Abstract To assess vestibular inXuences on recovery ofbalance during sudden falls, we measured the posturalresponses of Wve healthy subjects to a hold and release per-turbation coupled with galvanic vestibular stimulation(GVS). Two electrode pairs were located with the anteriorelectrode of each pair over the mastoid process and the pos-terior electrode over the trapezius muscle on the same side.Bilateral unipolar GVS was generated 60 ms after a holdingforce against the sternum was released by individually driv-ing left and right electrode pairs as cathode or anode at1 mA for 12 s or 2 mA for 6 s. We computed the frequencyand damping parameters of a multi-link inverted pendulummodel of the body which best Wt the transient postural oscil-lations after release for each subject. These parameters didnot diVer signiWcantly across conditions indicating the GVSdid not modify the preset overall strategy of postural recov-ery. The intensity and polarity of GVS signiWcantly biasedboth the postural lean during the oscillatory period and theresting postural stance achieved during stimulation, deviat-ing them forward for cathodal stimulation and backward for
anodal. The residuals of the multi-link Wt, the frequencyspectra of the actual body sway ripples about the modeledsway, were diVerent across conditions. Because GVSaVected postural bias but not dynamics, it is likely that itprovided erroneous velocity signals leading to vestibulospi-nal compensations in segmental stiVness and dampingmechanisms. Our Wndings are consistent with theoreticalanalyses of the inXuence of GVS on the semicircular canalsand otolith organs of the inner ear.
Galvanic stimulation (GVS) has long been used as a tool tostudy vestibular function. In recent years, largely throughthe work of Day and Fitzpatrick and their colleagues, thevalue of GVS as a technique to activate selectively diVerentcomponents of the vestibular system while also varyingpostural, visual, and proprioceptive components of posturalcontrol and orientation has been amply demonstrated (Dayand Cole 2002; Day et al. 1997, 2002; Fitzpatrick and Day2004; Fitzpatrick et al. 1994, 1999, 2002). An importantaspect of their work is demonstrating that the postural andperceptual responses to GVS depend on head orientationrelative to the rest of the body and relative to gravity and onwhether the body is passively supported or actively stand-ing. For example, in the case of mechanically restrainedstance bilateral unipolar GVS, with anodes over both mas-toid processes, and cathodes placed anteriorly, will elicitapparent motion of the body in the posterior direction. Bycontrast, with unrestrained free stance the same stimulationwill lead to forward sway of the body.
S. B. Bortolami · S. Castellani · P. DiZio · J. R. Lackner (&)Ashton Graybiel Spatial Orientation Laboratory, MS 033, Brandeis University, P.O. Box 549110, Waltham, MA 02454-9110, USAe-mail: [email protected]
J. T. InglisSchool of Human Kinetics, The University of British Columbia, Vancouver, BC, Canada
P. DiZio · J. R. LacknerVolen Center for Complex Systems, Waltham, MA, USA
123
124 Exp Brain Res (2010) 205:123–129
Several aspects of GVS are important for its use as anexperimental tool: (1) the response to GVS does not habitu-ate or adapt, remaining comparably sized over repeatedapplications, (2) the EMG and postural responses to GVShave an early and late component, with the early responsebeing a step function and the later a ramp function that havebeen attributed to activation of otolith and semicircularcanal receptors, respectively, (3) GVS primarily aVects theactivity of muscles involved in active support of the bodyagainst gravity, and (4) GVS mainly aVects the dischargelevels of vestibular aVerent irregular Wbers. Importantly,GVS evokes not a change in the apparent vertical or per-ceived orientation of the base of support but rather in thesense of an unanticipated or unplanned head movement(Fitzpatrick and Day 2004). With the subject freely stand-ing, the postural response will be to tilt the head and bodyin the direction opposite the experienced head movement.
GVS has been used to evaluate the contribution of ves-tibular signals to step initiation and walking. During theactual initiation of a step GVS does not aVect center ofmass control or upper body orientation (Bent et al. 2002a).However, GVS presented during execution of a step duringlocomotion deviates body motion and center of mass trajec-tories (Bent et al. 2002b; Fitzpatrick et al. 2006, 1999).Accordingly, vestibular activation by GVS has a greaterinXuence during the dynamic phases of gait compared withstep initiation.
In the present study, we applied GVS during thedynamic phase of balance recovery following a destabiliz-ing perturbation. Our goal was to determine whether vestib-ular information aVects the regaining of balance from arecoverable “fall” induced by a hold and release (H&R)perturbation. In the H&R paradigm, a standing subjectactively resists a force applied at the sternum by the experi-menter (Bortolami et al. 2003). The subject has to innervatethe quadriceps femoris and tibialis anterior muscles (amongothers) to hold the body in place against the force at thesternum. When this force is suddenly removed or released,the body pivots forward because during the hold period thecenter of pressure of the feet is displaced toward the heelsand is behind the vertical projection of the center of mass.Consequently, after release, the body pivots forward at theankles until the muscle activations that were resisting thehold force are suddenly decreased and compensatory mus-cle activity occurs to arrest body motion and to restorestatic balance.
The advantage of the H&R paradigm is that it induces asudden forward fall about the ankles of the whole bodysimultaneously. We have previously shown that a multi-link model of the body with constant stiVness and dampingparameters can Wt the entire period of postural oscillationfrom release to the resumption of the Wnal settled posture(Bortolami et al. 2003). This pattern means that a postural
synergy is preselected and then implemented in a singlestep timed to occur so that the initial muscular forces brakethe fall and is maintained unchanged until oscillations aredamped and balance is regained. Combining GVS with theH&R perturbation provides a way to examine whether thispostural synergy can be modiWed during the dynamic phaseof recovering from a fall. GVS applied after initialization ofthe new synergy seemed an ideal way to test whether thesynergy can be disrupted because it inXuences the controlof muscles actively involved in postural control and thusdirectly implicates the leg muscles involved in recoveryfrom hold and release.
Fitzpatrick and Day (2004) have shown how diVerentpatterns of GVS can be used preferentially to bias vestibu-lar signals to inXuence balance control and locomotion. Ourgoal was to use GVS to aVect postural control in the sameanterior-posterior plane in which H&R perturbations causemotion of the body. To this end, we used bilateral monopo-lar GVS in which electrodes of the same polarity are placedon the mastoids on either side with the reference electrodeson the trapezius muscles. With this conWguration, cathodalGVS increases the activity of all responsive canal and oto-lith units irrespective of their directional sensitivity. Thenet result of the canal activation should be an almost com-plete cancellation of the signals from the anterior and pos-terior semicircular canals on each side of the head becauseGVS aVects them oppositely leaving at most a small pitchforward component. The net result for the otolith organs isa signal consistent with backward tilt of the body. Thus,cathodal stimulation after release should lead to an increasein the forward position of the body. Anodal stimulationshould have the reverse eVect. These calculations aredirectly based on the elegant and comprehensive analysesand model presented by Fitzpatrick and Day (2004).
The use of GVS allows us to test our hypothesis (Bortolamiet al. 2003) that the response to H&R is a synergy that isactivated on release and that remains virtually invariantuntil the body equilibrium is reset. Consequently, the tran-sient postural oscillations should reXect the dynamics of anattractor deWned by mechanical structure and constant jointstiVness and damping parameters. GVS which begins afterthe stiVness and damping parameters have already been setshould bias the attained equilibrium posture of the dynami-cal system deWned by a multi-link system with Wxed jointstiVness and damping. In other words, bilateral monopolarGVS should aVect the rest state the body returns to afterrecovery, displacing the body slightly forward for cathodaland backward for anodal stimulation relative to controlvalues. When GVS is terminated, body posture should driftto that of the control condition if the initially establishedsynergy is GVS-invariant.
An eVect of GVS on postural bias but not on posturaloscillations would also be consistent with the proposal put
forward by Feldman and Latash (2005) regarding how par-allel, independent reXexive mechanisms might inXuenceposture. The � equilibrium point hypothesis (Feldman1966) asserts that central commands specify a virtual equi-librium posture which depends on biomechanical andreXexive states such that muscle activation is generateduntil the actual and virtual postures match, within a thresh-old. Feldman and Latash (2005) stipulated that the Wnal set-tled posture can fall outside the window deWned by thecentral virtual posture if unexpected, non-central com-mands are issued. Our previous H&R studies support theview that postural oscillation and the associated patterns ofmuscle activation reXect Wxed attractor dynamics. Thus,introducing GVS after the reXex state has been set shouldbias the Wnal posture.
Materials and methods
Subjects
Five healthy men, ages 28–47, participated after providinginformed consent to a protocol approved by the BrandeisHuman Subjects Committee. All were physically active andwithout a history of skeletal, muscular, or inner ear diseaseor impairment.
Apparatus
An AMTI (model Z-ALS-600) force plate was used torecord the subject’s center of foot pressure (CP). AnOptotrak system (Northern Digital model 3020) was usedto monitor (sample rate = 200 Hz) with an accuracy ofapproximately 0.1 mm the locations of infrared emittingdiodes aYxed to the subject’s head, shoulder, hip, knee,and ankle. A custom-built, instrumented dynamometerprobe with a Xat contact surface was used to apply a forceat the subject’s sternum during the hold period and to sig-nal the exact time of release. EMG signals were recordedfrom the gastrocnemius, biceps femoris, rectus femoris,and tibialis anterior muscles using an MT8 EMG record-ing device (MIE Medical Research, UK). The signals wereband-pass Wltered, 10–500 Hz, and sampled at 1,300 Hz.GVS was delivered by means of left and right electrodepairs on the mastoid processes and on the trapezius mus-cles near the neck. The electrodes (Alimed Inc., 60091) onthe mastoids were trimmed to Wt and had a surface areaof + 12 cm2, the trapezius electrodes had an area of45.6 cm2. Stimulation was provided by two independentA-M systems Analog Stimulus Isolators (Model 2200)(one per ear) that were controlled by digital timing cir-cuitry. The signals were time synchronized and recordedon two PCs for later analysis.
Procedure
Subjects stood in stocking feet, eyes closed and their armscrossed, on the force platform in the standard Rombergposition feet side by side but not touching. The experi-menter pushed with a steady force against the subject’ssternum with a hand held dynamometer. The subjectactively resisted this force to maintain a straight uprightposture. The hold force was adjusted in magnitude fordiVerent sized subjects to elicit similar sway responses. Theexperimenter without warning within the next 10 s sud-denly withdrew the dynamometer probe thereby perturbingthe subject’s balance. The subject attempted to regain stablebalance as quickly as possible without moving either feet orarms.
Each subject participated on two separate days. On eachtest day, three conditions were run with four repetitions in abalanced order. The conditions included right and left elec-trodes on the mastoids driven as cathode, anode, or neutral(no stimulation) relative to a bilateral electrode pair on thetrapezii of the opposite polarity, or neutral. On one test day,stimulation was 1 mA for 12 s, on the other 2 mA for 6 s.These current values were chosen because they are levelsknown to aVect postural control. The order was alternatedacross subjects. In trials involving GVS, the stimulationwas initiated 60 ms after release. This brief delay periodwas adopted before biasing the vestibular activity becausewe wanted the body to be in the initial stage of a fall afterresponse selection but prior to the appearance of EMGactivity in muscles that could brake the forward motion ofthe body (cf. Bortolami et al. 2003).
Results and analysis
EMG activity
Figure 1 shows the performance of one representative sub-ject for 1 mA, 12 s duration stimulations. The EMGresponses to H&R are highly patterned. The rectus femorisand tibialis anterior which act to resist the hold force aredeactivated within 60 ms after release. The gastrocnemius-biceps femoris muscle pair, which brings the center of pres-sure forward, is activated about 50 ms later. This activationbrakes and reverses the forward motion of the body. It isthen followed by activity in the tibialis anterior-rectusfemoris pair which moves the center of pressure backward.We calculated the average muscle onset latencies afterrelease of gastrocnemius-biceps femoris and of tibialisanterior-rectus femoris EMG activations. The latency of theWrst pair was about 110 § 10 ms across conditions whilethe second pair averaged about 170 § 10 ms across condi-tions. As can be seen from Fig. 1, which is characteristic of
all subjects, the onset latency of EMG activity in responseto release varies little for the anodal, cathodal, and controlconditions. An ANOVA performed on the average onsettimes of the pairs across subjects indicated no signiWcantdiVerences (P < 0.05) between stimulation conditionsalthough there is a trend for cathodal GVS to delay theEMG bursts in the gastrocnemius and biceps femoris forboth GVS stimulation levels.
Postural sway
Sway traces for the head, shoulder, hip, and knee weresynchronized across diVerent repetitions of conditions usingthe release onset as the referent. To take into account diVer-ences in hold forces across diVerent trials, the sway traceswere normalized in relation to the center of foot pressurevariation, �CPx, from the hold period to after the release.Normalization is appropriate because the postural response tothe H&R perturbation is linear (cf. Bortolami et al. 2003).For the test conditions involving 2 mA, 6 s GVS stimulation(and control), �CPx averaged 49 § 5 mm, and for the 1 mA,12 s stimulation and control, 41 § 12 mm. A dynamic swaybias (D1) was determined for each condition (anodal, cath-odal, control) and subject. It was calculated as the meanvalue of normalized shoulder position within a 2.5 s (for2 mA, 6 s GVS) or 5 s (for 1 mA, 12 s GVS) window preced-ing the end of GVS, and in equivalent windows for the
control conditions. These time windows were chosenbecause in the absence of GVS, postural recovery from H&Rwould be complete before these time periods. Thus, any bias-ing eVect of GVS on postural stance can be determined.Figure 2 shows representative trials from one subject for2 mA, 6 s anodal and cathodal stimulations and a control trial.
D1 values were computed from normalized shoulder dis-placement values and thus are fractions of �CPx or purenumbers. For the 1 mA, 12 s condition, D1 averaged¡0.021 § 0.057, 0.017 § 0.022, and 0.110 § 0.078,respectively, for the anodal, control, and cathodal condi-tions. D1 values in the 2 mA, 6 s test conditions were¡0.165 § 0.154, 0.038 § 0.130, and 0.329 § 0.261 for theanodal, control, and cathodal trials, respectively. Pairwisecomparisons (Tukey–Kramer) indicated that D1 was sig-niWcantly more anterior in the cathodal trials than theanodal or control trials (P < 0.05), for both GVS amplitude/duration conditions.
The diVerences between conditions following recoveryof stable posture (with GVS still present) were also ana-lyzed in terms of whether the whole body was leaningbackward or forward in relation to the control trials (seeFig. 2). For 1 mA, 12 s stimulation, anodal versus controlwas ¡0.1° and cathodal versus control was +0.33°. For2 mA, 6 s GVS, anodal versus control was ¡0.59°, andcathodal versus control +0.25°. (At the level of the head,across subjects, a 1° displacement equals about 3 cm.)
Fig. 1 Representative EMG responses to H&R of one subject for the control condition and for anodal and cathodal GVS (1 mA, 12 s duration). GVS is initiated 60 ms after release from hold: the green and red arrows indicate where statistically signiWcant diVerences were Wrst detectable between anodal and cathodal GVS conditions relative to the control EMG traces. Across subjects, these diVerences are non-signiWcant
Anodal GVS thus led to a backward lean and cathodal aforward lean relative to the control trials.
We also computed the D1 values and the non-normalizedangular and linear displacements from the control posture ina 2.5 s (for 2 mA, 6 s GVS) or 5 s (for 1 mA, 12 s GVS)window starting 3 s after GVS oVset. There were no signiW-cant diVerences between both GVS conditions and the con-trol condition for any of the measures and stimulus proWles.
System identiWcation
A four degrees of freedom inverted pendulum model wasused to Wt average sway traces for each subject. The proce-dure and model are described in Bortolami et al. (2003). Foreach condition (cathodal, anodal, and control) and subject,the normalized head, shoulder, hip and knee displacementsfor the four repeated trials were averaged. The model param-eters that produced the best Wt were the frequency and damp-ing values of the Wrst and second modes. These values arepresented in Tables 1 and 2. ANOVAs and Friedman testsrevealed no signiWcant diVerences (P > 0.05; �2 = 0.17 and0.13 for 1 mA, 12 s and 2 mA, 6 s stimuli, respectively)between the parameters across conditions. This result meansthat the dynamic response of the body to H&R was statisti-cally equivalent across all of the conditions.
Residuals
Each model Wt produced four traces of “raw residuals”(diVerences between model Wts and sway traces), one each
for head, shoulder, hip, and knee. The residuals representthe discrepancy between the experimental data and thefour-link-pendulum approximation. Then, we determinedthe spectra for each residual trace (i.e., for head, shoulder,hip, and knee). The energies from each trace’s spectrumwere used to calculate an energy-weighted average spec-trum for each condition (anodal, cathodal, control) for eachGVS level and for each subject. Finally we calculated themedian of the spectral residuals across the subjects foreach type and level of stimulation. These residuals repre-sent a Wne-grained “roughness” superimposed on the main
Fig. 2 Representative postural sway responses for anodal and cathodal GVS (2 mA, 6 s dura-tion) and a control condition for one subject. The traces are nor-malized across subjects because the postural response to H&R is linear. GVS is initiated 60 ms after release from hold. D1 is the 2.5 s period prior to the end of GVS when posture has settled in the control condition and the eVect of GVS on the resulting posture of the body can be mea-sured. After GVS is terminated the body assumes the same pos-tural conWguration as present in the control condition
Table 1 Modal parameters identiWed from sway data of the diVerentconditions of stimulation at 1 mA for 12 s
12 s—1 mA Anodal Neutral Cathodal
�1 [Hz] 0.10 § 0.08 0.14 § 0.06 0.13 § 0.07
�2 [Hz] 0.39 § 0.09 0.40 § 0.09 0.28 § 0.12
�1 0.94 § 0.13 0.97 § 0.05 0.93 § 0.14
�2 0.52 § 0.25 0.59 § 0.12 0.67 § 0.30
Table 2 Modal parameters identiWed from sway data of the diVerentconditions of stimulation at 2 mA for 6 s
postural sway captured by the dynamics of the multi-linkinverted pendulum model representing the “ideal per-former”. To determine whether this signature changed sys-tematically as a consequence of GVS, we made pairwisecomparisons across conditions over the frequency span ofDC-2.5 Hz. For both the 1 mA, 12 s and 2 mA, 6 s GVS,the cathodal stimulation residuals were larger and statisti-cally diVerent from those of the anodal and control condi-tions (Wilcoxon, P < 0.001 all comparisons).
Discussion
Our goal was to test the hypothesis that GVS would bias theposition of the body during and following an H&R posturalperturbation but would not aVect the patterning of bodyoscillations and muscle activation producing recovery ofbalance. This hypothesis is based on the concept that theresponse to an H&R perturbation is the release of a synergyspecifying constant stiVness and damping parameters thatare maintained until stable upright posture is regained. Weinitiated GVS 60 ms after the hold force on the subject’ssternum was removed so that a forward “fall” of the bodywould be in progress and a braking synergy selected, butpostural muscles would not yet have been innervated toarrest and reverse the impending body displacement.Fitzpatrick et al. (1994) found that for standing subjectsGVS elicits both short latency, +60 ms, and medium latency,+110 ms, EMG responses that are of opposite sign. Themedium latency but not the short latency responses evokepostural alterations with sway in the direction of the anodalelectrode. These actions are well within the period—whichlasts several seconds—before the body is stabilized follow-ing an H&R perturbation. An important feature of the H&Rperturbation is that it aVects all segments of the body simul-taneously rather than sequentially as is the case with pertur-bations delivered by moving the stance base (Bortolamiet al. 2003).
We found no signiWcant diVerences in modal frequenciesand damping parameters across the GVS and control condi-tions, nor across the EMG latencies. The lack of alterationsin modal behavior means that GVS stimulation did not dis-rupt the multi-link inverted pendulum character of the H&Rpostural response even though it occurred in time to inXu-ence that aspect of recovery. GVS did aVect the centroidaround which postural oscillations occurred and the restingpostural vertical attained after H&R, biasing these positionsslightly backward for anodal stimulation and forward forcathodal. This signiWcant change in the equilibrium point ofposture during GVS and the lack of signiWcant change inthe equilibrium point of posture after GVS oVset are consis-tent with the proposal that GVS generates both a stepresponse and a ramp response, with the former arising from
the pars medialis receptors of the utricles and the latterfrom the semicircular canals (Fitzpatrick and Day 2004).This pattern thus accords with a preset centrally com-manded synergy specifying postural equilibrium positionsand recovery dynamics and with GVS eliciting posturalcompensations that act in parallel on the equilibriumposition.
Our analysis of the residuals after model Wts shows thatGVS also aVected the Wne-grained dynamics embedded inthe slower postural oscillations during recovery fromrelease. A static, unidirectional bias of the gravity referencesignal, the “step signal” from the utricles should not aVectthe smoothness of the postural dynamic response. Such asignal would arise from the semicircular canals, the rampresponse described by Fitzpatrick and Day (2004). OurWndings are in accord with studies cited in the “Introduc-tion” showing that GVS has little inXuence on the initiationof gait but does inXuence the trajectory of ongoing bodydisplacement through space. The entire pattern of ourresults is consistent with the analysis presented by Fitzpatrickand Day (2004) of GVS inXuences on semicircular canaland otolith receptors and their consequences for posturalcontrol. They also follow the view that GVS acts in parallelwith � equilibrium point control of posture, as described byFeldman and Latash (2005).
Acknowledgments Support was provided by NSF grants 0815577and BCS-0925878 (JRL) and NSERC (JTI). We thank Prof. BradfordJ. McFadyen for his helpful advice.
References
Bent LR, Inglis JT, McFadyen BJ (2002a) Vestibular contributionsacross the execution of a voluntary forward step. Exp Brain Res143:100–105
Bent LR, McFadyen BJ, Inglis JT (2002b) Visual-vestibular interac-tions in postural control during the execution of a dynamic task.Exp Brain Res 146:490–500
Bortolami SB, DiZio P, Rabin E, Lackner JR (2003) Analysis ofhuman postural responses to recoverable falls. Exp Brain Res151:387–404
Day BL, Cole J (2002) Vestibular-evoked postural responses in theabsence of somatosensory information. Brain 125:2081–2088
Day BL, Severac Cauquil A, Bartolomei L, Pastor MA, Lyon IN(1997) Human body-segment tilts induced by galvanic stimula-tion: a vestibularly driven balance protection mechanism.J Physiol 500(Pt 3):661–672
Day BL, Guarraz M, Cole J (2002) Sensory interactions for humanbalance control revealed by galvanic vestibular stimulation. AdvExp Med Biol 508:129–137
Feldman AG (1966) Functional tuning of the nervous system duringcontrol of movement or maintenance of a steady posture. III.Mechanographic analysis of the execution by man of the simplestmotor tasks. Biophysics 11:766–775
Feldman AG, Latash ML (2005) Testing hypotheses and the advance-ment of science: recent attempts to falsify the equilibrium pointhypothesis. Exp Brain Res 161:91–103
