained and untrained) and muscle (agonist and antagonist) for each group (Mean ± SE) Training group (n = 12) Pre Agonist (mV) Antagonist (mV)
Imagery group (n = 11) Post
Pre
Post
Untrained
0.215 ± 0.045
0.308 ± 0.046*
0.264 ± 0.036
0.238 ± 0.039
Trained
0.336 ± 0.040
0.416 ± 0.174*
0.376 ± 0.030
0.320 ± 0.027
Untrained Trained
0.090 ± 0.026 0.090 ± 0.019
0.062 ± 0.011 0.092 ± 0.025
0.075 ± 0.018 0.072 ± 0.013
0.069 ± 0.014 0.055 ± 0.007
*Change score for the agonist muscle pooled across arms significantly greater than imagery
123
84
Brain Topogr (2007) 20:77–88
Fig. 4 Average activation maps for three slices (z = 47 mm; z = 52 mm; z = 57 mm) depicting pre and post central sulcus (M1 and S1) activation pre- and post-training for the MRI sub-sample (n = 8; four training and four imagery). The z-coordinate is centered on zero at the posterior commissure after Talairach transformation. All activated regions exceed a threshold of g = 0.60 (t = 3.573; P \ 0.01), and are scaled for intensity (1.00 = max intensity on an arbitrary scale). The anatomical maps display right = left. Note the increased activation in medial/dorsal and anterior regions of pre central sulcus associated with the untrained left limb after training
inferior temporal gyrus, right anterior middle temporal gyrus, right prefrontal cortex, and left S1. Figure 5 displays shared and unique activations for preand post-training for both the right and left contraction conditions for the imagery group (t(3) = 3.537, P \ 0.001). The left condition showed shared activation in right M1 and S1, left inferior occipital cortex, and left anterior
cerebellum. Regions of activation unique to post training included left S2, left ventral PMC, medial occipital cortex, left prefrontal cortex, right M1 and S1, and right cerebellum. Regions of activation unique to pre-training included bilateral dorsal PMC, left anterior superior temporal gyrus, bilateral superior temporal gyrus/angular gyrus (Wernicke’s area), and posterior cerebellum. The right
Fig. 5 Shared and unique activations for pre- and post-training during right- and left-hand conditions for the training and imagery group. Yellow regions represent activations unique to post-training and blue regions represent activations unique to pre-training. Red regions represent activations shared by pre- and post-training. Total pre-training activation can be inferred by combining red and blue
regions, and total post-training activation can be inferred by combining red and yellow regions. Note the areas of increased activation after training for the untrained left arm of the training group. Activation regions exceed a threshold of g = 0.60 at the individual level, and group average activation amplitudes pass a onetailed t-test against zero for the group (P \ 0.001)
123
Brain Topogr (2007) 20:77–88
condition exhibited shared activation in left M1 and S1, left PMC, bilateral middle and inferior temporal gyrus, medial cerebellum, and superior frontal gyrus. Regions of unique activation for post training included bilateral ventral M1, bilateral prefrontal cortex, right PMC, right superior temporal gyrus, right lateral cerebellum, left dorsal M1 and S1, left SMA, and left inferior occipital lobe. Regions of activation unique to pre-training included right inferior temporal gyrus, anterior cerebellum, and superior frontal gyrus. There were no significant regions of activation in subcortical structures including basal ganglia, thalamus, and corpus callosum, for any condition in either group; therefore, anatomical views of these structures are not shown.
Discussion Several functional imaging studies have been able to confirm that changes in activation in the brain accompany movement task learning [17, 18, 26, 45], but this study is the first to demonstrate changes in cortical activation associated with strength training. Although others have speculated about the role of supraspinal adaptations in cross-education, this study provides some evidence that potent strength increases in the untrained limb are accompanied by changes in brain activation. Consistent with recent studies in our lab [12], substantial cross-education of strength was observed after right hand/arm training (47.7% increase in strength of the untrained limb) and was similar in magnitude to strength gained in the trained arm (45.3%; Fig. 2). Surprisingly, the magnitude of effect here (105%) was vastly larger than the typical magnitude of cross-education relative to the trained limb (35–60%) previously reported by others [6, 35, 56], and it substantially exceeds the average pre- to post-training strength increase in the untrained limb (7.6%) outlined by an updated meta-analysis [6]. The higher magnitude of cross-education in this study is likely a function of using the preferential direction of transfer (right-to-left) and an unfamiliar strength task [12]. Regions of new cortical activation after training associated with the untrained arm are consistent with areas implicated in motor learning [17]. Major regions of increased activation associated with cross-education of strength after training included the contralateral sensorimotor cortex (M1 and S1), and ipsilateral temporal lobe. The region of activation in sensorimotor cortex for the untrained arm after training was noticeably enlarged (yellow regions added to red regions in Fig. 5). Increased M1 activation is consistent with previous studies associating M1 plasticity or enlarged M1 activation with motor learning [17, 26, 40, 41]. Dorsal regions of sensorimotor
85
cortex were activated before training, but not after (blue regions in Fig. 5). This type of change in motor areas where new activations are accompanied by the suppression of previously activated regions has been reported with motor learning [42, 48, 51]. Harmonious with similar relative strength changes in both the trained and untrained arm with training (Fig. 2), changes in activation in contralateral sensorimotor cortex were evident in both arms. A similar change in activation in the contralateral motor strip with both arms suggests a potential role of interhemispheric communication regarding motor planning for producing strength. Inter-hemispheric communication is a theory regarding the mechanism of interlateral or bilateral transfer of skills. This suggests that cross-education of strength might be similar in nature to cross-education of skills [12, 29]. The ‘cross-activation model’ states that motor engrams for a skill or task are stored in both hemispheres with unilateral acquisition [39]. The model suggests that task acquisition with the more proficient system, such as with the dominant limb in our study, would provide a better-stored engram for the opposite limb. Imamizu and Shimojo [23] state that perfect transfer should occur if adaptations at the level of the brain are primarily involved in transfer of learning and intermediate transfer should occur if adaptation occurs in both the brain and the periphery. However, we observed perfect transfer ([100%) with evidence of peripheral and central adaptation. Notably, skill learning does not induce muscle hypertrophy, which obviously contributes to performance improvement with strength training. Since muscle hypertrophy accompanies strength gain only for the trained arm in this study, arguably, the relative contribution of neural adaptation was far greater in the untrained arm than the trained arm. The patterns of strength increase and muscle hypertrophy for both limbs following training provide evidence of this. Muscle thickness was almost identical in both arms prior to training (Table 1), yet the trained arm (dominant right) was significantly stronger than the untrained arm (non-dominant left), indicating a difference in neuromuscular coordination/recruitment strategies between arms favoring the dominant side. The 45.3% increase in strength of the trained arm was induced by a combination of muscle hypertrophy (Table 1) and neural adaptation. In comparison, the 47.7% increase in strength in the untrained limb was not accompanied by muscle hypertrophy (Table 1) indicating neural adaptation was largely responsible. In addition to increased activation in contralateral sensorimotor cortex, post training activation patterns with the untrained arm include ipsilateral (left) temporal lobe activation as part of new activation after training (mainly left middle temporal gyrus). In contrast, for the left arm of the imagery group, there is no temporal lobe activation in either hemisphere after the intervention period. Left middle
123
86
temporal gyrus has been implicated as the region for retrieval of motion knowledge [32]. Left temporal lobe is implicated in semantic memory, which refers to general worldly knowledge, including knowledge about words, objects, and how one interacts with them [31, 54]. If temporal lobe activation is involved in supraspinal adaptation with unilateral training the exact mechanism is uncertain, and the role of left versus right temporal lobe activation remains unclear in this context. Memory retrieval might be pertinent for cross-education of strength in terms of providing an internal representation of a movement task previously acquired by the opposite limb through training. Obayashi [38] theorized that the magnitude of transfer may primarily depend on the retrieval of prior memory and its use in creating an internal representation, as if re-living a past experience. Previous motor learning studies have reported temporal lobe activation with lefthand learning in right-handers [18]. New activations after training were present in the cerebellum for both arms after training. The cerebellum is thought to play an important role in motor learning [24, 46]. The cerebellum functions in comparing actual movement with intended in order to shape ongoing movement and improve accuracy [15]. Obayashi [38] argues that there is a strong relationship between prior memory and the cerebellum, and the mechanism of transfer. Increased activation in the cerebellum may reflect more precise timing of agonist, antagonist, synergist, or postural muscle activation, which results in improved coordination. The current study provides evidence of increased agonist muscle activation (Table 1) accompanying strength gain, but no measures of synergist or postural muscle activation were taken. Unilateral strength training was effective for increasing strength, muscle activation, and brain activation in both arms. Despite showing no effect for strength and no changes in muscle activation, the imagery group exhibited changes in brain activation in regions that were somewhat similar to the training group (e.g., M1). This could signify some adaptation in cortex from the imagery sessions [43]. Several studies report that motor imagery activates similar regions of cortex as actual movement [2, 36, 44]. Therefore, motor imagery involves activation of cortex, and repeated sessions of motor imagery could effectively train the brain, and lead to changes in brain activation over time. Since both groups showed changes in brain activation, yet only the training group improved strength, dissimilarities between groups can be emphasized as possible sites of supraspinal adaptation associated with cross-education. In general, the imagery group exhibited more new activation in the ipsilateral rather than the contralateral hemisphere for both the right and left arm. This contrasts with the training group, where activation patterns for the trained
123
Brain Topogr (2007) 20:77–88
right arm are more lateralized to the contralateral hemisphere (pre-training activation shown in the right hemisphere was not activated after training). Physical training appeared to result in a more focused contralateral activation pattern (Fig. 5), with the exception of the activation in the ipsilateral (left) temporal lobe for the untrained left arm. As compared to the training group, the imagery group showed much larger regions of shared activation (between pre and post) for the right arm (Fig. 5). For the left arm, the training group shows larger regions of shared and new activation in contralateral motor cortex (Fig. 5). The disparity in temporal lobe activation between the two groups is puzzling, and raises questions regarding the precise mechanistic role of temporal lobe activation for cross-education of strength. Left temporal lobe activation was evident at pre-, but not at post-test, for the left arm of the imagery group, whereas the training group exhibited left temporal lobe activation as part of new activation posttraining for the untrained left arm. For the left arm of the imagery group, an accurate internal representation of the task with which to evaluate the accuracy of the movement may not have been encoded, since no physical training of the either arm took place. The changes in activation associated with the right arm of the imagery group included temporal lobe, where decreased activation, shared activation, and a small amount of new activation was evident (Fig. 5). The disparities in temporal lobe activation between the right and left arm in the imagery group might be explained by the fact that the imagery sessions were completed only with the right arm. Verbal communication (written) involved in the testing procedure could be responsible for some of the temporal lobe activation shown in either group. However, the verbal prompts where the same for each testing condition, and therefore, similar temporal lobe activation would be expected in all conditions. Despite exhaustive procedures to control the consistency of the MRI experiment before and after training, the possibility of statistical artifact cannot be definitively ruled out because of the small sample size (n = 4). Replication of this study, with larger a sample size is required in order to confirm the changes in brain activation with cross-education of strength we have shown here. For strength to improve in an untrained limb after unilateral training there must be alteration in the motor recruitment strategy. The changes in cortical activation in the imagery group did not result in appropriate changes in neural recruitment necessary to elicit any increase in strength. For example, new activations in the hand region of the contralateral sensorimotor cortex were only shown for the untrained left arm of the training group (Fig. 5). Unfortunately, the exact resultant excitatory and inhibitory
Brain Topogr (2007) 20:77–88
output from the brain to the periphery cannot be determined by fMRI [1]. We hoped to gain insight into the peripheral changes with EMG. Peripheral changes at the muscle support the strength results (Table 1), in that agonist muscle activation increased to a similar extent in both arms after training. For the imagery group, a non-significant change in muscle activation (Table 2) coincided with non-significant changes in strength (Table 1). Admittedly, the accuracy of EMG for estimating muscle activation is questionable [10] and because our analysis did not include postural and synergist muscles, it does not represent the full scope of changes in neural coordination that might be reflected by activation changes in the brain. Nonetheless, there is limited evidence here to support both central and peripheral neural adaptation in conjunction with crosseducation of strength. In this paper we provide, for the first time, a modest indication that cross-education of strength may be controlled by changes in brain activation with training. We compared a unilateral physical training group to an imagery group, and found cross-education of strength (untrained limb) was associated with an enlarged region of activation in contralateral sensorimotor cortex (M1 and S1). In addition, ipsilateral temporal lobe activation was uniquely associated with the untrained limb, indicating a possible role of memory retrieval of prior movements acquired by the trained arm. These changes in brain activation with training might suggest a role of inter-hemispheric communication of an improved motor plan that can provide the untrained limb with a reference for preparation and execution of future movement. The precise role of interhemispheric communication in cross-education of strength remains an empirical question for future research. We encourage future studies to continue to investigate neural adaptation with strength training using neuro-imaging techniques and transcranial magnetic stimulation to better understand the process of neural adaptation with unilateral strength training. Our brain activation findings are limited by low subject numbers and the restrictions associated with the MRI environment. Nonetheless, this study represents a significant step towards understanding the complex higher order processes accompanying inter-limb interactions with strength training or motor learning. These findings have implications for rehabilitation strategies in response to unilateral injury (limb immobilization) or unilateral impairment (stroke), where training of the uninjured limb may involve neural adaptations beneficial for increasing strength of the injured or impaired contralateral limb. Acknowledgments Jonathan Farthing was supported by a graduate scholarship from the Natural Sciences and Engineering Research Council of Canada. Gordon Sarty, Ron Borowsky, and Gord Binsted are each supported by grants from the Natural Sciences and Engineering Research Council of Canada, which contributed to this
87 research. We acknowledge the technical support of Jennifer Hadley during the MRI scanning procedures and MRI data processing. We acknowledge Heather Whelan and Doug Jacobson for their technical assistance and support for the testing of participants.
References 1. Arthurs OJ, Boniface S. How well do we understand the neural origins of the fMRI BOLD signal? Trends Neurosci 2002;25: 27–31. 2. Binkofski F, Amunts K, Stephan KM, Posse S, Schormann T, Freund HJ, Zilles K, Seitz RJ. Broca’s region subserves imagery of motion: a combined cytoarchitectonic and fMRI study. Hum Brain Mapp 2000;11:273–85. 3. Borowsky R, Loehr J, Friesen CK, Kraushaar G, Kingstone A, Sarty G. Modularity and intersection of ‘what’, ‘where’, and ‘how’ processing of visual stimuli: a new method of fMRI localization. Brain Topogr 2005;18:67–75. 4. Bryden MP. Measuring handedness with questionnaires. Neuropsychologia 1977;15:617–24. 5. Carolan B, Cafarelli E. Adaptations in coactivation after isometric resistance training. J Appl Physiol 1992;73:911–7. 6. Carroll TJ, Herbert RD, Munn J, Lee M, Gandevia SC. Contralateral effects of unilateral strength training: evidence and possible mechanisms. J Appl Physiol 2006;101:1514–22. 7. Cox RW. AFNI: software for analysis and visualization of functional MR neuroimages. Comput Biomed Res 1996;29: 162–73. 8. Ehrsson HH, Fagergren A, Jonsson T, Westling G, Johansson RS, Forssberg H. Cortical activity in precision- versus power-grip tasks: an fMRI study. J Neurophysiol 2000;83:528–36. 9. Evetovich TK, Housh TJ, Housh DJ, Johnson GO, Smith DB, Ebersole KT. The effect of concentric isokinetic strength training of the quadriceps femoris on electromyography and muscle strength in the trained and untrained limb. J Strength Cond Res 2001;15:439–45. 10. Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG. J Appl Physiol 2004;96:1486–95. 11. Farthing JP, Chilibeck PD. The effects of eccentric training at different velocities on cross-education. Eur J Appl Physiol 2003;89:570–7. 12. Farthing JP, Chilibeck PD, Binsted G. Cross-education of arm muscular strength is unidirectional in right-handed individuals. Med Sci Sports Exerc 2005;37:1594–600. 13. Farthing JP, Cummine J, Borowsky R, Chilibeck PD, Binsted G, Sarty GE. False activation in the brain ventricles related to taskcorrelated breathing in fMRI speech and motor paradigms. Magn Reson Mater Phy 2007;20:157–68. 14. Gerardin E, Sirigu A, Lehericy S, Poline JB, Gaymard B, Marsault C, Agid Y, Le Bihan D. Partially overlapping neural networks for real and imagined hand movements. Cereb Cortex 2000;10:1093–104. 15. Ghez C. The cerebellum. In: Kandel ER, Schwartz JH, Jessell TM, editors. Principles of neural science. Norwalk, Connecticut: Appleton and Lange; 1991. 16. Gibala MJ, MacDougall JD, Sale DG. The effects of tapering on strength performance in trained athletes. Int J Sports Med 1994;15:492–7. 17. Grafton ST, Hazeltine E, Ivry RB. Functional anatomy of sequence learning in normal humans. J Cogn Neurosci 1995;7:497–510. 18. Grafton ST, Hazeltine E, Ivry RB. Motor sequence learning with the nondominant left hand: a pet functional imaging study. Exp Brain Res 2002;146:369–78.
123
88 19. Hellebrandt FA. Cross education: ipsilateral and contralateral effects of unimanual training. J Appl Physiol 1991;4:136–44. 20. Herbert RD, Dean C, Gandevia SC. Effects of real and imagined training on voluntary muscle activation during isometric contractions. Acta Physiol Scand 1998;163:361–8. 21. Hortoba´gyi T. Cross education and the human central nervous system. IEEE Eng Med Biol Mag 2005;24:22–8. 22. Hortoba´gyi T, Lambert NJ, Hill JP. Greater cross-education following training with muscle lengthening than shortening. Med Sci Sports Exerc 1997;29:107–12. 23. Imamizu H, Shimojo S. The locus of visual-motor learning at the task or manipulator level: implications from intermanual transfer. J Exp Psychol Hum Percept Perform 1995;21:719–33. 24. Ito M. Mechanisms of motor learning in the cerebellum. Brain Res 2000;886:237–45. 25. Jueptner M, Frith CD, Brooks DJ, Frackowiak RSJ, Passingham RE. Anatomy of motor Learning. II. Subcortical structures and learning by trial and error. J Neurophysiol 1997;77:1325–37. 26. Karni A, Meyer G, Jezzard P, Adams MM, Turner R, Ungerleider LG. Functional MRI evidence for adult motor cortex plasticity during motor skill learning. Nature 1995;377:155–8. 27. Kristeva R, Cheyne D, Deecke L. Neuromagnetic fields accompanying unilateral and bilateral voluntary movements: topography and analysis of cortical sources. Electroencephalogr Clin Neurophysiol 1991;81:284–98. 28. Lagerquist O, Zehr EP, Docherty D. Increased spinal reflex excitability is not associated with neural plasticity underlying the cross-education effect. J Appl Physiol 2006;100:83–90. 29. Lee M, Carroll TJ. Cross education: possible mechanisms for the contralateral effects of unilateral resistance training. Sports Med 2007;37:1–14. 30. Liu JZ, Dai TH, Elster TH, Sahgal V, Brown RW, Yue GH. Simultaneous measurement of human joint force, surface electromyograns, and functional MRI-measured brain activation. J Neurosci Method 2000, 101:49–57. 31. Martin A. Functional neuroimaging of semantic memory. In: Cabeza R, Kingstone A editors. Functional neuroimaging of semantic memory. Cambridge: MIT; 2001. 32. Martin A, Haxby JV, Lalonde FM, Wiggs CL, Ungerleider LG. Discrete cortical regions associated with knowledge of color and knowledge of action. Science 1995;270:102–5. 33. Moritani T, deVries HA. Neural factors verses hypertrophy in the time course of muscle strength gain. Am J Phys Med 1979;58:115–30. 34. Muellbacher W, Ziemann U, Wissel J, Dang N, Kofler M, Facchini S, Boroojerdi B, Poewe W, Hallet M. Early consolidation in human primary motor cortex. Nature 2002;415:640–44. 35. Munn J, Herbert RD, Gandevia SC. Contralateral effects of unilateral resistance training: a meta-analysis. J Appl Physiol 2004;96:1861–6. 36. Naito E, Kochiyama T, Kitada R, Nakamura S, Matsumura M, Yonekura Y, Sadato N. Internally simulated movement sensations during motor imagery activate cortical motor areas and the cerebellum. J Neurosci 2002;22:3683–91. 37. Nirkko AC, Ozdoba C, Redmond SM, Burki M, Schroth G, Hess CW, Weisendanger M. Different ipsilateral representations for distal and proximal movements in the sensorimotor cortex: activation and deactivation patterns. NeuroImage 2001;13:825–35. 38. Obayashi S. Possible mechanism for transfer of motor skill learning: implication of the cerebellum. Cerebellum 2004;3: 204–11.
123
Brain Topogr (2007) 20:77–88 39. Parlow SE, Kinsbourne M. Asymmetrical transfer of training between hands: implications for interhemispheric communication in normal brain. Brain Cogn 1989;11:98–113. 40. Pascual-Leone A, Grafman J, Hallet M. Modulation of cortical motor output maps during development of implicit and explicit knowledge. Science 1994;263:1287–9. 41. Pascual-Leone A, Nguyet D, Cohen LG, Brasil-Neto JP, Cammarota A, Hallett M. Modulation of muscle responses evoked by transcranial magnetic stimulation during the acquisition of new fine motor skills. J Neurophysiol 1995;74:1037–45. 42. Pearce AJ, Thickbroom GW, Byrnes ML. Functional reorganization of the corticomotor projection to the hand in skilled racquet players. Exp Brain Res 2000;130:238–43. 43. Ranganathan VK, Siemionow V, Liu JZ, Sahgal V, Yue GH. From mental power to muscle power-gaining strength by using the mind. Neuropsychol 2004;42:944–56. 44. Romero DH, Lacourse MG, Lawrence KE, Schandler S, Cohen MJ. Event-related potentials as a function of movement parameter variations during motor imagery and isometric action. Behav Brain Res 2000;117:83–96. 45. Sakai K, Ramnani N, Passingham RE. Learning of sequences of finger movements and timing: frontal lobe and action-oriented representation. J Neurophysiol 2002;88:2035–46. 46. Sanes JN. Neocortical mechanisms in motor learning. Cur Opin Neuro Biol 2003;13:225–31. 47. Sarty GE, Borowsky R. Functional MRI activation maps from empirically defined curve fitting. Concepts Magn Reson Part B (Magn Reson Eng) 2005;24B:46–55. 48. Shadmehr R, Holcomb HH. Neural correlates of motor memory consolidation. Science 1997;277:821–5. 49. Shima N, Ishida K, Katayama K, Morotome Y, Sato Y, Miyamura M. Cross education of muscular strength during unilateral resistance training and detraining. Eur J Appl Physiol 2002;86:287–94. 50. Sohn YH, Jung HY, Kaelin-Lang A, Hallett M. Effect of levetiracetam on rapid motor learning in humans. Arch Neurol 2002;59:1909–12. 51. Staines WR, Padilla M, Knight RT. Frontal-parietal event-related potential changes associated with practicing a novel visuomotor task. Brain Res Cogn Brain Res 2002;13:195–202. 52. Talairach J, Tournoux P. Co-Planar steroetaxic atlas of the human brain. Stuttgart: Theime; 1988. 53. Toma K, Honda M, Hanakawa T, Okada T, Fukuyama H, Ikeda A, Nishizawa S, Konishi J, Shibasaki H. Activities of the primary and supplementary motor areas increase in preperation and execution of voluntary muscle relaxation: an event-related fMRI study. J Neurosci 1999;19:3527–34. 54. Tulving E. Elements of episodic memory. New York NY: Oxford University Press; 1983. 55. Yue G, Cole KJ. Strength increases from the motor program: comparison of training with maximal voluntary and imagined muscle contractions. J Neurophysiol 1992;67:1114–23. 56. Zhou S. Chronic neural adaptations to unilateral exercise: mechanisms of cross-education. Exerc Sport Sci Rev 2000;28:177–84. 57. Ziemann U, Muellbacher W, Hallett M, Cohen LG. Modulation of practice-dependent plasticity in human motor cortex. Brain 2001;124:1171–81. 58. Zijdewind I, Toering ST, Bessem B, Van Der Laan O, Diercks RL. Effects of imagery motor training on torque production of ankle plantar flexor muscles. Muscle Nerve 2003;28:168–73.
