Influence of Stance Width on Frontal Plane Postural Dynamics and Coordination in Human Balance Control

Abstract

The influence of stance width on frontal plane postural dynamics and coordination in human bipedal stance was studied. We tested the hypothesis that when subjects adopt a narrow stance width, they will rely heavily on nonlinear control strategies and coordinated counter-phase upper and lower body motion to limit center-of-mass (CoM) deviations from upright; as stance increases, the use of these strategies will diminish. Freestanding frontal plane body sway was evoked through continuous pseudorandom rotations of the support surface on which subjects stood with various stimulus amplitudes. Subjects were either eyes open (EO) or closed (EC) and adopted various stance widths. Upper body, lower body, and CoM kinematics were summarized using root-mean-square and peak-to-peak measures, and dynamic behavior was characterized using frequency-response and impulse-response functions. In narrow stance, CoM frequency-response function gains were reduced with increasing stimulus amplitude and in EO compared with EC; in wide stance, gain reductions were much less pronounced. Results show that the narrow stance postural system is nonlinear across stimulus amplitude in both EO and EC conditions, whereas the wide stance postural system is more linear. The nonlinearity in narrow stance is likely caused by an amplitude-dependent sensory reweighting mechanism. Finally, lower body and upper body sway were approximately in-phase at low frequencies (<1 Hz) and out-of-phase at high frequencies (>1 Hz) across all stance widths, and results were therefore inconsistent with the hypothesis that subjects made greater use of coordinated counter-phase upper and lower body motion in narrow compared with wide stance conditions.

INTRODUCTION

The maintenance of stable human bipedal stance requires that the vertical projection of the body's center-of-mass (CoM) remains within the base-of-support (BoS). In freestanding humans, the BoS is defined by the area under and between the feet, and the CoM is determined by the orientation of individual body segments. To keep CoM motion within the BoS, humans can potentially use one or more of three basic strategies: 1) limit motion between individual body segments but regulate sway via nonlinear mechanisms that limit CoM sway by altering sensitivity to perturbations, 2) allow larger motion of body segments relative to one another, but in a coordinated manner so that the body's overall CoM motion remains small, and/or 3) increase the BoS by standing with feet farther apart, thereby permitting the CoM greater range of motion before reaching the limits of the BoS. Very little is known about how these strategies interact. The goal of this study is to understand how changes in the BoS in the frontal plane influence the way humans use the first two balance control strategies listed above.

In the first strategy, body segments move en bloc above the ankle joint with the upper body segment (UB; body mass located above the pelvis) aligned to the lower body (LB), analogous to the “ankle strategy” described for sagittal plane sway (Horak and Nashner 1986). Subjects using this strategy while responding to external perturbations could limit overall CoM motion by using a nonlinear control scheme that reduces the sensitivity of body sway responses to perturbations as the perturbation amplitude increases. Previous studies have shown nonlinear stimulus-response behavior whereby the stance control system reduced the influence of larger amplitude perturbations (Cenciarini and Peterka 2006; Maurer et al. 2006; Oie et al. 2002; Peterka 2002). This nonlinear stimulus-response behavior has been shown for visual tilt (Peterka 2002), surface tilt (Maurer et al. 2006; Peterka 2002), and external force (Maurer et al. 2006) stimuli that evoke sagittal plane sway, and for surface tilt (Cenciarini and Peterka 2006), visual translation (Oie et al. 2002), and tactile stimuli (Oie et al. 2002) that evoke frontal plane sway in narrow stance conditions. However, the influence of stance width on this nonlinear behavior has not been studied.

The second strategy is to use coordinated counter-phase motion of UB and LB segments to make rapid corrections in CoM position and to keep the overall CoM motion small relative to BoS even though the motion of individual body segments may be large. This strategy, termed the “hip strategy,” has long been recognized to contribute to stance control in conditions where the BoS is very narrow (Horak and Nashner 1986) and when very rapid corrections are necessary (Kuo 1995). The hip strategy can also be considered one of the normal “eigen” modes of control (Alexandrov et al. 2005).

Although it is convenient to describe the hip strategy separate from the ankle strategy, in fact, both are present simultaneously during unperturbed stance in the sagittal plane (Creath et al. 2005; Zhang et al. 2007). That is, at low frequencies, the UB and LB are nearly aligned and “in-phase” resembling the ankle strategy; however, at frequencies above ∼1 Hz, the UB and LB move in opposite directions and exhibit an “out-of-phase” hip strategy. This simultaneous in-phase and out-of-phase body segment motion is also present when subjects respond to an external surface (Alexandrov et al. 2005; Creath et al. 2008) or visual (Kiemel et al. 2008) stimulus in the sagittal plane. Similar dynamics have been found during spontaneous sway in the frontal plane (Zhang et al. 2007), but the role of stance width in shaping counter-phase behavior is unknown.

It is clear from previous studies that nonlinear stimulus-response behavior and coordinated counter-phase motion of body segments are important strategies to keep the CoM within the BoS in some conditions. However, humans normally maintain a wider stance in the frontal plane than those adopted in the previous studies (Cenciarini and Peterka 2006; Oie et al. 2002) and often increase stance width during everyday activities that challenge balance such as standing on a moving train. A few studies have investigated the role of stance width in the frontal plane on balance control. Increasing stance width has been shown to be associated with reductions in frontal plane spontaneous body sway (Day et al. 1993; Kirby et al. 1987), reductions in responses to galvanic stimulation of the vestibular nerve (Day et al. 1997; Welgampola and Colebatch 2001), and reductions in center of pressure motion, trunk motion, and muscle activation levels during sudden surface translations (Henry et al. 2001). Although these reductions in balance-related measures seem to be consistent across previous studies, the underlying cause of these reductions is unclear because the complex interaction between stance width in the frontal plane and neural control strategies for balance is still poorly understood.

Stance width in the frontal plane plays an important role by directly affecting the allowable range over which the CoM can move (Horak and Macpherson 1996), changing the mechanics of the LB, and modifying the proprioceptive sensory information available for balance control and the intrinsic mechanical properties of the LB by changing the relationship between LB sway and the stretching or shortening of muscles and tendons that span the hip joints (Day et al. 1993; Scrivens et al. 2008). Thus it seems reasonable to predict that changing stance width would influence the dynamic characteristics of the frontal plane balance control system.

In addition, changes in stance width dramatically affect the relationship between LB sway and the pelvis orientation in space. For example, when a subject's stance width is narrower than the distance between their hip joints, the pelvis always tilts in the same direction as the LB sways. However, as a subject progressively increases stance width, pelvis orientation transitions from being in the same direction as LB sway, to being invariant to, and then being in the opposite direction of LB sway. Because the control of UB orientation is influenced by pelvis orientation (Goodworth and Peterka 2009), changes in stance width are expected to have an impact on UB segment dynamics and thus influence the coordination strategy used for balance control.

In this study, we explore the influence of stance width on the dynamic response properties of frontal plane body sway by characterizing the linearity of responses to surface tilt stimuli of varying amplitude and by quantifying segment coordination. We test the hypothesis that when subjects adopt a small stance width, they will use a nonlinear control strategy and coordinated counter-phase motion of the UB and LB to control their CoM because stability requires minimal deviation of their CoM from upright and must therefore be tightly controlled. However, as the stance increases, CoM deviations from upright can be larger without jeopardizing stability. Therefore we also test the hypothesis that, as subjects adopt larger stance widths, nonlinear stimulus-response behavior and coordinated counter-phase motion will decrease. To test these hypotheses, we evoked body sway in the frontal plane through external rotations of the surface on which subjects stood with various stimulus amplitudes. Subjects were either eyes open or closed and adopted various stance widths.

If nonlinear control behavior is used to limit CoM excursions when stance width is narrow, we expect to see subjects become relatively less responsive to surface tilt stimuli as the stimulus amplitude increases. In contrast, as stance width increases, we expect that it would become less necessary for subjects to use a potentially more complex nonlinear control strategy to maintain balance because stimulus-evoked CoM excursions would never come close to the extended BoS afforded by wide stance. Therefore subjects could adopt a simpler linear control strategy. Additionally, if increasing stance width reduces the need for coordinated counter-phase motion of the UB and LB, we expect to see greater use of counter-phase motion in narrow stance to reduced CoM motion compared with wide stance conditions.

METHODS

Subjects

Eight healthy subjects (4 male, 4 female) with no history of balance disorders participated in this experiment. All subjects gave their informed consent before being tested using a protocol approved by the Institutional Review Board at Oregon Health & Science University. The subjects had a mean age of 29 ± 7 (SD) yr, mass of 69 ± 6.7 kg, height of 173 ± 6.0 cm, L4/L5 above surface of 106 ± 3.9 cm, greater trochanter above surface of 89.3 ± 4.9 cm, and distance between hip joint centers of 17.3 ± 1.2 cm.

Experimental setup

Body sway was evoked in the frontal plane in freestanding subjects through continuous rotations of the support surface (SS) on which subjects stood. SS rotations were controlled by a servomotor, and the rotation axis was horizontal and perpendicular to the subject's frontal plane at ankle height halfway between the subject's heels. Subjects faced a half-cylinder shape (70 cm radius) lined with a complex checkerboard pattern of white, black, and three gray levels that was illuminated by fluorescent lights attached to the right and left edges of the surround (Peterka 2002). For each test, subjects maintained a stance width of either 5, 12, 21.5, or 31 cm between the medial malleoli, defined as the intermalleolar distance (IMD). Because ankle geometry and foot width varied for each subject, IMDs corresponded to different BoS (distance between outside edges of the feet) for each subject. Specifically, IMDs of 5, 12, 21.5, and 31 cm corresponded to mean BoS widths across subjects of 24.0, 31.0, 40.5, and 50.0 ± 0.66 cm. For the remainder of this study, we refer to these stance widths as narrow, parallel, medium, and wide. The stance widths used in this study encompass a range comparable to those used previously (Day et al. 1993). Maintaining balance in narrow stance width during the largest perturbations was challenging but could be performed comfortably by all test subjects. In the parallel stance, subjects' mid-heel to mid-heel distance was approximately equal to the distance between hip joint centers. The wide stance width was limited by the dimensions of our platform, and the medium width was selected to be about halfway between parallel and wide. For comparison, a previous study showed that the preferred stance width is ∼17 cm between heel centers (McIlroy and Maki 1997), corresponding to an IMD of ∼11 cm.

Data collection

In all experiments, stimulus delivery and data sampling occurred at 200 Hz. Sampled data included SS angular position and potentiometer outputs (part CP-2URX-04, Midori America) that varied with lateral displacements of the upper trunk and pelvis. Lateral displacements were measured by means of “sway rods” that were attached to the rotation axis of the earth-fixed potentiometers. The other end of the sway rods rested on small lightweight metal hooks attached to the body. The UB hook was positioned on the midline of the trunk between the C6 and T3 vertebrae, and the LB hook was on the center of buttock at approximately the hip joint level. The sway rods could slide freely on the hooks. From the known positions of the potentiometers, the distances along the sway rods from the potentiometers to the hooks in the upright position, the potentiometer outputs, and height of the hooks relative to the assumed rotational axes of the upper and lower body segments, appropriate trigonometric conversions were used to calculate frontal plane linear displacements of the upper trunk and pelvis from rotational motion of each rod recorded by the potentiometer. Upper trunk and pelvis displacements were used to calculate angular displacements of the UB and LB with respect to earth-vertical (Fig. 1A). The UB sway angle was defined as the rotation angle about vertical calculated using the UB displacement at the UB hook with respect to an axis located midway between the hip joint centers (Goodworth and Peterka 2009; Seidel et al. 1995), and LB sway angle was defined as the rotation angle about vertical calculated using the LB displacement at the LB hook with respect to an axis located midway between the ankle joints.

Fig. 1.

Schematic of the 2-link representation of lower body (LB) and upper body (UB) sway. A: definition of LB and UB sway angles. B: validation of the sway measurement technique performed by manually tracking a continuous surface rotation with a rigid inverted-pendulum (IP). Measured LB and UB sway were essentially identical to IP sway across a wide bandwidth of frequencies.

In addition, whole body CoM displacements in the frontal plane were estimated using a biomechanical model that included two legs, one pelvis, and a detailed representation of the UB geometry that accounted for head and arm locations (Erdmann 1997). Body segment dimensions were estimated via anthropomorphic measures (Erdmann 1997; Winter 2005). Arms (crossed at waist level), head, and trunk were assumed to move together. LB segment orientations were based on a three-segment model of the LB (legs and pelvis) where leg sway angles were set equal to the LB sway angle, and it was assumed that no knee bending occurred.

The horizontal and vertical positions of the CoM of all body segments were calculated at each time step based on measured SS, LB, and UB angles. Whole body CoM displacements, calculated from the mass-weighted summation across all body segments, were converted to angular displacements to define the CoM sway angle as the rotation angle about vertical of the whole body CoM with respect to the axis located midway between the ankle joints. UB, LB, and CoM sway angles were considered to be the response variables, whereas SS rotation angle was considered to be the stimulus variable.

Our measurement technique was validated through the following procedure. A rigid inverted-pendulum (IP) was situated with its axis of rotation aligned with the surface rotation axis (Fig. 1B). A separate potentiometer was connected to the IP rotation axis to directly measure the IP sway angle. UB and LB sway rods with their associated potentiometers were used to measure UB and LB displacements and to calculate UB and LB sway angles in a manner identical to the methods used with test subjects. A continuous surface rotation was presented, and surface rotation and IP sway angles were shown on an oscilloscope. A manual tracking task was performed by the experimenter where the goal was to match the IP sway angle to the surface rotation. Times series of UB, LB, and IP sway angles were nearly identical, with R² values of 0.9995 and 0.9999 for the UB versus IP and LB versus IP regressions, respectively (Fig. 1B, middle column). In addition, a frequency domain analysis of LB and UB sway relative to IP sway was essentially identical, with a gain ratio of 1 and phase difference of 0 across a wide range of stimulus frequencies (Fig. 1B, right column; see Analysis for more details on frequency domain analysis).

External stimuli

SS rotational stimuli were presented continuously according to a pseudorandom waveform based on a pseudorandom ternary sequence (PRTS) of numbers (Davies 1970; Peterka 2002). Each number was assigned an angular velocity value of either +a, 0, or –a that was maintained constant for a specified state duration of Δt s. The angular velocity waveform was mathematically integrated to derive the angular position waveform. The angular position waveform was scaled to a specific peak-to-peak value for each test condition and was used to drive the SS rotation.

SS stimuli were created from a 2,186-length PRTS with 0.02 s state duration and cycle length of 43.72 s, giving a power spectrum of stimulus velocity with approximately equal amplitude spectral components ranging from 0.023 to ∼16.7 Hz. Seven PRTS cycles were presented in the lowest amplitude test, and six cycles were presented in the higher amplitude tests because responses to higher amplitude tests were large enough that low variance estimates of mean responses could be made with less averaging across individual stimulus cycles. A stimulus based on a PRTS was used because there are advantages to using periodic wide-bandwidth stimuli compared with random white-noise type stimuli for obtaining lower variance estimates of stimulus-response functions (Pintelon and Schoukens 2001).

Protocol

Subjects performed a total of 41 tests in randomized order to offset potential biases caused by fatigue and learning. All subjects were able to complete the tests in either three or four test sessions. Each test session was limited to 2.5 h and took place on a separate day. The 41 tests included 8 spontaneous sway tests (no SS stimulus 4 IMDs of 5, 12, 21.5, or 31 cm; EO or EC), 24 SS tests (3 amplitudes of 1, 2, or 4° peak-to-peak; 4 IMDs of 5, 12, 21.5, or 31 cm; EO or EC), and 9 tests where a visual surround stimulus was presented (data not presented in this study). Each test lasted ∼5.5 min, and subjects were given the opportunity to rest after every test.

Subjects were instructed to maintain straight knees throughout the test and to respond naturally. Subjects wore headphones and listened to their choice of novels or short stories to mask environment and equipment sounds and to maintain alertness.

Analysis

Experimental data were analyzed by calculating frequency-response functions (FRFs), coherence functions, and impulse-response functions (IRFs). FRFs enable the detection of frequency-dependent changes across test conditions, such as in-phase to out-of-phase transitions between the UB and LB, as well as the identification of changes in response sensitivity. IRFs enable the detection of time-dependent changes across test conditions, such as onset delays of sensory integration mechanisms. These analyses, which provide a linear analysis of the system dynamics under the given test condition, have been previously described in detail (Goodworth and Peterka 2009) and are briefly described below.

FRFs.

FRFs were defined as the ratio of the discrete Fourier transform of the response signal to the discrete Fourier transform of the stimulus signal (Pintelon and Schoukens 2001). FRFs were calculated for each stimulus cycle (except the 1st cycle to avoid transient behavior) and were smoothed by first averaging FRFs over the stimulus cycles and then averaging FRFs across adjacent frequency points. An increasing number of adjacent points were averaged with increasing frequency to reduce the variance of estimates at higher frequencies while maintaining adequate frequency resolution (Otnes and Enochson 1972). The final FRF estimates were approximately equally spaced on a logarithmic scale ranging from 0.023 to 5.9 Hz (the upper frequency range was limited by the signal-to-noise ratio of the experimental data).

Each FRF was expressed as a set of gain and phase values that vary with frequency. Each gain value indicates the ratio of the response amplitude to the stimulus amplitude at its particular frequency, and each phase value indicates the relative timing of the response compared with the stimulus (expressed in degrees and in most analyses was “unwrapped” using the “phase” function in the Matlab Signal Processing Toolbox, The MathWorks, Natick, MA). The stimulus signal was the SS angle and response signals were UB, LB, or CoM sway angles. Use of these angular response signals allows FRF measures to indicate the extent to which the UB, LB, and CoM aligned to the SS stimuli at any particular stimulus frequency. That is, a gain of 1 and phase of 0° at a particular frequency indicates perfect alignment to the stimuli with no lead or lag in timing.

COHERENCE FUNCTIONS.

Coherence functions measure the extent to which power in the sway response was linearly related with the power in the stimulus. Coherence function values vary from 0 to 1, with values of 1 indicating a perfect linear relationship between stimulus and response with no noise in the system or measurements. Coherence functions were defined as the squared magnitude of the stimulus-to-response cross-power spectrum divided by the product of the stimulus power spectrum and response power spectrum (Bendat and Piersol 2000). The cross-power and power spectra were calculated from the discrete Fourier transform of each stimulus cycle and were smoothed by averaging across individual stimulus cycles and by averaging adjacent frequency points. The coherence functions had the same frequency spacing as the FRFs.

IRFs.

IRFs were calculated for SS stimuli using an appropriately scaled cross-correlation between the ideal SS PRTS velocity waveform and the UB, LB, and CoM response waveforms (Davies 1970; Goodworth and Peterka 2009). The cross-correlation provides an IRF estimate because the PRTS velocity waveform is an approximate white noise stimulus (Davies 1970). The IRFs displayed in figures were convolved with a unit impulse of stimulus velocity and are shown with units of angular velocity.

The IRF of a linear system is the time domain equivalent of the frequency domain FRF (Davies 1970; Westwick and Kearney 2003). Although IRF and FRF representations of the system dynamics are equivalent for linear time-invariant systems, system properties are often easier to appreciate in one representation compared with another. For example, a time delay is easier to recognize in an IRF than an FRF where its effects are distributed.

PEAK-TO-PEAK AND ROOT-MEAN-SQUARE MEASURES.

Peak-to-peak (PP) and root-mean-square (RMS) measures of the LB, UB, and CoM were calculated for each subject and test condition. First, LB, UB, and CoM time series were averaged over the stimulus cycles (excluding the 1st cycle). Then, the RMS was calculated as the root-mean-square of the averaged and zero-meaned LB, UB, and CoM time series, and the PP was calculated as the maximum minus the minimum values of the averaged LB, UB, and CoM time series.

STATISTICS.

To test whether stimulus amplitude, visual availability (EO compared with EC), and stance width (IMD) had statistically significant effects on RMS sway of the LB, UB, and CoM, we used repeated-measures ANOVAs with three experimental factors: stimulus amplitude, visual availability, and stance width. Stimulus amplitude and stance width were continuous variables in the statistical model. Null hypothesis rejection was set to P < 0.05 for all tests. In addition, mean LB, UB, and CoM FRFs include 95% CIs that were determined using the percentile bootstrap method with 1,000 bootstrap samples (Zoubir and Boashash 1998), and 95% CIs on mean coherence functions were calculated according to Otnes and Enochson (1972).

RESULTS

Sway responses to SS stimuli

CoM SWAY.

SS stimuli and the CoM sway response for the three stimulus amplitudes, four stance widths, EO, and EC are shown in Fig. 2A. CoM sway waveforms (averaged over all subjects) generally followed the tilting SS stimulus, meaning that subjects tended to align to the SS. However, the extent of this alignment toward the SS depended on visual availability, stimulus amplitude, and stance width.

Fig. 2.

Center-of-mass (CoM) responses to support surface (SS) stimuli. A: average cycle of CoM sway obtained by averaging across subjects and across all individual stimulus cycles showed dependency on stance width [intermalleolar distance (IMD)], stimulus amplitude, and visual availability [eyes open (EO) or eyes closed (EC)]. B: root-mean-square (RMS) CoM sway as a function of stance width for each stimulus amplitude (mean ± SD). C: CoM RMS sway as a function of SS stimulus amplitude for each stance width (mean ± SD).

Increasing stance width generally reduced RMS sway at low stimulus amplitudes but increased RMS sway at higher stimulus amplitudes (Fig. 2B). Specifically, for RMS sways across combined EO and EC conditions, increasing stance width resulted in a statistically significant decrease in RMS sway during spontaneous sway (P < 0.01) and the 1° stimulus amplitude (P < 0.01), and a significant increase in RMS sway during the 2° (P < 0.01) and 4° stimulus amplitudes (P < 0.01; Fig. 2B). Visual availability (EO compared with EC) resulted in significant reductions in RMS sway during spontaneous sway (P < 0.01) and all stimulus amplitudes (P < 0.01). There was a significant interaction between visual availability and stance width at 1° (P = 0.016) and 2° (P = 0.042) stimulus amplitudes. These interactions related to the fact that visual availability reduced sway to a greater extent in the narrow and parallel stances compared with the medium and wide stances. Similar interaction trends were evident for spontaneous sway (P = 0.082; Fig. 2B, top) and 4° stimulus amplitude (P = 0.077; Fig. 2B, bottom), but these interactions were not statistically significant.

Increasing stimulus amplitude resulted in a significant increase in RMS sway at all stance widths (Fig. 2C; P < 0.01 at all stance widths). This increase in RMS sway with SS amplitude was approximately linear in medium and wide stances, whereas in narrow and parallel stances, RMS sway showed some tendency toward saturation. Across all stimulus amplitudes, visual availability resulted in significant reductions in RMS sway in narrow (P < 0.01) and parallel stances (P < 0.01) but not in medium (P = 0.32) or wide stances (P = 0.34; Fig. 2C). A significant interaction between visual availability and SS amplitude was present in narrow stance (P < 0.01; Fig. 2C, left). This interaction relates to the fact that increasing stimulus amplitude increased RMS sway to a lesser extent in EO compared with EC. A similar interaction trend was evident for the parallel stance condition, but this interaction was not statistically significant (P = 0.091).

LB AND UB SWAY.

UB and LB sway averaged across all subjects during the 4° SS test for narrow and wide stance are shown in Fig. 3A. UB and LB RMS sway as a function of SS stimulus amplitude, stance width, and visual availability is shown in Fig. 3, B and C. Variability in UB and LB RMS sway tended to be higher across subjects compared with CoM sway (cf. error bars in Figs. 2 and 3). LB and CoM RMS sway exhibited very similar patterns as a function of SS amplitude, stance width, and EO/EC (cf. Figs. 2B and 3B left column, and cf. Figs. 2C and 3C top row). For RMS sways across combined EO and EC conditions, increasing stance width resulted in a statistically significant decrease in LB RMS sway during spontaneous sway (P < 0.01), no significant effect in the 1° stimulus amplitude (P = 0.051), and a significant increase in RMS sway during the 2° (P < 0.01) and 4° (P < 0.01) stimulus amplitude (Fig. 3B, left column). Visual availability resulted in significant reductions in LB RMS sway during the 1° stimulus amplitude (P = 0.019), and there was a significant interaction between visual availability and stance width during the 1° stimulus amplitude (P = 0.013; Fig. 3B, left column).

Fig. 3.

LB and UB responses to SS stimuli. A: average cycle of LB and UB sway responses obtained by averaging across subjects and across all stimulus cycles for the 4° SS tests in narrow and wide stance conditions. B: LB and UB RMS sway as a function of stance width for each stimulus amplitude (mean ± SD). C: LB and UB RMS sway as a function of stimulus amplitude for each stance width (mean ± SD).

Increasing stimulus amplitude resulted in a significant increase in LB RMS sway in all stance widths (Fig. 3C, top row; P < 0.01 for all stance widths). For RMS sways across stimulus amplitudes, visual availability significantly reduced LB RMS sway in narrow (P < 0.01) and parallel stances (P = 0.045) but not in medium (P = 0.68) or wide (P = 0.75) stance widths (Fig. 3C, top row).

Patterns of UB RMS sway were different from LB and CoM. For RMS sways across combined EO and EC conditions, increasing stance width significantly reduced UB RMS sway during the 1° (P < 0.01), and 2° (P < 0.01) stimulus amplitudes (Fig. 3B, right column). Visual availability resulted in significant reductions in UB RMS sway during 1° (P < 0.01), 2° (P < 0.01), and 4° (P < 0.01) stimulus amplitudes (Fig. 3B, right column). There was no significant interaction effect between vision and stance width on UB RMS sway during any SS test condition.

Increasing stimulus amplitude resulted in a significant increase in UB RMS sway in all stance widths (P < 0.01; Fig. 3C, bottom). Visual availability significantly reduced UB RMS sway in all stance widths (P = 0.025 for narrow and P < 0.01 for remaining stance widths; Fig. 3C, bottom). There was a significant interaction between stimulus amplitude and visual availability in UB RMS sway in parallel (P < 0.01) and wide (P < 0.01) stance conditions (Fig. 3C, bottom). Interaction trends were also evident in narrow and medium but were not statistically significant (P = 0.11 in narrow and P = 0.064 in medium). Taken together, these results imply that subjects used visual information to reduce their UB sway in all stance widths. This result is in contrast to LB sway, where subjects only used visual information to reduce their LB sway in narrow and parallel stances.

CoM DISPLACEMENTS RELATIVE TO BASE-OF-SUPPORT.

To determine the extent to which subjects' CoM approached the limits of their BoS, PP CoM displacements were divided by BoS for each test condition (represented as white bars in Fig. 4). PP CoM displacements did not exceed 26% of the BoS on any test condition. Larger stimulus amplitudes were associated with larger PP CoM displacements, similar to RMS sway. Increasing stance width (and therefore also the BoS) resulted in lower PP CoM displacements relative to BoS. EO PP CoM displacements were lower than EC, especially in narrow stance.

Fig. 4.

Peak-to-peak (PP) CoM displacements divided by base-of-support (BoS). A: schematic of BoS, experimental CoM displacement, and hypothetical CoM displacement with UB and LB perfectly aligned. B: PP CoM displacements, where experimental CoM is represented as white bars and hypothetical CoM with UB and LB alignment represented as black bars. The gray bars represent a predicted linear increase in CoM displacement from 1 to 4° stimuli. Across-subject mean with ±SD shown for the experimental CoM displacements. Note that in this study, PP measures primarily reflect low-frequency behavior because most stimulus-response power was contained within the lower frequencies.

We quantified the role of body segment coordination in shaping PP CoM displacements. First, a hypothetical PP CoM displacement that would have occurred if the UB was perfectly aligned was calculated (represented as black bars in Fig. 4B). This hypothetical PP CoM displacement was compared with the experimental CoM displacement. Experimental CoM displacement would have been less than the hypothetical if either 1) the actual UB PP sway angle was less in magnitude than the LB PP sway angle or 2) the actual UB motion was out-of-phase with the LB so that peak UB sway did not occur at the same point in time as peak LB sway. In all test conditions, experimental CoM displacements were only slightly lower than those predicted if the UB was perfectly aligned. Specifically, in narrow stance, experimental CoM displacements were 91% in EC 1°, 97% in EC 4°, 90% in EO 1°, and 89% in EO 4° SS tests of the CoM displacements that would have occurred if the UB and LB were aligned. In wide stance, experimental CoM displacements were 91% in EC 1°, 89% in EC 4°, 86% in EO 1°, and 80% in EO 4° SS tests of the CoM displacement that would have occurred if the UB and LB were aligned.

In addition to coordination of body segments, PP CoM displacements could have been reduced if the responsiveness of the postural system diminished with increasing stimulus amplitude such that the sway responses did not scale linearly with SS amplitude. To quantify the role of a stimulus-response nonlinearity in limiting the PP CoM displacements with respect to BoS, we compared the experimentally measured PP CoM displacement on the 4° SS test to the linear prediction that CoM displacement would be 4 times larger on the 4° SS test compared with the 1° SS test (represented as gray bars in Fig. 4B). There were no conditions where experimental PP CoM displacements increased linearly from 1 to 4°. In narrow stance, experimental CoM displacements during 4° SS stimuli were much smaller than the linear prediction (white bars/gray bars = 51% in EC and 43% in EO). However, in wide stance, experimental PP CoM displacements during 4° SS were much closer to the linear prediction (87% in EC and 80% in EO) than in narrow stance.

It is important to note that PP measures described above primarily reflect sway behavior at low frequencies and are much less influenced by sway behavior at higher frequencies because most of the power in the integrated PRTS stimulus was contained within the lower frequencies (Goodworth et al. 2009; Peterka 2002) and because body inertia would tend to limit the amplitude of sway responses to higher frequency components of any surface stimulus.

Frequency-response analysis of sway responses

CoM ANALYSIS.

The variation across frequency of FRF gains was similar for all stance widths, stimulus amplitudes, and EO/EC conditions (Fig. 5). Gains in the 0.02- to 0.6-Hz range increased with increasing frequency to reach a peak value around 0.3–0.6 Hz. Gains decreased rapidly above 1 Hz, showed a minor peak or plateau around 4 Hz, and decreased again for frequencies >4 Hz. In frequency regions where gains were >1, subject CoM sway amplitude exceeded that of the SS stimulus.

Fig. 5.

CoM frequency-response functions (FRFs) and coherence functions averaged across all subjects showed that stimulus amplitude and/or visual availability influenced FRFs more in narrow compared with wide stance conditions. Error bars on FRFs show 95% CIs on mean gain and phase at each frequency.

Subjects exhibited amplitude-dependent changes in FRF gains in narrow and parallel stances, but amplitude-dependent changes were very minor in medium and wide stances. In narrow and parallel stances, increases in stimulus amplitude resulted in gain reductions at frequencies below ∼1–1.5 Hz. Similarly, visual availability resulted in gain reductions below ∼1–1.5 Hz for narrow and parallel stances but had little effect on gains in medium and wide stances.

All phase curves showed some phase lead relative to the SS stimulus at frequencies below ∼0.1 Hz. Phases generally decreased (more phase lag) with increasing frequency and exhibited a small peak or plateau around 3 Hz (Fig. 5). At each stance width, neither stimulus amplitude nor visual availability had a noticeable impact on phase curves. However, increasing stance width resulted in slightly reduced phase leads at frequencies <0.1 Hz, slightly less phase lag at frequencies between 0.1 and 3 Hz (and consequently an apparent reduction in the peak at 3 Hz). Above 3 Hz, phases did not change as a function of stance width.

All SS stimuli resulted in coherences between 0.6 and 0.98 at frequencies <1 Hz, and coherences decreased sharply at frequencies >1 Hz (Fig. 5). Increasing stimulus amplitude and increasing stance width generally resulted in higher coherences across all measured frequencies.

LB AND UB ANALYSIS.

For all EC tests, UB segment dynamics were similar across subjects. However, in five EO tests (2° and 4° narrow, 2° and 4° parallel, 2° medium), one subject exhibited a clearly different UB control strategy compared with the remaining seven subjects (see Fig. 11 for more detail). Therefore to make comparisons across test conditions, this particular subject's data were not included in figures describing UB and LB sway responses to SS stimuli (Figs. 6, 7, 8, and 10).

Fig. 11.

Example data from the single subject that used a counter-phase control strategy at low frequencies in several EO test conditions. Results of EO narrow UB analysis in the individual subject compared with mean results in the remaining 7 subjects are shown for (A) sway waveforms and (B) frequency-response functions.

Fig. 6.

LB FRFs and coherence functions for narrow and wide stance showed that patterns of LB FRFs were very similar to CoM FRFs (compare with Fig. 5). Error bars on FRFs show 95% CIs on mean gain and phase at each frequency.

Fig. 7.

UB FRFs and coherence functions. Phases were not “unwrapped” like LB and CoM phases in Figs. 5 and 6 because several UB FRFs did not vary smoothly across all stimulus frequencies. Consequently unwrapping algorithms did not perform consistently, thus giving a false impression that phases differed greatly between some trials. Error bars on FRFs show 95% CIs on mean gain and phase at each frequency.

Fig. 8.

Ratios of UB gain to LB gain and phase difference between UB and LB vs. stimulus frequency for both EO and EC conditions.

Fig. 10.

IRFs for (A) LB and (B) UB EC and EO conditions. Across-subject mean with ±SE shown for the 1° stimulus.

LB gains and phases were similar to CoM for all tests. Figure 6 shows LB gain and phase curves for the narrow and wide stance (cf. Fig. 5). In the narrow stance condition, there were LB gain reductions at frequencies below ∼1 Hz with increasing stimulus amplitude and in EO compared with EC conditions (Fig. 6A). In the wide stance condition, LB gains were nearly invariant across stimulus amplitude and were only slightly reduced for EO compared with EC in the 0.2- to 1-Hz range (Fig. 6B). LB phase curves were essentially invariant across stimulus amplitude and between EO/EC conditions. LB coherences were similar to CoM coherences in that LB coherences were generally high (>0.7) at low frequencies and decreased with increasing frequency (Fig. 6). At frequencies >1 Hz, LB coherences were higher than CoM coherences.

UB gains differed in several ways from LB and CoM gains. Most UB gain curves exhibited notches around 0.8–1 Hz, where gains were much lower than surrounding frequencies (Fig. 7). Low gains indicate that the UB was upright in space and therefore relatively unaffected by the SS stimulus. Increasing stance width generally resulted in lower and more constant gains at frequencies <0.8 Hz and larger gains between 2 and 6 Hz. Increases in stimulus amplitude were associated with gain reductions in all stances at frequencies <0.8–1 Hz, but these gain reductions were most systematic in narrow and parallel stances. EO UB gains were more variable than EC, and EO UB gains were lower than EC gains at frequencies below ∼4 Hz for the narrow stance and below 0.8–1 Hz for all remaining stance widths.

EC UB phases were similar across all stance widths and stimulus amplitudes. Specifically, EC UB phases were nearly in-phase with the SS stimulus at the lowest frequencies (<0.1 Hz) and declined (more lag) with increasing frequency until reaching a value near −180° at ∼0.8–1 Hz, which corresponds with the frequency where most gain curves showed a notch. The EC phase curves wrap to approximately +180° at ∼1 Hz and continue to show a monotonic decrease that reached approximately −140° at 5.9 Hz.

There were many similarities but also some differences between EC and EO UB phases. Across all stance widths and stimulus amplitudes, both EC and EO phases were close to 0° at frequencies <0.1 Hz and showed a monotonic decrease with increasing frequency at frequencies >2 Hz. On some generally lower amplitude trials (1 and 2° narrow stance and the 1° parallel, medium, and wide stances), EO and EC phase curves were very similar across all stimulus frequencies. However, on other generally higher amplitude trials, EO phases showed amplitude-dependent effects in the 0.3- to 2-Hz frequency range.

UB coherences were high at low frequencies, decreased with increasing frequency to a minimum around 0.5 at 1.3 Hz, increased with increasing frequencies up to a peak around 3–4 Hz, and decreased at frequencies >3–4 Hz (Fig. 7). The peaks and valleys in UB coherence curves coincide with peaks and valleys in UB gain curves, consistent with reduced signal-to-noise ratios in the experimental data at frequencies where gains were low.

RELATIVE LB AND UB MOTION.

UB gains relative to LB gains (i.e., the ratio of UB gain/LB gain) and UB phase relative to LB (i.e., the difference of UB-LB phase) are shown in Fig. 8. At the lowest stimulus frequency (0.023 Hz), UB gains were between 0.33 and 1.5 times LB gains. Above 0.023 Hz, in all stance widths, UB gains decreased relative to LB gains until ∼1 Hz, where a sharp transition occurred, and UB gains increased relative to LB gains. At frequencies >3 Hz, UB gains were 2.1–3.2 times larger than LB gains.

Increasing stimulus amplitude resulted in larger UB gains relative to LB gains at low stimulus frequencies (less than ∼0.5 Hz) in EC/EO narrow and EC parallel conditions. Visual availability had the most consistent effect on medium and wide stance conditions where EO UB gains relative to LB gains were 0.81 and 0.55 times EC averaged across frequencies <0.4 Hz in medium and wide, respectively.

On all EC tests, UB phase relative to LB showed that UB sway lagged the LB at frequencies less than ∼0.1 Hz, was approximately in-phase with the LB at ∼0.18 Hz, and transitioned to being out-of-phase (approximately −180°) with the LB at frequencies >2 Hz. Transitions from in-phase to out-of-phase behavior were similar across stance widths and were not influenced by stimulus amplitude on EC tests.

The relative phase on EO SS tests were similar to EC at low frequencies (<0.3 Hz) and at higher frequencies (>2 Hz), but in the frequency range between 0.3 and 2 Hz, transitions from in-phase to out-of-phase showed some dependence on stimulus amplitude and stance width. The most obvious difference between EO and EC phases occurred on the 4° SS tests, where the phase of UB relative to LB increased systematically with increasing frequency from 0.3 to 1 Hz before becoming out-of-phase (±180°) at higher frequencies.

Frequencies where the phase transitions occurred coincided with frequencies where UB gains relative to LB gains were lowest. UB gains relative to LB gains were highest when the UB moved 180° out-of-phase with the LB.

Time-domain analysis

Time-domain analysis of the dynamic responses to SS perturbations was accomplished by computing IRFs. Although continuous rotations were used to estimate IRFs, an IRF can be intuitively thought of as the time course of a subject's rotational sway velocity evoked by a sudden SS tilt of 1° (i.e., a velocity impulse). Negative IRF values indicate sway velocity in a direction opposite to the SS tilt, and positive IRF values indicate sway velocity in the same direction as the SS.

CoM IRFs.

Figure 9A shows the computed SS IRF velocity impulse, which includes the dynamics of the SS actuator and servo-control, and Fig. 9B shows the CoM IRFs. In all SS tests, CoM IRFs exhibited a very brief small negative peak at ∼0.05 s, followed by a sharp positive rise up to a first local maximum of 0.9–5.1°/s at ∼0.2 s, followed by a brief dip or plateau before another positive rise up to a second local maximum (usually corresponding to the overall IRF maximum) of 1.6–6.4°/s at ∼0.3–0.55 s. IRFs decreased to a minimum value of −0.3 to −2.3°/s at ∼0.8–1.6 s and finally decayed to ∼0°/s within 4 s.

Fig. 9.

CoM impulse-response functions (IRFs) from SS stimuli. A: SS tilt stimulus IRFs calculated between the ideal PRTS velocity and the actual support surface velocity. B: EC and EO CoM IRFs derived from the 3 different SS stimulus amplitudes. Across-subject mean with ±SE shown for the 1° stimulus. C: EC minus EO CoM IRFs averaged across subjects and stimulus amplitudes (mean ± SE) and displayed between 0 and 0.5 s to show time delay before onset of visual contribution, ∼0.15 s (vertical dashed line).

The detailed time course of IRFs depended on stance width, stimulus amplitude, and visual availability. CoM IRFs showed stance-width dependency after 0.05 s. Increasing stance width resulted in larger first local maximum values but did not affect the time needed to reach the first local maximum at ∼0.2 s. In contrast, the time needed to reach the second local maximum decreased as stance width increased (∼0.3 s for wide stance compared with ∼0.55 s for narrow stance). Also, increases in stance width were associated with IRFs that decreased in value faster after the IRF peak and reached a minimum in less time (∼0.8 s for wide stance compared with ∼1.6–1.8 s for narrow stance).

Stimulus amplitude-dependent changes were most pronounced in narrow and parallel stances. In narrow and parallel stances, increases in stimulus amplitude were associated with a slight decrease in CoM IRF magnitude beginning at ∼0.1 s (not clearly evident on the scale of Fig. 9B plots), followed by a larger decrease in CoM IRF magnitude beginning at ∼0.2–0.25 s.

The availability of vision affected CoM IRFs after a time delay of ∼0.15 s in all stance widths, where EO IRFs had lower magnitudes than EC (Fig. 9C). Differences between EO and EC CoM IRFs were greatest in narrow stance and least in wide stance conditions.

LB AND UB IRFs.

The overall shape of LB IRFs (Fig. 10A) was similar to CoM IRFs (Fig. 10B); however, LB IRF curves exhibited more pronounced features compared with CoM. Specifically compared with CoM, LB IRFs exhibited larger and longer initial negative peaks at ∼0.05 s, larger first local maximums of 1.6–8.5°/s at ∼0.2 s, larger second maximums of 1.8–7.3°/s at ∼0.3–0.55 s, and larger minimums of −0.31 to −2.7°/s at ∼0.8–1.6 s. Changes in stance width, stimulus amplitude, and visual availability had very similar effects on both LB and CoM IRFs.

The time course of UB IRFs differed from LB and CoM (Fig. 10B). UB IRFs began with a large initial positive peak of 4.2–11°/s at ∼0.05 s, followed by a negative peak of −1.3 to −7.7°/s at ∼0.2 s, followed by a positive rise up to a local maximum of 0.6 to 5.5°/s at ∼0.4–0.7 s, and eventual decay to 0°/s by 3 s. At times <0.25 s, the initial positive and negative peaks in UB IRFs generally coincided in time with, but were opposite in sign to, the initial negative and positive peaks in LB IRFs. This result means that, for the first ∼0.25 s after a sudden SS rotation, UB and LB sway velocities were opposite in direction in all SS tests.

Increasing stance width affected UB IRFs in several ways. Wider stances resulted in larger initial positive peaks at ∼0.05 s and larger negative peaks at ∼0.2 s. Increases in stance width were also associated with IRFs that reached local maximums with less delay (∼0.4 s in wide compared with ∼0.7 s in narrow) and local minimums with less delay (∼1 s in wide compared with ∼2 s in narrow). Increases in stimulus amplitude and visual availability were associated with lower UB IRF magnitudes evident in all stance widths. This result is different from LB and CoM IRFs where stimulus amplitude- and visual-dependent changes were more evident in narrow and parallel stance widths compared with medium and wide.

Counter-phase low-frequency UB sway in one subject

In five EO tests (2° and 4° narrow stance, 2° and 4° parallel stance, and 2° medium stance), one subject exhibited a clearly different control strategy at low frequencies compared with all other tests and compared with the remaining seven subjects. Figure 11 compares the UB sway of this subject to the mean results from the remaining seven subjects during the 4° EO SS test, narrow stance condition. Figure 11A shows that UB sway tended to be counter-phase with the SS in this particular subject, whereas UB sway tended to align toward the SS stimulus in the remaining seven subjects. The counter-phase UB sway in this particular subject had larger gains compared with the remaining seven subjects at the lowest stimulus frequency (0.023 Hz) and between 0.4 and 0.6 Hz, but at all other frequencies, gains were similar across all subjects (Fig. 11B). UB phases in this particular subject differed most from the remaining seven subjects at low frequencies, where phases were out-of-phase with the SS at the lowest stimulus frequency and monotonically decreased with increasing stimulus frequency to approximately −80° at 5.9 Hz (Fig. 11B).

DISCUSSION

Coordination strategy hypothesis

One potential strategy to limit CoM excursions from upright is to use coordinated counter-phase motion of UB and LB segments to keep the overall CoM motion small relative to the BoS even though the motion of individual body segments may be large. We tested the hypothesis that coordinated counter-phase motion of the UB and LB is used in narrow stance conditions to limit CoM displacements relative to the small BoS. Additionally, we hypothesized that this coordinated counter-phase strategy is used less at wider stance widths because CoM displacements can be larger without jeopardizing stability. If these hypotheses are true, we expected to see greater use of coordinated counter-phase LB and UB motion in narrow compared with wide stance conditions.

Coordinated counter-phase UB and LB motion and perfect alignment of the UB and LB represent two extremes on a continuum of possible coordination strategies. Perfect alignment means that UB and LB have identical gains and phases (in-phase with each other), whereas coordinated counter-phase sway means that UB and LB have opposite phases. Neither perfect alignment nor coordinated counter-phase sway occurred across all stimulus frequencies. Instead, general strategies emerged in a frequency-dependent manner.

At low stimulus frequencies (<0.5–1 Hz) and in narrow stance conditions, subjects' UB and LB motion was approximately in-phase (Fig. 8) and had similar gain values. At low stimulus frequencies (<0.5–1 Hz) and in wide stance conditions, subjects' UB and LB were also approximately in-phase, but UB gains were smaller than LB gains (Figs. 6 and 7), indicating less body segment alignment compared with narrow stance. At high stimulus frequencies (>1–2 Hz), UB and LB sway was approximately out-of-phase (Fig. 8) across all stance widths. Thus the hypothesis that coordinated counter-phase UB and LB motion is higher in narrow stance compared with wide stance was not supported.

There was one exception to the preceding conclusion. In several EO test conditions (generally conditions with small BoS and higher stimulus amplitudes), one subject exhibited UB sway that was out-of-phase with the LB across all stimulus frequencies (Fig. 11). Thus in select conditions, this subject used coordinated movements that were beneficial in the sense of reducing CoM sway. We do not have an explanation for this behavior and can only point out that visual availability must have had an important influence on the control strategy used by this subject, because the UB phase relative to LB in this subject was similar to all other subjects during EC tests. Previous studies have shown that numerous factors can influence the use of a hip versus an ankle strategy (Horak and Macpherson 1996) including availability of sensory information (Horak et al. 1990). It also seems plausible that other subjects might have used a counter-phase coordination strategy at low stimulus frequencies if sway was evoked with larger stimulus amplitudes than those used in this study.

Nonlinear stimulus-response strategy hypothesis

A second potential strategy to limit CoM excursions from upright is to use a nonlinear control scheme that reduces the sensitivity to perturbations of increasing amplitude. Overall results indicate that a nonlinear control strategy played an important role in limiting CoM excursions in narrow stance conditions. The role of this strategy decreased with increasing stance width, such that its contribution was very small in the wide stance condition. For example, in narrow stance, as stimulus amplitudes increased, and in EO compared with EC, there were CoM FRF gain reductions at frequencies below 1–1.5 Hz with minimal phase changes (Fig. 5). Furthermore, CoM gain reductions were accompanied by UB and LB gain reductions typically at frequencies <1 Hz, whereas UB and LB phases were approximately in-phase at corresponding frequencies (Figs. 6–8). Thus the amplitude-dependent nonlinearity present in narrow stance in both EO and EC conditions was not attributable to a change in coordination.

On the other hand, in both EO and EC wide stance conditions, CoM RMS sway and PP displacements (reflecting primarily low-frequency sway) increased nearly in direct proportion to the SS amplitude, but this nearly linear increase was not present in narrow stance (Figs. 2 and 4B). Furthermore, CoM FRF gains showed only minor reductions with increasing SS stimulus amplitude both in EO and EC wide stance conditions, but there were clear reductions in gains in narrow stance (Fig. 5). Similarly, there were only minor reductions in IRF peak amplitudes associated with increases in SS stimulus amplitude in wide stance IRFs compared with the much larger reductions in narrow stance (Fig. 9). Thus in wide stance conditions, subjects adopted a “simpler” linear control strategy compared with the nonlinear control present in narrow stance conditions.

Evidence for sensory reweighting

Because changes in stance width alter LB mechanics, it was not possible to determine what factors (i.e., mechanics and/or neural control) contributed to the changes in postural dynamics as a function of stance width. However, for a given stance width, LB mechanics are essentially fixed. Therefore the observed amplitude-dependent changes in FRFs at a given stance width must be caused by changes in the underlying control system. There is previous evidence that the nonlinear stimulus-response behavior, particularly prominent at narrower stance widths, is attributable to a sensory reweighting phenomenon.

There are two previous studies of frontal plane stance control where sensory reweighting was identified as a dominant contributor to the experimentally observed system nonlinearity (Cenciarini and Peterka 2006; Oie et al. 2002). The study of Oie et al. (2002) identified an amplitude-dependent nonlinearity in visually evoked sway when subjects maintained a very narrow stance (tandem-Romberg). They determined that their data were more accurately explained by a control system model that attributed the nonlinear behavior to an amplitude-dependent reweighting of sensory orientation cues rather than to a change in control properties (i.e., changes in the stiffness and damping of a neural controller).

The study by Cenciarini and Peterka (2006); which used varying amplitude pseudorandom SS rotations to evoke frontal plane sway in subjects standing with feet close together (∼10 cm IMD), is more directly comparable to this study. In addition to the SS stimulus, the previous study simultaneously presented a pulsed galvanic stimulation of the vestibular system that was mathematically uncorrelated with the SS stimulus. With increasing SS stimulus amplitude, subjects became more responsive to the galvanic stimulation and relatively less responsive to the SS stimulus exhibiting gain reductions with minimal change in phase at frequencies <1–2 Hz. The increased responsiveness to galvanic vestibular stimulation and the gain reductions and with increasing SS amplitude were attributed to a sensory reweighting mechanism whereby subjects shifted away from reliance on proprioceptive cues that orient the body to the SS and toward reliance on vestibular cues that orient the body more upright. The similarities between the narrow stance FRF results at frequencies <1–2 Hz in this and the previous study (Cenciarini and Peterka 2006) provide evidence that sensory reweighting contributed to a relatively lower responsiveness to SS stimuli as the SS amplitude increased. By extension to wider stance width conditions, where FRF gains showed less dependency on stimulus amplitude, the diminished amplitude-dependent behavior is consistent with a diminished role of sensory reweighting in compensating for perturbations of different amplitudes.

IRF results are also consistent with the interpretation that sensory reweighting contributed to the amplitude-dependent nonlinearity in narrower stance conditions. Because IRFs represent the time course of a subject's response to a sudden SS tilt, IRFs can be used to detect sensorimotor time delays in the balance control system. Sensorimotor time delays include the time needed to process sensory information, transmit signals through axons, synapses, and neuromuscular junctions and activate muscle contractions. If the relative reductions in CoM responsiveness to SS stimuli with increases in SS amplitude were caused by a neurally mediated sensory reweighting mechanism, we would expect to see changes in IRFs across SS amplitude after a time delay consistent with delays associated with sensory integration for balance control, considered to be ≥100 ms (Peterka 2002; van der Kooij et al. 1999). In contrast, IRF changes at shorter delays would be indicative of changes in muscle stiffness (possibly caused by co-contraction) or short-latency reflex activation.

In narrower stance conditions with both EO and EC, CoM IRFs showed minimal differences across SS amplitudes in the first ∼150 ms (Fig. 9B). After ∼150 ms, IRF peak amplitudes decreased with increasing stimulus amplitude. Thus after a time delay consistent with delays associated with sensory integration, subjects became less responsive to the SS stimulus as SS amplitude increased. The difference between the time courses of EO versus EC CoM IRFs is also consistent with visual availability contributing to changes in stimulus-response dynamics through a sensory reweighting mechanism. Specifically, visual availability altered the time course of the EO IRF relative to the EC IRF only after a delay of ∼150 ms (Fig. 9C).

Biomechanical contributions to UB IRFs

The time course of UB IRFs (Fig. 10B) gives insight into the mechanisms contributing to the control of UB orientation on the pelvis. The initial sharp rise in the UB IRFs occurred with no time delay, indicating a mechanical origin to this early feature of these IRFs. The biphasic time course over the first ∼0.2 s increased in magnitude with increasing stance width but was relatively unaffected by visual availability and stimulus amplitude. Short-latency reflexes, such as a stretch reflex, could have contributed to later portions of the initial biphasic time course (Cresswell et al. 1994; Skotte et al. 2005), but the invariance of the early portion of the UB IRF to stimulus amplitude indicates that any reflex contribution scaled proportionally with the stimulus amplitude. For example, in narrow stance, a sudden SS tilt of 1° would produce a sudden pelvis tilt that is <1°. However, as stance width increases, a sudden SS tilt of 1° would produce larger pelvis tilts, and the tilt would exceed 1° for medium and wide stance widths. Therefore intrinsic stiffness and reflexive mechanisms that tend to orient the UB perpendicular to the tilted pelvis would produce torque in proportion to the angle between the pelvis and UB, and this torque would increase as stance width increases. This larger torque at wider stances would produce a larger magnitude early time course of UB IRFs and the initial UB IRF deflection would be in the same direction of the tilted pelvis and SS. In addition, this sharp rise in UB velocity would influence the LB through an interactive torque that would act to move the LB in a direction opposite to the UB. This interaction torque likely contributed to the early negative peak in the LB IRFs (Fig. 10A).

UB and LB phase behavior

UB and LB phase results are in agreement with previous studies that have shown in-phase sway of the UB and LB at frequencies <1 Hz and out-of-phase sway >1 Hz during spontaneous sway in the sagittal and frontal plane (Creath et al. 2005; Horlings et al. 2009; Zhang et al. 2007) and during frontal plane balance responses to visual stimuli (Kiemel et al. 2008). The underlying cause of in-phase and out-of-phase sway in the UB and LB is still unknown. For example, loss of lower-leg proprioception influences the high-frequency out-of-phase sway behavior (Horlings et al. 2009), but there is also evidence that the out-of-phase behavior arises from the mechanical properties of the balance control system “plant,” which includes the musculoskeletal multisegmented body (Kiemel et al. 2008). We found that the overall feature of in-phase UB and LB sway at low frequencies and out-of-phase at high frequencies was present in all conditions of visual availability, stimulus amplitude, and stance width, consistent with the notion that out-of-phase behavior arises from the plant. However, if the plant produces out-of-phase behavior, it is surprising that the transition from in-phase to out-of-phase behavior was minimally influenced by stance width because changes in stance width do affect the mechanical properties of the system.

The transition from in-phase to out-of-phase behavior was influenced by sensory information and stimulus amplitude. Specifically, in both EC and EO tests, there was a transition from in-phase to out-of-phase behavior around 1 Hz (Fig. 8), but in EO tests, the UB-LB phase showed an increasing phase lead beginning at ∼0.2 Hz that was not present on EC tests. A previous study also found the UB leading LB at frequencies above ∼0.2 Hz using a sum-of-sines visual stimulus to evoke sagittal plane sway (Kiemel et al. 2008). The similarity between sagittal plane segmental motion evoked in the previous study (Kiemel et al. 2008) and frontal plane segmental motion evoked in this study suggests that common balance control strategies are used in all directions of sway when the BoS is small.

Finally, the transition from in-phase to out-of-phase sway was associated with a notch in the UB gain curve around 1 Hz (Fig. 7). This notch in UB gain predicts that UB orientation would remain relatively vertical in response to a 1-Hz stimulus. The low UB gain at 1 Hz may facilitate maintaining a vertical UB orientation during gait because natural gait generates an oscillatory frontal plane body sway at ∼1 Hz (one half the normal stepping rate of ∼2 steps/s; Bauby and Kuo 2000).

Limits of interpretation in this study

Because the CoM is determined by the orientation of individual body segments, understanding how individual body segments are controlled is necessary for a thorough understanding of balance control. Accurate conclusions about how individual body segments were controlled in this experiment cannot be made without accounting for the physical laws of multisegment motion where motion of one body segment generates interaction torques on adjacent body segments (Zajac and Gordon 1989). The interaction torques that arise during multisegment motion mean that sensory information used to orient one body segment will necessarily influence the orientation of other body segments. Therefore LB and UB sway responses in this study cannot be viewed as independent of each other. Interpretation of segmental behavior would benefit from a mathematical model that can represent sensory systems and capture behavior of multilink dynamics.

GRANTS

This work was supported by National Institute on Aging Grant AG-17960.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).