Balance maintenance as an acquired motor skill: Delayed gains and robust retention after a single session of training in a virtual environment

Available online at www.sciencedirect.com

www.elsevier.com/locate/brainres

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 2

http://dx.doi.org/10.0006-8993/& 2015 El

nCorresponding autE-mail addresse

[email protected]

Research Report

Balance maintenance as an acquired motorskill: Delayed gains and robust retention aftera single session of training in a virtual environment

Orit Eliona,e,n, Itamar Selab, Yotam Bahata, Itzhak Siev-Nerc,Patrice L. (Tamar) Weissd, Avi Karnib,e,f

aThe Center for Advanced Technologies in Rehabilitation, Rehabilitation Hospital, C. Sheba Medical Center, Tel Hashomer,IsraelbThe Edmond J. Safra Brain Research Center for the Study of Learning Disabilities, University of Haifa, IsraelcDepartment for Orthopedic Rehabilitation, The Rehabilitation Hospital, C. Sheba Medical Center, Tel Hashomer, IsraeldDepartment of Occupational Therapy, Faculty of Social Welfare and Health Sciences, University of Haifa, Haifa, IsraeleDepartment of Human Biology and Sagol Department of Neurobiology & Ethology, Faculty of Natural Sciences,University of Haifa, IsraelfFMRI Unit, Department of Diagnostic Radiology, C. Sheba Medical Center, Tel Hashomer, Israel

a r t i c l e i n f o

Article history:

Accepted 11 March 2015

Does the learning of a balance and stability skill exhibit time-course phases and transfer

limitations characteristic of the acquisition and consolidation of voluntarymovement sequences?

Available online 19 March 2015

Keywords:

Motor skill

Acquisition

Balance

Virtual reality

Time-course

1016/j.brainres.2015.03.02sevier B.V. All rights rese

hor. Tel: þ972 54 4688597s: [email protected] (O(P.L. (Tamar) Weiss), avi

a b s t r a c t

Here we followed the performance of young adults trained inmaintaining balance while standing

on a moving platform synchronized with a virtual reality road travel scene. The training protocol

included eight 3 min long iterations of the road scene. Center of Pressure (CoP) displacements

were analyzed for each task iteration within the training session, as well as during tests at 24 h, 4

weeks and 12 weeks post-training to test for consolidation phase (“offline”) gains and assess

retention. In addition, CoP displacements in reaction to external perturbations were assessed

before and after the training session and in the 3 subsequent post-training assessments (stability

tests). There were significant reductions in CoP displacements as experience accumulated within

session, with performance stabilizing by the end of the session. However, CoP displacements

were further reduced at 24 h post-training (delayed “offline” gains) and these gains were robustly

retained. There was no transfer of the practice-related gains to performance in the stability tests.

The time-course of learning the balance maintenance task, as well as the limitation on

generalizing the gains to untrained conditions, are in line with the results of studies of manual

movement skill learning. The current results support the conjecture that a similar repertoire of

basic neuronal mechanisms of plasticity may underlay skill (procedural, “how to” knowledge)

acquisition and skill memory consolidation in voluntary and balance maintenance tasks.

& 2015 Elsevier B.V. All rights reserved.

0rved.

.. Elion), [email protected] (I. Sela), [email protected] (I. Siev-Ner),[email protected] (A. Karni).

Fig. 1 – Within the training session and between-sessions gains.Data points represent the total CoP displacement in single roadscene iterations. Error bars - standard deviation. The grey linerepresents a power function regression fit of the performanceduring the training session. * � po0.05; ** � po0.01.

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 2 55

1. Introduction

Successful motor performance often requires the integration ofpostural adjustments, i.e., the ability to maintain and restorebalance, with voluntary, task-specific movements (Sela andKarni, 2012; Kurtzer et al., 2005; Keshner, 1994). This ability is ofparamount importance in determining the effectiveness ofrehabilitation programs after brain injury since balance is akey factor in enabling independent daily function (Vearrieret al., 2005).

There have been important advancements in our under-standing of the time-course (phases) and constraints on theacquisition of volitional, task oriented, motor skills in adults(Born, 2010; Stickgold and Walker, 2005; Robertson et al., 2005;Doyon and Ungerleider, 2003; Karni et al., 1998; Karni, 1996) butwhether these characterize the acquisition of balance skills aswell is not known. The learning of task-related movementroutines and specifically the generation of long-term proce-dural memory for the performance of task oriented movementsequences can be characterized by several distinct phaseswhich have been delineated in a number of laboratory tasks(e.g., Meital et al., 2013; Korman et al., 2007; Robertson, 2005;Stickgold and Walker, 2005; Sosnik et al., 2004; Maquet et al.,2003; Doyon and Ungerleider, 2003; Korman et al., 2003;Hikosaka et al., 1999; Karni et al., 1998; Shadmehr andBrashers-Krug, 1997). Rapid gains in performance occur earlyon in training (“fast learning”, novelty effect) but after a certainnumber of within-session task iterations, performance levelsoff if task conditions during the training session are unchanged(e.g., Adi-Japha et al., 2008; Korman et al., 2003; Hauptmannand Karni, 2002; Karni and Sagi, 1993). This plateau phase maybe followed by a latent phase when significant gains inperformance evolve. These delayed, ‘offline’, gains areexpressed hours after the termination of training and oftenrequire an interval of sleep to become effective (Born, 2010;Korman et al., 2007; Fischer et al., 2002; Walker et al., 2002;Karni et al., 1998). These delayed “offline” performance gainsreflect a procedural memory consolidation phase, i.e., theestablishment of improved, stable, task solution routines(Korman et al., 2003; Walker et al., 2003; Karni et al., 1998;Karni and Sagi, 1993). The consolidation phase often leads toknowledge that is less transferable across physical parametersof the task and training condition (Keetch et al., 2008; Kormanet al., 2003; Shadmehr and Brashers-Krug, 1997).

In the current study, we used an integrated system compris-

ing a virtual environment (VE) simulation of travel on a rough

and winding road, standing on a synchronized platform moving

in all three planes, to study balance maintenance learning.

Virtual reality systems afford interactive VEs in which the

intensity of practice and sensory feedback can be systematically

manipulated to provide individualized, real-life-like motor train-

ing situations and flexible, controlled settings for the study of

complex experiential learning (Rizzo and Kim, 2005). The main

goals were to characterize the phases in the acquisition of a

balance skill by healthy adults and its retention in long term

memory and to determine whether aspects of the postural skill

acquired in the VE were transferable to tests of balance and

stability, outside the context of the trained task conditions.

2. Results

2.1. Overall learning, between-sessions gains andretention

A single training session on the road scene VE resulted in robustwithin-session gains in stability, as well as delayed, post-sessiongains which were expressed at 24 h post training, and in robustlong-term retention of the gains as reflected in performance at 4and 12 weeks post training. These changes in performance wereexpressed by a significant reduction of the total CoP displace-ment across the 5 time-points (mean of 2 successive iterations;at session initiation, session final, 24 h, 4 weeks, 12 weeks post-training) (F(3.38,23.67)¼18.640, Po0.001) (Figs. 1 and 2). Tests com-paring the CoP displacement at pairs of time-points showed thatthere were significant reductions in sway not only within-session (P¼0.003) but also in the 24 h interval following thesession (P¼0.03) i.e., significant delayed gains. Also, compared toperformance at the end of the training session, the CoP displace-ment was reduced at 4 weeks (P¼0.03) and at 12 weeks (P¼0.015)after the training session, indicating robust retention not only ofthe within-session gains but of the gains achieved in the 24 hpost-training interval as well (i.e., no significant change in CoPdisplacement at 4 and 12 weeks compared to the 24 h post-training level) (Fig. 1).

2.2. Within-session and delayed gains expressed at 24 hpost-training

There was a significant reduction in the CoP displacementacross the 8 road scene iterations afforded in the trainingsession (F(1.86,13.03)¼21.204, Po0.001) (Figs. 1 and 2). A summaryof the main results of the statistical analyses is provided inTable 1. These incremental reductions in CoP displacementduring the training session were well fitted by a power function(R2¼0.89) (Fig. 1). The increase in skill was reflected in thereduction of the variability of balancemaintenance performanceas reflected in the CoP displacement during the training session.To this end, the individuals’ standard deviations (SD) of theCoP's distance from origin of axis (in the X and Z directions;right-left and anterior-posterior, respectively) at 1 s time bins

Fig. 2 – An example of the platform displacements (black) and the corresponding Centre of Pressure displacements (gray) inthe X axis (right� left) during the performance of the ‘road scene’ in a single participant. Movements are shown for a timewindow of 50–90 s from start, during the first (A) and final (B) task iterations of the training session.

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 256

across each of the task iterations was calculated. This measureof the variance of the Cop displacement was compared acrossthe successive training iterations. The variance was significantlyreduced along the X axis (right-left) but remained unchangedalong the Z axis (anterior�posterior) (Table 1).

The training session could be divided into two distinctphases along lines previously suggested in studies of volitionalskill acquisition (Logan, 1992; Hauptman and Karni, 2002;Korman et al., 2003; Hauptman et al., 2005; Adi-Japha et al.,2008): a fast learning phase consisting of the first four iterations,and a plateau phase, consisting of the last 4 iterations (Fig. 1).An analysis with iterations and phase (early, late) as withinsubject factors showed a significant phase effect (F(1,7)¼22.629,P¼0.002) as well as a significant phaseniteration interaction forthe total CoP displacement indicating (F(2,14)¼24.928, Po0.001),indicating that the rate of changes (reduction) in the total CoPdisplacement across iterations differed in the two phases(Table 1). Post-hoc comparisons confirmed that in the initialphase there was a clear reduction of the total CoP displacement(difference between 1st and 4th iterations, P¼0.011). However,stability leveled off in the later phase, with no significantdifferences between the total CoP displacements betweeniterations 5 and 8 (P¼0.98) (Fig. 1).

Although performance leveled off (i.e., attained a plateau)in the second half of the training session, there were addi-tional reductions in sway when participants were tested the

next day; CoP displacements were significantly reduced by24 h post training compared to the final iterations of the taskat the end of the training sessions (P¼0.03) (Fig. 1).

2.3. Transfer to stability tests

To assess the ability of the participants to transfer the gainsacquired in training to conditions outside the context of the trainedtask, the CoP displacements in the two quiet stance conditionswere compared using a repeated measures ANOVA with the 5time-points and the 2 stance conditions (sway open, sway closed)as within subject factors and group (training group, control group)as between subjects factor. There was no significant interaction ofconditionntime-pointngroup (F(2.83,39.56)¼0.432, P¼0.72). The CoPdisplacement significantly decreased across the five time-points(F(2.55,35.65)¼4.797, P¼0.009), however, therewas no significant groupeffect, and the interaction of groupntime-point was also notsignificant (F(2.55,35.65)¼0.269, P¼0.81) (Table 1).

In the perturbation tests, the CoP displacements in the fourperturbation directions were compared using a repeated mea-sures ANOVA with the 5 time-points and the four perturbationdirections ((forward, backward, right and left)) as within subjectfactors and group (training group, control group) as a betweensubjects factor. The CoP displacements were significantlydecreased across the five time-points (F(4,56)¼8.096, Po0.001);however, there was no significant group effect as well as no

Table 1 – Main results of the repeated measures ANOVAs and post-hoc tests, comparing performance across task iterationswithin-session and the ability to transfer the gains to untrained conditions at different time-points.

Condition Comparison F P

Within session training effects Within session gains (8 task iterations) F(1.86,13.03)¼21.204 o0.001Phase (within session 2 phases) F(1,7)¼22.629 0.002phaseniteration(within the training session) interaction F(2,14)¼24.928 o0.001Within session sd F(2.18,8.73)¼5.073 0.033right-to-left displacementWithin session sd F(1.37,6.84)¼1.205 n.s.anterior-posterior displacement

Transfer to Quiet stance (eyes open and eyes closed) Time-point (CoP displacement) F(2.55,35.65)¼4.797 0.009Group F(1)¼0.144 n.s.Condition F(1)¼0.007 n.s.Groupntime-point F(2.55,35.65)¼0.269 n.s.Conditionntime-pointngroup F(2.83,39.56)¼0.432 n.s.

Transfer to Perturbation test performance Time-points (CoP displacements) F(4,56)¼8.096 o0.001Group F(1, 14)¼0.426 n.s.Direction of perturbations F(3.42)¼27.780 0.001Groupntime-point F(4.56)¼0.821 n.s.Groupndirection F(3,42)¼0.462 n.s.Directionntime-pointngroup F(5.1,71.47)¼0.807 n.s.

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 2 57

significant interaction of groupntime-point nor of directionn-time-pointngroup (Table 1), (Fig. 3). There was a significantdifference in CoP displacements in the different directions ofthe perturbation (F(3.42)¼27.780, Po0.001). However, the interac-tion of groupndirection was not significant (Table 1) (Fig. 3).Thus, the results of both transfer tests indicated no significanttransfer of the gains acquired in training to conditions outsidethe trained task.

3. Discussion

Altogether, the results of the current study show that a singletraining session in balancing on a moving platform in thecontext of a VE was sufficient to trigger significant changes inbalance maintenance, a decrease in postural sway, expressed byreduced CoP displacements, not only within-session, i.e., con-currently with the accumulation of experience in the taskconditions, but also between-sessions, in young adults. Thechanges in sway were well retained across intervals of severalweeks, indicating the establishment of a well-maintained taskroutine. Thus, the training experience afforded during thesession was sufficient to trigger not only within-session gains,but also an ‘off-line’, post-training, phase of performancechange and long-termmemory and reflected in robust retentionover a three months interval.

The data indicate the existence of several distinct phasesin learning the balance maintenance task: an early phase ofrapidly accrued gains (fast learning) followed by a plateauphase during the initial training session, a between-sessionsphase of further changes in performance without additionalpractice, and long term retention of both the within-sessionand the between sessions gains. Similar phases have beendescribed and proposed as characteristic of skill (proceduralknowledge) acquisition and skill (procedural memory) con-solidation in the learning of perceptual as well as motor tasks

(Meital et al., 2013; Penhune and Steele, 2012; Dayan andCohen, 2011; Rozanov et al., 2010; Ari-Even Roth et al., 2005;Hauptmann et al., 2005; Stickgold and Walker, 2005; Kormanet al., 2003; Maquet et al.,2003; Karni et al. 1998; Karni andSagi, 1993). Thus, the current results lend support to thenotion of shared repertoire of basic neural mechanisms forprocedural memory consolidation underlying the acquisitionof experience-driven skill in multiple brain systems (Karniet al., 1998; Karni, 1996).

We propose therefore that the performance changes inducedby training in the current task represent a process of procedurallearning by the balance maintenance system, i.e., experiencedriven skill acquisition, specifically the setting up of a taskrelevant balancing routine. In line with this notion, the gains intask performance were not transferable to balancing perfor-mance in conditions/context unrelated to the trained condi-tions. Skill training often results in limitations on the ability totransfer the gains accrued in the training conditions to novel,untrained conditions; a limitation presumably reflecting thesetting up of task specific routines rather than general solutionswhen training conditions are kept constant (e.g., Censor, 2013;Rozanov et al., 2010; Breslin et al., 2010; Scholtz et al., 2009;Keetch et al., 2008; Sosnik et al., 2004; Korman et al., 2003;Hikosaka et al., 1999; Karni et al., 1998).

A decrease in CoP displacements does not necessarily meanbetter stability (Prado et al., 2007; Blanchard et al., 2005; Palmieriet al., 2002). Sway reduction may in some conditions reflectincreased stiffness that can impair postural adjustments (Rocchiet al., 2006). However, in these studies sway was considered as aparameter characterizing balancing performance at a certaintime-point in a given task or pathological condition (i.e., anindicator of ability and disability) rather than as a parameter ofchange across an extended learning experience, in a normalfunctioning population. In the context of the current studyconditions, the decrease in CoP displacement, with training, isconsidered a positive effect, i.e., an adaptive gain in performance

Fig. 3 – Performance in the perturbation tests (transfer tests). Perturbations were applied in the 4 cardinal directions. The CoPdisplacements of the Training Group (diamond) and the Control Group (circle) across the 5 assessments: (1st and 2nd – beforeand after the training session, respectively; 3rd, 4th and 5th � 24 h, 4 weeks and 12 weeks after the training session). Errorbars � group standard deviations.

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 258

triggered by training (Huys et al., 2004; Fransson et al., 2003). Animportant indication that balancing skills in the given taskconditions are increased is the finding that the variance in CoPdisplacement was decreased. Decreases in variance are consid-ered a hallmark of the setting up of consistent task solutionroutines and their “automatization” (Cohen and Sternad, 2009;Braun et al., 2009; Chein and Schneider, 2005). The reduction ofthe variance, in CoP displacement across the training iterationswhich was found in the X axis of CoP displacement only (not inthe Z axis) may support the notion that the balance and posturecontrol system may individually represent and control in theright� left (X) axis, presumably reflecting hip movements andseparately controls sway in the anterior�posterior (Z) axis,reflecting the ankle strategy (Winter et al., 1998; Winter, 1996).The results further suggest that training selectively improved thesub-system controlling sway in the right� left (X) axis, but hadlittle effect in the anterior�posterior (Z) axis; presumablybecause the more restricted movement repertoire of the anklesis well established in young healthy individuals, while hipmovements are amenable to undergo task related shaping(improved strategy).

In the context of the VR task used in the current study,there is no good reason to assume that the decrease in swaywith practice reflects an adverse effect of training rather thanthe acquisition and consolidation of a balancing task solutionroutine. It is also clear that the repeated (5 sessions) testingafforded in the perturbation tests (transfer conditions) led toa significant reduction in sway in both the training and thecontrol groups. Reductions in sway may characterize thecoping of young healthy adults with repeated patterns ofbalance perturbations (Van Ooteghem et al., 2008).

The current results support the notion that the time-course ofbalance task learning is similar to the time-course of voluntary,procedural skill acquisition. According to the general model ofprocedural, voluntary skill acquisition (Walker, 2005; Maquet et al.,2003; Karni et al., 1998; Karni, 1996), fast within-session gains inperformance reflect the setting of task specific processing rou-tines, including adaptation processes, trial-and-error learning andthe tuning up of the newly acquired routines, leading if sufficientpractice is afforded, to the attainment of a stable, effective taskperformance (Korman et al., 2003; Karni et al., 1998). The rapid

gains in performance within the training session can be consid-ered to be a marker for the novelty of the experience (Kormanet al., 2003). This novelty effect is supported by the evidence of anearly reorganization of the cortico-cortical network subtendingmotor-sequence learning across repeated practice during training(Maquet et al. 2003), and by the concurrent activation of cerebello-cortical and cortico-striatal networks (Orban et al., 2011; Doyonet al., 2003). The within session learning was well-fitted with apower functionmodel, a model that is considered characteristic ofskill learning (Adi-Japha et al. 2008; Chein and Schneider, 2005;Hauptman et al., 2005; Korman et al., 2003; Hauptman and Karni,2002; Logan, 1992). The initial fast learning phase was followed bya subsequent within-session leveling off of performance. A level-ing off of performance, i.e., the attainment of end-of-sessionplateau performance, was described in previous studies of voli-tional skill acquisition (Adi-Japha et al., 2008; Hauptman et al.,2005; Korman et al., 2003; Hauptman and Karni, 2002; Logan,1992). A further indication for skill acquisition was the decrease inperformance variability, a phenomenon that has been consideredto be a hallmark of the establishment of a stable movementroutine and automaticity in task performance ( Adi-Japha et al.2008; Chein and Schneider, 2005; Logan, 1988).

The expression of delayed gains in performance, as reflectedin task performance at 24 h post-training, has been ascribed tolatent memory consolidation processes whereby improved neu-ronal representations of the trained task are established and,presumably, structural synaptic changes are completed (Xuet al., 2009; Korman et al. ,2007; Walker, 2005; Monfils et al.,2005; Kleim et al., 2003; Maquet et al., 2003; Hess and Donoghue,1994; Karni and Sagi, 1993). The current results therefore, lendsupport to the notion that a latent memory consolidation phaseoccurs also in the acquisition of balance and posture main-tenance skills. Supporting this notion was the finding of robustretention of the gains, both those acquired during the trainingsession and the additional, delayed gains that were expressed at24 h post training. This pattern of long term retention of thegains is in line with previous data on volitional skill learning(Meital et al., 2013; Korman et al., 2003; Karni et al., 1998).

Altogether, the data from the current study suggest thatthe training protocol used in the current study was sufficient,to not only attain stable performance (a plateau phase) by the

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 2 59

end of the session but also to trigger between-session gains,that presumably reflect an effective memory consolidationphase (Walker, 2005; Maquet et al., 2003; Karni et al., 1998)and to generate long lasting procedural (“how-to”) memory(Wright et al., 2010; Robertson, 2009; Robertson et al., 2005;Sosnik et al., 2004; Korman et al., 2003; Maquet et al., 2003;Fischer et al., 2002; Karni et al., 1998, 1995).

The results of the current study, however, point to thelimits on the ability to transfer the gains acquired in the VEbalance task during the training session to performance oftests of balance and stability outside the context of thetrained task conditions. This limitation of transfer, i.e., thespecificity of the acquired task solution routine, not only toits physical parameters, but perhaps also to its context, is inagreement with previous studies (Keetch et al., 2008; Keshneret al., 2004; Korman et al., 2003; Karni et al., 1998). Thedecrease in the CoP displacements in response to the struc-tured perturbations in the perturbation tests was apparent inboth groups and thus can be ascribed to the repeatedexperience of the perturbations during the 5 tests (time-points) (Van Ooteghem et al., 2008). This pattern of resultsis in line with learning in the Finger Opposition Sequencetask, wherein the gains were consistently found to depend onthe order of the trained sequence (Rozanov et al., 2010;Korman et al., 2003; Karni et al. 1998), as well as with skilllearning in more complex tasks (Keetch et al., 2008; Keshneret al., 2004). Nevertheless, important aspects of a trainingexperience can transfer to novel task conditions, dependingon the training phase (e.g., Korman et al., 2003) and pre-sumably on the specific neuronal population that undergoesthe training related gains (Censor, 2013; Karni, 1996). Thespecificity of the acquired gains that was clearly indicated inthe transfer tests, suggests that balance training protocolsshould perhaps be much more effective for everyday activityif a variety of task conditions, rather than the performance ofa single, specific task condition, is trained (Hue et al., 2004);however, in tasks such as the finger opposition sequencelearning task, sequential training in multiple task conditionscan sometimes lead to interference effects (Brown andRobertson, 2007; Balas et al., 2007a, 2007b; Korman et al.,2007; Walker, 2005; Brashers-Krug, 1996).

The current results, from a small number of participants,both young and healthy, should not be generalized to differ-ent populations and especially to clinical populations beforeadditional studies on larger numbers of participants, andspecifically, older participants are undertaken. The currentstudy did not explore the learning of balance maintenance inreal life conditions. The role of visual and proprioceptiveinputs on balance control (e.g., Slaboda et al., 2009; Carveret al., 2006) as well as the effects of concurrent (multi) taskingas would occur in everyday situations (Blümle et al., 2006;Pelleccia, 2003) on the learning of the balance skill requirespecific testing. Also, the ecological validity, or the ability totransfer the gains acquired in the VR balance trainingexperience to real life conditions was not directly tested inthis study.

In conclusion, the results of the current study suggest acorrespondence between the time-course of learning andmemory consolidation in the acquisition of a voluntary motorskill and the time-course of the learning of a balance

maintenance skill. The study provides evidence for a memoryconsolidation phase (“off-line” learning phase) in balancelearning. The results also underscore possible limits on theability to transfer gains acquired in VE training conditions toother untrained conditions, in young healthy adults. Thus,the research paradigm developed in the current study affordsa new approach to studying the processes subserving thelearning of balance skills in a conceptual framework similarto that used in understanding volitional skill learning.

4. Experimental procedures

4.1. Population

Sixteen healthy participants were divided into two groups: atraining group (N¼8, male: female 1:1, mean age (7SD)¼26.4(71.7) years) and a control group (N¼8, male: female 1:1,mean age (7SD)¼30.0 (72.7) years). Due to factors related tolab availability, the lengthy preparation of each participantfor each session and the long protocol, the number ofparticipants per group was kept small. This is in agreementwith similar studies on skill acquisition (e.g., Katnak et al.,2010; Rozanov et al., 2010; Breslin et al., 2010; Rinaldi et al.,2009; Leonard et al., 2009; Karni et al., 1998, 1995). There wasno significant difference between the age of both groups(P¼0.17). Participants reported no history of disease or injuryto the central or peripheral nervous systems, or to themusculoskeletal system; none of them used medications thataffect the Central Nervous System (CNS) or motor perfor-mance. The study was approved by the ethics committee forhuman experimentation at the C. Sheba Medical Center.

4.2. Instrumentation

All experiments were conducted in a hospital-based motorperformance laboratory. The VE was generated and displayedusing the CAREN Integrated Reality System (ComputerAssisted Rehabilitation Environment) with CAREN III soft-ware (MOTEK BV, Amsterdam, The Netherlands, http://www.motekmedical.com). This system operates in real-time andenables the creation of a variety of controlled and repeatablesimulated environments with 3D visual, sound, and proprio-ceptive stimuli. It includes two force-plates embedded withina moving platform (AMTI http://www.amtiweb.com) that canbe manipulated with six degrees of freedom (Fig. 4A).

4.3. Procedures

Immediately before the training session, a 10-min structuredbattery of stability assessment was performed. Following a 5-minbreak, the participants in the training group performed thetraining task which consisted of eight iterations of a virtualscenario. To measure the consolidation and long-term retentionphases, participants performed two iterations of the training task24 h (Day 2), 4 weeks (Day 3) and 12 weeks (Day 4) post training.The stability test was repeated immediately after the trainingsession and again at 24 h (Day 2), 4 weeks (Day 3) and 12 weeks(Day 4) post training (before performing the training task. (Fig. 4B).The control group had no training of the balance task but

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 260

performed the same stability tests at the same time points as thetraining group. Instead of training (Day 1), the participants in thecontrol group sat on a stool on the stationary platform in the VRlaboratory for a period of time equivalent to the length of thetraining session.

4.4. Basic position for assessment

The participants stood at the centre of the platform with feetplaced apart by about a pelvis width (i.e., in the stancerecommended by Duarte and Freitas, 2010) and each foot placedon a separate force plate. The foot placing was clearly markedusing an adhesive ribbon. Participants were secured by aharness and two (slack, no movement restriction while upright)safety ropes. Participants were instructed to maintain the homeposition and to return to it if forced to move their feet by aplatform perturbation.

4.5. The training condition

A simulated road scenario was used for training. Participantswere required to maintain their balance while standing on theplatform (with the feet placed in marked positions) and “advan-cing” along a “road” that was displayed onto a screen(2.5�3m2, viewed from a distance of 2.4 m). The completion

Fig. 4 – Diagram of the CAREN system (A) and the overall designof the experiment (B). In the CAREN system data obtained by theforce-plates (and the motion capture system, data not reported)is transferred to the main computer which also controls themotion of the platform and the synchronized visual display;6DoF – 6 degrees of freedom. Time-course of the experiment:white bar � balance assessments (and perturbations, transfertests), black bar – road scene (VE) training session (day 1); graybar – road scene (VE) tests (days 2–4).

of each run (i.e., route along the road) took 2:48 min. Theplatform's movements were correlated with the visual stimuli.The “road” included flat, straight sections, sections with verticaldisplacements, sections with right and left tilts and sectionswith right and left translations. Maximal platform rotations onthe x-axis (pitch) were 3.41 forward and 4.21 backward. Themaximal translations on the x-axis (sway) were 12 cm to theright and 13 cm to the left. On the z-axis (roll) the maximumplatform tilt was 6.51 to the right and 8.21 to the left. On they-axis only translation movements were made (maximum,8 cm). These parameters were decided following several pilottests. The aimwas to optimize the parameters to be challengingyet not to cause excessive stepping. This task was sufficientlynew to participants and they could have not practiced it prior tothe study, and could have only practiced it in the lab, therefore,true learning could be investigated.

4.6. Stability tests

The structured tests of stability included: quiet stance stand-ing for 30 s with eyes open (sway open condition) and witheyes closed (sway closed condition) and a perturbations test.The perturbations test constituted of a set of 12 pseudo-random external perturbations realized by movements of theplatform at a velocity of 40 cm/s with a total displacement of10 cm over 0.25 s. After the perturbation, the platformremained stationary for 3 s and then slowly (2 s) returned tothe initial position. The set consisted of three perturbationsin each of four directions in the horizontal plane (forward,backward, right and left).

4.7. Performance measures

The outcome measures used to assess performance, learningand retention, were the total displacements of the CoP (x andy axes) in each task iteration. Since the parameters of theplatform displacements were identical for all participants,the total CoP displacement was calculated without subtract-ing the platform displacements.

The raw data were filtered and processed using MatLabsoftware tool kits (Version 2008a, The Mathworks Inc., Natick,MA � http://www.mathworks.com). A 4th order Butterworthlow pass filter with a cutoff frequency of 6 Hz was used.

4.8. Statistical analyses

To test for overall learning, including within training sessiongains and the gains achieved between sessions, five time-pointswere used as within-subject factors in a repeated measuresANOVA: the first two iterations (iterations 1 and 2) and the finaltwo task iterations (iterations 7 and 8) in the training session, aswell as the 2 iterations performed at 24 h (Day 2), 4 weeks (Day3) and 12 weeks (Day 4) post training. Post hoc tests were usedto test for performance changes between pairs of successivetime-points. To directly test for within-session gains, a repeatedmeasures ANOVA was used to compare the performancemeasures across the 8 task iterations within the session. Inaddition, to test for a performance stabilization phase, in thelatter half of the training session, performance in the four final

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 2 61

iterations was compared to performance in the four initialiterations of the session.

As an assessment of the ability to transfer the gainsacquired in training to conditions outside the context of thetrained task, the performance in the stability tests wascompared using a repeated measures ANOVA with 5 time-points as within subject factors (performance on Day 1, beforeand after training, Day 2 (24 h), Day 3 (4 weeks) and Day 4 (12weeks) after the training session) and group (training group,control group) as between subjects factor. The two quietstance conditions and in a separate analysis, the four per-turbation directions were compared as within-subject factors.

All statistical analyses were performed using SPSS (Ver-sion 19, SPSS Inc., Chicago IL). Whenever sphericity was notsupported, the Greenhouse-Geisser method was used tocorrect the degrees of freedom and determine the signifi-cance of results.

Acknowledgements

The study was part of a Ph.D. dissertation at the Center forAdvanced Technologies in Rehabilitation and the SagolDepartment of Neurobiology, University of Haifa (O.E.). O.E.'s current affiliations are the Gertner Inst. for Epidemiology& Health Policy Res, Tel Hashomer, and the Department ofPhysical Therapy, Ariel University, Israel. O.E. was supportedin part by the Sheba Medical Center, Israel as well as by aresearch grant from the Sagol Foundation (A.K.) and from theONO Academic Center, Israel. We thank MOTEK Medical, TheNetherlands, for the technical support.

r e f e r e n c e s

Adi-Japha, E., Karni, A., Parnes, A., Loewenschuss, I., Vakil, E.,2008. A shift in task routines during the learning of a motorskill: group-averaged data may mask critical phases in theindividuals’ acquisition of skilled performance. J. Exp.Psychol.: Learn. Memory Cogn. 34 (6), 1544–1551.

Ari-Even Roth, D., Kishon-Rabin, L., Hildesheimer, M., Karni, A.,2005. A latent consolidation phase in auditory identificationlearning: time in the awake state is sufficient. Learn. Memory12, 159–164.

Balas, M., Nester, S., Giladi, N., Karni, A., 2007a. Interference toconsolidation gains in learning a novel movement sequenceby handwriting: dependence on laterality and the level ofexperience with the written sequence. Exp. Brain Res. 180,237–246.

Balas, M., Roitenberg, N., Giladi, N., Karni, A., 2007b. whenpractice does not make perfect: well practiced handwritinginterfers with the consolidation gains in learning a movementsequence. Exp. Brain Res. 178, 499–509.

Blanchard, Y., Carey, S., Coffey, J., Cohen, A., Harris, T., Michlik, S.,2005. The influence of concurrent cognitive tasks on posturalsway in children. Pediatr. Phys. Ther. 17 (3), 189–193.

Blumle, A., Maurer, C., Schweigart, G., Mergner, T., 2006. Acognitive intersensory interaction mechanism in humanpostural control. Exp. Brain Res. 173, 357–363.

Born, J., 2010. Slow-wave sleep and the consolidation of long-term memory. World J. Biol. Psychiatry 11 (S1), 16–21.

Brashers-Krug, T., 1996. Consolidation in human motor memory.Nature 382, 252–255.

Braun, D.A., Aertsen, A, Wolpert, D.M., Mehring, C., 2009.Learning optimal adaptation strategies in motor tasks.J. Neurosci. 29 (20), 6472–6478.

Breslin, G., Hodges, N.J., Kennedy, R., Hanlon, M., Williams, M.A.,2010. An especial skill: support for a learned parametereshypothesis. Acta Psychol. 134, 55–60.

Brown, R.M., Robertson, E.M., 2007. Inducing motor skillimprovements with a declarative task. Nat. Neurosci. 10 (2),148–149.

Carver, S., Kiemel, T., Jeka, J.J., 2006. Modeling dynamics ofsensory re-weighting. Biol. Cybern. 95, 123–134.

Censor, N, 2013. Generalization of perceptual and motor learning:a causal link with memory encoding and consolidation.Neuroscience 250 (10), 201–207.

Chein, J.M., Schneider, W., 2005. Neuroimaging studies of practicerelated change: fmri and meta analytic evidence of a domaingeneral control network for learning. Cogn. Brain Res. 25 (3),607–623.

Cohen, R.G., Sternad, D., 2009. Variability in motor learning:relocating, channeling and reducing noise. Exp.Brain Res. 193(1), 69–83.

Dayan, E., Cohen, L.G., 2011. Neuroplasticity subserving motorskill learning. Neuron 72 (3), 443–454.

Doyon, J., Penhune, V., Ungerleider, L.G., 2003. Distinctcontribution of the cortico-striatal and cortico-cerebellarsystems to motor skill learning. Neuropsychologia 41 (3),252–262.

Duarte, M., Freitas, S., 2010. Revision of posturography based onforce plate for balance evaluation. Rev. Bras Fisioter. 14 (3),183–192.

Fischer, S., Hallschmid, M., Elsner, A.L., Born, J., 2002. Sleep formsmemory for finger skills. Proc. Natl. Acad. Sci. 99 (18),11987–11991.

Fransson, P.A., Johansson, R., Tjernstrom, F., Magnusson, M.,2003. Adaptation to vibratory perturbations in posturalcontrol. IEEE Eng. Med. Biol., 53–57.

Hauptmann, B., Karni, A., 2002. From primed to learn: thesaturation of repetition and the induction of long-termmemory. Cogn. Brain Res. 13 (3), 313–332.

Hauptmann, B., Reinhart, E., Stephan, A., Karni, A., 2005. Thepredictive value of the leveling off of within-sessionperformance for procedural memory consolidation. Cogn.Brain Res. 24 (2), 181–189.

Hess, G., Donoghue, J.P., 1994. Long-term potentiation ofhorizontal connections provides a mechanism to recognizecortical motor maps. J. Neurophysiol. 71 (6), 2543–2547.

Hikosaka, O., Nakahara, H., Rand, M.K., Sakai, K., Lu, X.,Nakamura, K., Miyachi, S., Doya, K., 1999. Parallel neuralnetworks for learning sequential procedures. Trends Neurosci.22 (12), 464–471.

Hue, O.A., Seynnes, O., Ledrole, D., Colson, S.S., Bernard, P.L., 2004.Effects of a physical activity program on postural stability inolder people. Aging Clin. Exp. Res. 16 (5), 356–362.

Huys, R., Daffrtshofer, A., Beek, P., 2004. Multiple time scales andmultiform dynamics in learning to juggle. Motor Control 7,188–212.

Karni, A., 1996. The acquisition of perceptual and motor skills: amemory system in the adult human cortex. Cogn. Brain Res. 5,39–48.

Karni, A., Meyer, G., Rey-Hipolito, C., Jezzard, P., Adams, M.,Turner, R., Ungerleider, L., 1998. The acquisition of skilledmotor performance: fast and slow experience-driven changesin primary motor cortex. Proc. Natl. Acad. Sci. USA 95,861–868.

Karni, A., Sagi, D., 1993. The time course of learning a visual skill.Nature 16, 250–252, http://dx.doi.org/10.1038/365250a0.

Karni, A., Meyer, G., Jezzard, P., Adams, M.M., Turner, R.,Ungerleider, L.G., 1995. Functional MRI evidence for adult

b r a i n r e s e a r c h 1 6 0 9 ( 2 0 1 5 ) 5 4 – 6 262

motor cortex plasticity during motor skill learning. Nature 377(6545), 155–158.

Katnak, S.S., Sullivan., K.J., Fisher, B.E., Knowlton, B., Winstein, C.J., 2010. Neural substrates of motor consolidation depend onpractice structure. Nat. Neurosci. 13 (8), 923–925.

Keetch, K.M., Lee, T.D., Schmidt, R.A., 2008. Especial skills:specificity embedded within generality. J. Sport Exerc. Psychol.30 (6), 723–736.

Keshner, E., 1994. Mechanisms controlling posture and balance.Phys. Ther. Pract. 3 (4), 207–217.

Keshner, E.A., Kenyon, R.V., Langston, J., 2004. Postural responsesexhibit multisensory dependencies with discordant visual andsupport surface motion. J. Vestib. Res. 14 (4), 307–319.

Kleim, J.A., Bruneau, R., Calder, K., Pocock, D., VandenBerg, P.M.,MacDonald, E., Monfils, M.H., Sutherland, R.J., Nader, K., 2003.Functional organization of adult motor cortex isdependent upon continued protein synthesis. Neuron 40 (1),167–176.

Korman, M., Doyon, J., Doljansky, J., Carrier, J., Dagan, Y., Karni, A.,2007. Daytime sleep condenses the time course of motormemory consolidation. Nat. Neurosci. 10 (9), 1206–1213.

Korman, M., Raz, N., Flash, T., Karni, A., 2003. Multiple shifts inthe representation of a motor sequence during the acquisitionof skilled performance. Proc. Natl. Acad. Sci. 100 (21),12492–12497.

Kurtzer, I., Herter, T.M., Scott, S.H., 2005. Random change incortical load representation suggests distinct control ofposture and movement. Nat. Neurosci. 8 (4), 498–504.

Leonard, J.A., Brown, R.H., Stapley, P.J., 2009. Reaching to multipletargets when standing: the spatial organization offeedforward postural adjustments. J. Neurophysiol. 101 (4),2120–2133.

Logan, G., 1988. Toward an Instance theory of automatization.Psychol. Rev. 95 (4), 492–527.

Logan, G.D., 1992. Shapes of reaction time distributions andshapes of learning curves: a test of the instance theory ofautomaticity. J. Exp. Psychol. 18 (5), 883–914.

Maquet, P., Laureys, S., Perrin, F., Ruby, P., Melchior, G., Boly, M.,Vu, T.D., Desseilles, M., Peigneux, P., 2003. Festina lente:evidence for fast and slow learning processes and a role forsleep in human motor skill learning. Learn. Memory 10,237–239.

Meital, N., Korinth, S.P., Karni, A., 2013. Plasticity in the adultoculomotor system: offline consolidation phase gains insaccade sequence learning. Brain Res. 1528, 42–48.

Monfils, M.H., Plautz, E.J., Kleim, J.A., 2005. In search of the motorengram: motor map plasticity as a mechanism for encodingmotor experience. Neuroscientist 11 (5), 471–483.

Orban, P., Peigneux, P., Debas, K., Barakat, M., Bellec, P., Benali, H.,Maquet, P., Doyon, J., 2011. Functional neuroanatomyassociated with the expression of distinct movementkinematics in motor sequence learning. Neuroscience 179,94–103.

Palmieri, R.M., Ingersoll, C.D., Stone, M.B., 2002. Center ofpressure parameters used in the assessment of posturalcontrol. J. Sport Rehab. 11 (1), 51–66.

Penhune, V.B., Steele, C.J., 2012. Parallel contributions ofcerebellar, striatal and M1 mechanisms to motor sequencelearning. Behav. Brain Res. 226, 579–591.

Prado, J., Stoffregen, T., Duarte, M., 2007. Postural sway duringdual tasks in young and elderly adults. Gerontology 53 (5),274–281.

Rinaldi, N.M., Polastri, P.F., Barela, J.A., 2009. Age-related changesin postural control sensory reweighting. Neurosci. Lett. 467,225–229.

Rizzo, A.A., Kim, G.J., 2005. A swot analysis of the field of virtualreality rehabilitation and therapy. Presence 14 (2), 1–29.

Robertson, E.M., Press, D.Z., Pascual-Leone, A., 2005. Off-linelearning and the primary motor cortex. J Neurosci. 25 (27),6372–6378.

Robertson, E.M., 2009. From creation to consolidation:a novel framework for memory processing. PLoS Biol. 7,e1000019.

Rocchi, L., Chiari, L., Cappello, A., Horak, F.B., 2006. Identificationof distinct characteristics of postural sway in parkinson’sdisease: a feature selection procedure based on principalcomponent analysis. Neurosci. Lett. 394 (2), 40–145.

Rozanov, S., Keren, O., Karni, A., 2010. The specificity of memoryfor a highly trained finger movement sequence change theending change all. Brain Res. 1331, 80–87, http://dx.doi.org/10.1016/j.brainres.2010.03.019.

Scholtz, J., Klein, M.C., Behrens, T.E.J., Johansen-Berg, H., 2009.Training induces changes in white-matter architexture. Nat.Neurosci. 12 (11), 1367–1368.

Sela, I., Karni, A., 2012. Differences in learning volitional (manual)and non-volitional (posture) aspects of a complex motor skillin young adult dyslexic and skilled readers. PLoS One 7,e43488.

Shadmehr, R., Brashers-Krug, T., 1997. Functional stages in theformation of human long-term motor memory. J. Neurosci. 17(1), 409–419.

Slaboda, J.C., Barton, J.E., Maitin, I.B., Keshner, E.A., 2009. Visualfield dependence influences balance in patients with stroke.Conf. Proc. IEEE Eng. Med. Biol. Soc., 1147–1150.

Sosnik, R., Hauptmann, B., Karni, A., Flash, T., 2004. Whenpractice leads to co-articulation: the evolution ofgeometrically defined movement primitives. Exp. Brain Res.156 (4), 422–438.

Stickgold, R., Walker, M.P., 2005. Memory consolidation andreconsolidation: what is the role of sleep?. Trends Neurosci. 28(8), 408–415.

Van Ooteghem., K., Frank, J.S., Allard, F., Buchanan, J.J., Oates, A.R., Horak, F.B., 2008. Compensatory postural adaptationsduring continuous, variable amplitude perturbations revealgeneralized rather than sequence specific learning. Exp. BrainRes. 187 (4), 603–611.

Vearrier, L.A., Langan, J., Shumway-Cook, A., Woollacott, M., 2005.An intensive massed practice approach to retraining balancepost stroke. Gait Posture 22 (2), 154–163.

Walker, M.P., Brakefield, T., Hobson, J.A., Stickgold, R., 2003.Dissociable stages of human memory consolidation andreconsolidation. Nature 425 (6958), 616–620.

Walker, M.P., Brakefield, T., Morgan, A., Hobson, A.J., Stickgold, R.,2002. Practice with sleep makes perfect: sleep dependentmotor skill learning. Neuron 35 (1), 205–211.

Walker, M., 2005. A refined model of sleep and the time course ofmemory formation. Behav. Brain Sci. 28, 51–104.

Winter, D.A., Patla., A.E., Prince, F., Ishac, M., Gielo-Perczak, K.,1998. Stiffness control of balance in quiet standing. J.Neurophysiol. 80, 1211–1221.

Winter, D.A., 1996. Unified theory regarding A/P and M/L balancein quiet stance 75 (6), 2334–2343.

Wright, D.L., Rhee, J.H., Vaculin, A., 2010. Offline improvementduring motor sequence learning is not restricted to developingmotor chunks. J. Motor Behav. 42, 317–324.

Xu, T., Yu, X., Perlik, A.J., Tobin, W.F., Zweig, J.A., Tennant, K.,Jones, T., Zuo, Y., 2009. Rapid formation and selectivestabilization of synapses for enduring motor memories.Nature 462, 915–919.