Postural adjustment errors reveal deficits in inhibition during lateral step initiation in older adults

Abstract

Postural dual-task studies have demonstrated effects of various executive function components on gait and postural control in older adults. The purpose of the study was to explore the role of inhibition during lateral step initiation. Forty older adults participated (range 70–94 yr). Subjects stepped to the left or right in response to congruous and incongruous visual cues that consisted of left and right arrows appearing on left or right sides of a monitor. The timing of postural adjustments was identified by inflection points in the vertical ground reaction forces (VGRF) measured separately under each foot. Step responses could be classified into preferred and nonpreferred step behavior based on the number of postural adjustments that were made. Delays in onset of the first postural adjustment (PA1) and liftoff (LO) of the step leg during preferred steps progressively increased among the simple, choice, congruous, and incongruous tasks, indicating interference in processing the relevant visuospatial cue. Incongruous cues induced subjects to make more postural adjustments than they typically would (i.e., nonpreferred steps), representing errors in selection of the appropriate motor program. During these nonpreferred steps, the onset of the PA1 was earlier than during the preferred steps, indicating a failure to inhibit an inappropriate initial postural adjustment. The functional consequence of the additional postural adjustments was a delay in the LO compared with steps in which they did not make an error. These results suggest that deficits in inhibitory function may detrimentally affect step decision processing, by delaying voluntary step responses.

the relationship between impaired executive function and reduced gait speed, reduced balance function, and falls has been well illustrated in large-scale epidemiologic studies (Anstey et al. 2006, 2009; Atkinson et al. 2007; Ble et al. 2005; Hausdorff et al. 2008; Holtzer et al. 2007; Inzitari et al. 2007a, 2007b; Rosano et al. 2005; Watson et al. 2010). The findings in these correlational studies have been supported by a multitude of experimental studies that have used dual-task methodologies to understand how balance performance changes when a person is engaged in another activity. Most of these studies can be classified as assessment of how divided attention influences balance performance. Attention is required for the preparation, maintenance, and processing of effortful tasks and for the coordination of multiple simultaneous tasks (Hartley et al. 1992). Different cognitive tasks have been used in the dual-task paradigms during standing and walking, including mental arithmetic (Brown et al. 1999; Stelmach et al. 1990), visuospatial tasks (Andersson et al. 1998; Kerr et al. 1985), reaction time tasks (Marsh and Geel 2000; Redfern et al. 2001; Shumway-Cook and Woollacott 2000), and word recall (Lindenberger et al. 2000). Normal aging appears to result in increased attentional requirements for balance during standing and walking (Brown et al. 1999; Chen et al. 1994, 1996; Ebersbach et al. 1995; Lajoie et al. 1993; Lindenberger et al. 2000; Lundin-Olsson et al. 1997; Maylor and Wing 1996; Means et al. 1998; Redfern et al. 2001; Shumway-Cook et al. 1997; Stelmach et al. 1989, 1990; Teasdale et al. 1991, 1993; Teasdale and Simoneau 2001).

Although the relationship between executive function measures and balance function is well established, it is not clear whether specific components of executive function have a more important relationship to balance than others. A component of executive function that may play a key role in balance and gait is inhibition. By definition, attention creates a focus on some information or action to the exclusion of others, that is, inhibition. For example, Redfern et al. (2009) found that increased magnitude of sway generated by older adults during conditions of sensory conflict (specifically degraded somatosensory feedback) was correlated with increased difficulty on an inhibition task, specifically perceptual inhibition (PI). However, no such correlation was found in young adults (Redfern et al. 2009). Inhibition has also been related to risk of falling. A recent prospective study found an association between fall incidence and performance on response inhibition tests (Stroop Color-Word test and Go-No-Go test) (Herman et al. 2010; Mirelman et al. 2012). In another study, differences between nonfallers and recurrent fallers were explained, in part, by delays in choice reaction time performance and increased errors during inhibition tests (Anstey et al. 2009). Additionally, differences in activation of brain regions responsible for inhibition control have been related to changes in fall risk (Nagamatsu et al. 2011). These associations between measures of inhibitory function and balance/gait outcomes suggest that inhibition does play a role in maintaining postural control, particularly in older adults.

Step initiation is a functional task that is also believed to require inhibitory function. Inhibition of inappropriate motor responses has been shown to be involved in step initiation in older and young adults (Cohen et al. 2011). During choice reaction time conditions of a step initiation task, older adults had a threefold greater chance of producing an error in selecting the correct motor program for stepping when the stimulus was not predictable. Furthermore, young and older subjects who made more errors had worse performance on the Stroop Color-Word test. Thus step initiation may provide a robust method for assessing the role of inhibitory function in the control of posture in older adults. Clinically normal older adults have increased step initiation times compared with young adults (Luchies et al. 2002; Melzer and Oddsson 2004; Rogers et al. 2001, 2003). Furthermore, older adults with balance impairments and higher fall risk have greater step initiation times compared with older adults without balance impairments (Lord and Fitzpatrick 2001; Medell and Alexander 2000; St George et al. 2007). The time to make a step increases further when there is uncertainty about the direction of the step, or when a dual task (such as a Stroop task, nonspatial or spatial working memory task) is simultaneously performed (Lord and Fitzpatrick 2001; Luchies et al. 2002; Melzer and Oddsson 2004; Patla et al. 1993; Rogers et al. 2003; St George et al. 2007; Sturnieks et al. 2008). In addition, the ability to step quickly is a critical factor in arresting a fall (van den Bogert et al. 2002).

Consequently, we have developed an inhibition test of step initiation that parallels a manual reaction time test of PI and motor inhibition (MI) (Germain and Collette 2008; Jennings et al. 2011; Nassauer and Halperin 2003). The primary purpose of the experiment was to examine, in older adults, the effects of inhibition on several measures of step initiation function obtained through analysis of vertical ground reaction forces (VGRF), including the latency of postural adjustments and foot liftoff. We developed this paradigm of this test to probe inhibition through the known tendencies of the motoric responses to be directed toward location of visuospatial attention (Simon 1969). By manipulating the locus of visuospatial attention with conflicting directional cues, the task requires inhibition of a motor response toward an irrelevant directional cue. Incorporating this concept into a step initiation test allows inhibitory function within stepping to be investigated. In addition to investigating the role of inhibition in stepping, we compared this with the role of inhibition in comparable manual responding. This permits us to ask the question of how differences in motor control and functional requirements, e.g., maintaining balance, might influence how inhibition is exercised.

METHODS

Subjects.

The study protocol was approved by the Institutional Review Board (IRB) at the University of Pittsburgh. Forty (11 men, 29 women) community-ambulating older adults aged 70–94 yr volunteered for the study (mean 75 yr, SD 5 yr) after being recruited from a registry of older adults and from participation in other research studies. There was no difference in mean age between the male and female subjects (F_1,38 = 0.03, P = 0.87). After subjects provided informed consent, a comprehensive screening examination was performed to exclude subjects with a history of neurological, cardiopulmonary, or orthopedic disease that would limit their mobility and balance function (e.g., stroke, Parkinson's disease, vestibular disease, multiple sclerosis, uncontrolled diabetes, uncontrolled hypertension, chronic obstructive pulmonary disease, significant peripheral neuropathy, significant vision loss, peripheral vestibulopathy, severe arthritis). Specific exclusion criteria included taking psychoactive medications (e.g., benzodiazepines), corrected visual acuity worse than 20/40, lower extremity nerve conduction velocity <40 m/s, and reduced neurocognitive function as determined by a score >1.5 SDs below age-adjusted mean in the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS; Pearson Education, San Antonio, TX) (Randolph et al. 1998).

Inhibition test of step initiation procedures.

Two types of PI stepping tasks and one type of MI task were performed, in addition to standard simple reaction time (SRT) and choice reaction time (CRT) stepping tasks (described below). The testing was conducted on the third visit of a larger experiment that assessed different aspects of step initiation. As such, the subjects were highly trained in the procedures of step initiation. Subjects donned a harness attached to an overhead carriage in order to prevent a fall from occurring. They stood on two independent force plates, arranged as in Fig. 1, which were used to quantify the timing and biomechanics of the step responses. A wooden platform surrounded the force plates at the same height as the top surface of the plate. Nonslip color-coded circles were placed on the force plates to designate the starting and target step locations. The lateral distance of the target locations from the starting position was 25% of the subject's leg length, defined as the distance from the floor to the greater trochanter. Subjects wore their own comfortable shoes.

Fig. 1.

Experimental setup showing a diagram of the foot placement on the forceplates (A) and a harnessed subject viewing the monitor with a congruous stimulus (B). SP, starting position; L1, left step location; R1, right step location.

Subjects stepped in response to visual stimuli generated by E-Prime 2.0 (Psychology Software Tools, Pittsburgh, PA). The visual stimuli were displayed on a 51-cm CRT monitor set just below eye height at a distance of 1 m from the subject. Thus the horizontal display of the screen encompassed a horizontal visual arc of 23°. While standing in the starting position, subjects fixated on the monitor that displayed, diagrammatically, an overhead view of the force platform configuration, with red circles indicating where their feet should be placed.

At the onset of the display of the visual stepping cues (described in detail below), subjects stepped to the left or right with the appropriate leading leg and followed with the trailing leg so that both feet came to rest laterally to the starting position. Subjects were instructed to step as quickly as possible when the visual stepping cue appeared and to continue looking at the monitor during the step, without worrying about the exact landing location of their foot. Subjects were given several trials to become consistent with their step length. After subjects returned to the starting position and achieved an equal weight distribution on both feet as measured by the force plates, the next stimulus occurred randomly during a 4- to 6-s window. Subjects were given verbal feedback to equalize their weight distribution if needed.

Table 1 details the 12 blocks that were performed. Blocks 1–3 were blocks that were used as baseline performance to compare with the previous testing sessions and were not of interest for this article.

Table 1.

Definition of blocks for inhibition test of step initiation

Block	Step Task	Measures
1-3	Baseline performance, results not reported
4	Simple reaction time	SRT
5	Simple reaction time	SRT
6	Choice reaction time	CRT
7	Perceptual inhibition-location	PI-LOC-CON, PI-LOC-INCON
8	Perceptual inhibition-direction	PI-DIR-CON, PI-DIR-INCON
9	Perceptual inhibition-location	PI-LOC-CON, PI-LOC-INCON
10	Perceptual inhibition-direction	PI-DIR-CON, PI-DIR-INCON
11	Motor inhibition-congruous	MI-CON
12	Motor inhibition-incongruous	MI-INCON

Blocks 4–6 served as SRT and CRT blocks, utilizing left and right arrows as the visual cues (Fig. 2). The color of the directional arrows was black, and the size of the arrows was 4° horizontal and 2° vertical. During block 4, subjects stepped to the left in response to leftward arrows appearing on the left side of the monitor. During block 5, subjects stepped to the right in response to rightward arrows appearing on the right side of the monitor. For the SRT blocks, two catch trials in which no arrow was displayed appeared in random order among the other stimuli. The durations of blocks 4 and 5 were 3 min, allowing 24 stimuli to be displayed. Block 6 was the CRT block containing both left and right arrows lateralized to left and right sides of the monitor, randomly ordered. The duration of block 6 was 4 min, allowing 16 of each stimulus to be displayed.

Fig. 2.

Visual stimuli for the inhibition task conditions. Labels at bottom left and right refer to the required step direction for each visual stimulus.

Blocks 7–10 constituted the blocks for the PI tasks of step initiation (Fig. 2). The PI task blocks required subjects to step laterally with their left foot or their right foot in response to left or right arrows (← or →) that were displayed on the left or right side of the monitor, depending on the instructional set given before the block started. The cues for step direction were congruous (left arrow on left side, right arrow on right) or incongruous (left arrow on right, right arrow on left). During the incongruous trials, a perceptual conflict arose between the location where the arrow was displayed, which has a strong effect on orienting attention (Simon 1969), and the directional meaning of the arrow, which also is a potent symbol for orienting visual attention (Hommel et al. 2001). Thus, depending on the instructions, subjects needed to inhibit responses to one of these two distractors of visual attention and respond to the relevant cue. In blocks 7 and 9, referred to as the perceptual inhibition-location task (PI-LOC), subjects stepped to the same side on which the arrow was displayed on the monitor and ignored the direction that the arrow was pointing. In blocks 8 and 10, referred to as the perceptual inhibition-direction task (PI-DIR), subjects stepped to the side in which the arrow was pointing, while ignoring its location on the screen. Interference in reaction times increases as the proportion of incongruous (e.g., ← on the right side of display) and congruous stimuli (e.g., ← on the left side of display) approaches 50%-50% (Logan and Zbrodoff 1979; MacLeod 1991), so an equal number of congruous and incongruous stimuli were displayed in random order during the 4-min blocks, resulting in ∼32 total stimuli.

Blocks 11 and 12 made up the MI test of step initiation (Fig. 2). Block 11 displayed the congruous trials, in which subjects stepped in the same direction of the centrally displayed arrow. Block 12 displayed the incongruous trials, in which subjects stepped in the opposite direction of the centrally displayed arrow. When instructed to step away from the arrow direction, subjects again had to inhibit stepping toward the focus of their visual attention. In both blocks, an equal number of left- and right-pointing arrows were displayed in random order. The first six subjects did not perform these blocks but instead repeated the PI-LOC and PI-DIR blocks. The order of blocks was fixed, consistent with the previous manual tasks (Jennings et al. 2011; Nassauer and Halperin 2003).

Outcome measures.

Ground reaction forces and moments were collected from the force plate at a sampling frequency of 1,000 Hz. Several critical events during each step were identified from the VGRF of the stepping leg (Fig. 3) (Melzer and Oddsson 2004; Rogers et al. 2003). For lateral steps taken during the SRT blocks, the VGRF assumed one of two profiles. In one profile, labeled as having one postural adjustment, the VGRF decreased monotonically from 50% body weight to 0% body weight. We defined the first deflection of the VGRF from 50% body weight as the latency of the first postural adjustment (PA1) and the point when the VGRF fell to 0% body weight as the liftoff time (LO). PA1 represented the reaction time, i.e., the instant when there was observable evidence that the body was responding to the stimulus. Meanwhile, LO represented the functional outcome of the task, because once the leg had lifted off of the force platform the subject had committed to taking a step in that direction. The latencies of PA1 and LO were determined as the moment when the derivative of the VGRF exceeded (PA1) or fell below (LO) a threshold of 2% body weight/s, which was based on prior investigations. In the second profile, having two postural adjustments, the VGRF initially increased in value, representing the force to propel the body toward the stance leg before decreasing to unload the stepping leg so that the step could be taken. Again, the first deflection of the VGRF was the PA1, and the LO occurred when all weight had been shifted to the stance leg and the VGRF declined to 0. The peak in the VGRF was defined as the second postural adjustment (PA2). In some cases, additional inflection points in the VGRF were observed, even though the subject stepped with the correct leg. As a result, every time there was a reversal of direction of the VGRF, it was considered to be another postural adjustment (PA2, PA3, etc.).

Fig. 3.

Representative vertical ground reaction force (VGRF) curves for lateral steps that had 1 and 2 postural adjustments. PA1 marks the first postural adjustment, PA2 is the second postural adjustment, and LO is the time of liftoff.

Motor and perceptual inhibition test.

The Motor and Perceptual Inhibition Test (MAPIT), defined by Jennings et al. (2011) and based on the work of Nassauer and Halperin (2003), is a computer-based neuropsychological test of inhibition. The MAPIT results were used as a corollary measure to examine whether there was a relationship between these inhibition measures assessed with manual key presses and our step initiation inhibition results. The MAPIT test was administered on a desktop computer using E-Prime 2.0 with the stimulus response box (Psychology Software Tools), consisting of five buttons aligned in a single row in front of the subject. Each task began with a short practice to confirm that participants understood the required movement. There were 40 trials for each of the 5 blocks (200 total reaction time trials), which were performed in fixed order. During all trials participants were encouraged to keep their vision fixed on the center of the computer screen; this was achieved by displaying a plus sign (+) at the central fixation point. The interval between the patient's response at any given trial and the subsequent trial was always 1.5 s. The five blocks, as detailed by Jennings et al. (2011), were motor congruous training task (block a), perceptual inhibition priming task (block b), perceptual inhibition task (block c; analogous to step blocks 8 and 10), motor congruous task (block d; analogous to step block 11), and motor incongruous task (block e; analogous to step block 12). The median reaction time was computed for each task. Measures of PI and MI were then calculated as per Jennings et al. (2011).

Statistical analysis.

We planned to compare the latency of PA1 following a stimulus as well as time of LO. However, when we analyzed the data we observed that the number of postural adjustments (i.e., the step behavior) varied from subject to subject and from task to task. The presence of multiple postural adjustments in some but not all subjects suggested, however, that this scoring was too simplistic in that liftoff did not uniformly follow a single postural adjustment. Therefore a grouping approach was used to account for this complexity, and each step was classified according to how many postural adjustments occurred (1, 2, or 3 or more). Furthermore, subjects were classified based on how many postural adjustments they typically generated during the SRT task conditions, when subjects knew unequivocally in which direction to step (blocks 4 and 5). On the basis of their step behavior in these blocks, we classified subjects into three groups: group SB1 preferentially had one postural adjustment during the SRT, group SB2 preferentially had two postural adjustments, and group SB3 did not have a preference. Assignment to group SB1 or SB2 required that a subject's proportion of steps with one postural adjustment or two postural adjustments was significantly greater than 50%, respectively, by the binomial test (P < 0.05). This subject group classification based on performance during the SRT condition was later used as an independent variable.

The primary dependent variables in the analysis were the step behavior (i.e., the proportion of steps with 1, 2, and 3 or more postural adjustments) and the latency of PA1 and LO. For all conditions, PA1 times <100 ms were considered to be an error and excluded from analysis. For the blocks, data were processed from correct step responses only (i.e., based on the correct limb taking the step, regardless of the initial behavior). In previous work, test-retest intraclass correlation coefficients of PA1 and LO times were >0.77, demonstrating acceptable reliability (Melzer et al. 2007). For each subject, median values were used as estimates of PA1 and LO for various combinations of the task conditions, in order to protect against extreme values that could influence the mean. In all analyses, a median value was calculated only if there were five or more valid steps performed for that combination. If fewer than five valid steps were performed, the value was considered to be missing, and this is noted in results. Linear mixed models (except where noted) using the SAS MIXED procedure (SAS Institute, Cary, NC) with unstructured working correlation matrix were fit to account for multiple measurements from the same subjects and the resulting stochastic nonindependence of observations.

Classification of step behavior.

First we examined whether step behavior changed as a function of the different task conditions and subject group classification. The dependent variable was the proportion of steps made that were the same as the subject's preferred step behavior. Since group SB3 did not have a preferred step behavior during the SRT condition, the statistical analysis used only the subjects from groups SB1 (preferred = 1 postural adjustment, nonpreferred = 2 or more postural adjustments) and SB2 (preferred = 2 postural adjustments, nonpreferred = 3 or more postural adjustments). The main effects of step behavior group (SB1 and SB2) and step task condition (SRT, CRT, PI-LOC-CON, PI-LOC-INCON, PI-DIR-CON, PI-DIR-INCON, MI-CON, MI-INCON) and the interaction effect between group and step task were included in the model as fixed effects. We hypothesized that when the inhibition demands of the task increased the step behavior would shift away from the preferential pattern exhibited during the SRT blocks, because more stepping errors would be produced. We used post hoc testing with preplanned contrasts to further investigate the role played by the type of inhibition task (i.e., perceptual and motor) and congruency of the stimuli.

Latency of first postural adjustment and liftoff.

Considering the entire subject sample (SB1, SB2, and SB3), we again fit a model consisting of step behavior group, step task, and their interaction. We hypothesized that when the inhibition demands of the task increased, the PA1 and LO would be delayed. Since it is possible that the latency of PA1 and LO depended on the type of step behavior exhibited by the subjects, we also stratified the analyses with the subject groups that had a strong step behavior preference (SB1 and SB2), using only the data from their preferred step type.

On the basis of a post hoc observation that the latency of the PA1 and LO differed depending on each subject's step behavior, we fit a model using each subject's step behavior type (preferred, nonpreferred), inhibition task type (PI-LOC, PI-DIR, MI), and subject group (SB1, SB2) as fixed effects. Note that for this analysis of the effect of step behavior on PA1 and LO latencies, the only inhibition task conditions in which there were sufficient numbers of steps (i.e., >5) to obtain a good estimate of the latency PA1 and LO were the incongruous conditions, because during the congruous trials subjects rarely displayed nonpreferred step behavior.

Association with MAPIT.

Correlation analyses were conducted between the manual MAPIT tests and the step initiation tests. Specifically, we examined the relationship between the PI and MI scores (i.e., incongruous reaction time − congruous reaction time) of the MAPIT tasks and the analogous scores for both PA1 and LO measures of the step initiation tasks.

RESULTS

Classification of group step behavior.

The group classifications (SB1, SB2, and SB3) were based upon the number of postural adjustments made during the SRT trials. Eighteen subjects had a strong preference for one postural adjustment (group SB1, range 79–100% of steps), 14 subjects had a strong preference for two postural adjustments (group SB2, range 71–100% of steps), and the remaining 8 subjects did not demonstrate a clear preference (group SB3). There was no difference in mean age across the groups (F_2,37 = 0.12, P = 0.89). Group membership was evenly distributed among women (10 SB1, 11 SB2, and 8 SB3), but more men were in group SB1 (n = 8) than in groups SB2 (n = 3) and SB3 (n = 0) (χ² likelihood ratio = 7.8, P = 0.021). The percentage of steps in which subjects actually stepped with the wrong leg, i.e., lifted off with the incorrect leg, was <1%.

Inhibition task on step behavior.

Step behavior was impacted by the type of task performed in groups SB1 and SB2. The percentage of each subject's preferred step type that occurred varied, depending upon the characteristics of the task (Fig. 4). ANOVA revealed a significant step task effect (F_7,21.2 = 19.63, P < 0.0001) and group × step task interaction (F_7,21.2 = 3.63, P = 0.010) but not a significant group effect (F_1,29.2 = 1.84, P = 0.19). Generally, the step task conditions induced significantly fewer steps of each subject's preferred type compared with the SRT condition and significantly more steps with at least one additional postural adjustment. The significant group × step task interaction indicated that group SB2 had a larger reduction in number of preferred steps during the incongruous trials within the PI-DIR (P = 0.0005) and MI (P = 0.008) conditions compared with group SB1. For the PI-DIR-INCON condition, 65% of the steps were of the preferred type for group SB1, compared with only 34% for group SB2. For the MI-INCON condition, 75% of the steps made by group SB1 were preferred, compared with 50% of the steps made by group SB2. Additional post hoc testing demonstrated that within both subject groups incongruous trials induced a lower percentage of preferred steps compared with congruous trials for all inhibition task conditions (P ≤ 0.02). In addition, subject groups SB1 and SB2 differed in their responses to the type of inhibition task. Whereas group SB1 had minimal changes of <10% in proportion of preferred steps among the incongruous trials or incongruous trials of the PI-LOC, PI-DIR, and MI conditions, group SB2 had greater variation in these proportions. In particular for group SB2, all pairwise differences were significant (P ≤ 0.005) between the proportion of preferred steps for the PI-LOC-INCON (73%), MI-INCON (50%), and PI-DIR-INCON (34%) conditions.

Fig. 4.

Percentage of steps of each subject's preferred type of step behavior for different task conditions. Group SB1 preferentially demonstrated 1 postural adjustment during the simple reaction time (SRT) task, and group SB2 preferentially demonstrated 2 postural adjustments during the SRT task. Error bars are SE. CRT, choice reaction time; PIL-Con, perceptual inhibition location-congruous trials; PIL-Incon, perceptual inhibition location-incongruous trials; PID-Con, perceptual inhibition direction-congruous trials; PID-Incon, perceptual inhibition direction task-incongruous trials; MI-Con, motor inhibition direction-congruous trials; MI-Incon, motor inhibition direction task-incongruous trials.

Effect of inhibition task on PA1 and LO.

Prior to conducting the main statistical analysis, we confirmed that there was no difference in onset of PA1 when stepping to the left or to the right. Consequently, the left and right step data were combined before the median was computed. Furthermore, there was no difference in onset of PA1 between blocks 7 and 9 (i.e., PI-LOC condition) or between blocks 8 and 10 (PI-DIR condition). Therefore, blocks 7 and 9 were combined before the median PA1 was computed, as were blocks 8 and 10.

An analysis of PA1 was performed to determine the impact of task type on the initiation time of the first postural adjustment. For the entire subject sample (SB1, SB2, and SB3), ANOVA indicated a significant effect of step task condition (F_7,36.3 = 48.61, P < 0.0001) and group × step task interaction (F_14,36.6 = 3.26, P = 0.0021) but a borderline significant overall group effect (F_2,36.9 = 3.25, P = 0.050) on PA1. The significant group × step task interaction indicated that the magnitude of changes in the PA1 that occurred as a result of the different step tasks depended on group membership. Post hoc testing showed that group SB3 had delayed PA1 times compared with groups SB1 and SB2 during the CRT condition (P < 0.021) and delayed PA1 compared with group SB2 during the PI-DIR-CON (P = 0.008) and MI-INCON (P = 0.01) conditions. To better visualize the general changes in PA1 due to the step task conditions, data are presented for the entire subject sample in Fig. 5. First, a significant increase in PA1 occurred from the SRT to CRT (P = 0.0008), as expected when a choice was added to the task. Second, PA1 was greater for all inhibitory task conditions compared with the SRT and CRT tasks (P ≤ 0.001). Within the inhibitory task conditions, PA1 from the incongruous step trials was always significantly greater than the congruous step trials (P ≤ 0.0002). The difference in latency of PA1 between incongruous and congruous stimuli was greatest for the MI task (92 ms), intermediate for the PI-DIR task (52 ms), and least for the PI-LOC task (27 ms). Finally, PA1 was progressively greater from the PI-LOC task to the PI-DIR task to the MI task. For the congruous stimuli, PA1 from the PI-DIR and MI tasks were significantly greater than the PA1 from the PI-LOC task (P < 0.0001). For the incongruous stimuli, PA1 during the MI task was significantly greater than PA1 of the PI-DIR task (P = 0.0002), which was significantly greater than PA1 from the PI-LOC task (P < 0.0001).

Fig. 5.

Mean latency of PA1 of all subjects for all steps. Error bars are SE.

Statistical analysis of LO timing again showed a significant step task effect (F_7,35.6 = 109.5, P < 0.0001). Across the different step task conditions, changes in LO generally followed the same trends as PA1 (Fig. 6). However, LO did not demonstrate a group × step task interaction (F_14,35.7 = 1.87, P = 0.066). Furthermore, there was an overall group effect (F_2,36.8 = 24.23, P < 0.0001) on LO, which occurred because groups SB2 and SB3 made more postural adjustments compared with group SB1 and thus had delayed LO times. In addition, group SB2 had increased LO times compared with group SB3, because SB2 had at least two postural adjustments more consistently than SB3.

Fig. 6.

Mean latency of foot liftoff (LO) of all subjects for all steps. Error bars are SE.

When we examined PA1 only for the preferred step behavior within each subject group (i.e., steps with 1 postural adjustment for SB1 and 2 postural adjustments for SB2), findings were similar to the analysis using all of the steps (3 of 32 subjects had missing data for the PI-DIR-INCON task). There was a significant main effect of step task (F_7,29.2 = 55.71, P < 0.0001) and group × step task interaction (F_7,29.2 = 5.79, P = 0.0003) but not a significant group effect (F_1,29.9 = 1.82, P = 0.19). As with the full data set, the relative magnitudes of PA1 were affected by inhibitory task condition and congruency of the stimuli. The significant interaction can be explained by relative differences in PA1 between step task conditions for the different groups.

Step behavior, initiation, and liftoff times.

When subjects made additional postural adjustments compared with their typical step behavior, it was labeled as an error in step initiation. Figure 7 demonstrates these erroneous step initiations within each subject group. In these examples, note that when additional postural adjustments were made PA1 occurred earlier than during the preferred behavior and LO occurred later. Therefore, we explored differences in PA1 and LO when nonpreferred step behavior occurred, compared with the preferred step behavior. Figure 8 displays PA1 for the different step behavior types (i.e., preferred and nonpreferred) that occurred during the incongruous conditions. The number of missing subjects with missing data was 10 of 32 for the PI-LOC-INCON task, 7 of 32 for the PI-DIR-INCON task, and 5 of 26 for the MI-INCON task. ANOVA demonstrated significant main effects for step behavior type (F_1,28.3 = 80.39, P < 0.0001) and inhibition task type (F_2,30.2 = 23.42, P < 0.0001) and a significant interaction between step behavior type and inhibition task type (F_2,20.8 = 12.56, P = 0.0003). There was not a significant subject group effect (F_1,29.6 = 2.32, P = 0.14), nor were there significant interactions of subject group with step behavior type (F_1,28.3 = 0.36, P = 0.55) and inhibition task type (F_2,30.2 = 1.46, P = 0.25). The step behavior × inhibition task interaction revealed that the difference in PA1 between preferred and nonpreferred step types depended on the inhibition task type. For the PI-LOC-INCON trials, there was no significant difference in PA1 between preferred and nonpreferred steps (19-ms mean difference, P = 0.162). In contrast, PA1 was 97 ms earlier (P < 0.0001) during nonpreferred steps compared with preferred steps during the PI-DIR-INCON trials and 75 ms earlier (P < 0.0001) during the MI-INCON trials.

Fig. 7.

Example VGRF curves displaying preferred (solid) and nonpreferred (dashed) step behaviors. Subject SI016 (A) preferentially generated 1 postural adjustment, and subject SI025 (B) preferentially made 2 postural adjustments. Solid VGRF curves represent the preferred step behavior during an incongruous trial from the perceptual inhibition-direction task (PI-DIR-INCON). Dashed VGRF curves represent a step in which the subject generated more than his/her customary number of postural adjustments. The first postural adjustment (PA1) and liftoff (LO) are marked.

Fig. 8.

Mean latency of PA1 of subjects when they generated steps of their preferred (P) and nonpreferred (NP) step type during the incongruous trials of the perceptual inhibition-location (PI-LOC), perceptual inhibition-direction (PI-DIR), and motor inhibition (MI) tasks. Error bars are SE.

Whereas PA1 occurred earlier in the step trials that contained a greater number of postural adjustments for the PI-DIR-INCON and MI-INCON trials, the consequence of the greater number of postural adjustments was that LO was delayed. ANOVA showed significant main effects of subject group (F_1,27.6 = 40.0, P < 0.0001), step behavior type (F_1,21.6 = 107.0, P < 0.0001), and inhibition task type (F_2,22.9 = 5.44, P = 0.012) and a significant interaction between step behavior type and inhibition task type (F_2,13.1 = 12.16, P = 0.001). Figure 9 displays this interaction effect for each subject group. Average LO delays of 151, 90, and 129 ms occurred during the nonpreferred steps of the PI-LOC-INCON (P < 0.0001), PI-DIR-INCON (P < 0.0001), and MI-INCON (P < 0.0001) conditions, compared with the preferred steps.

Fig. 9.

Mean latency of foot liftoff (LO) of groups SB1 (A) and SB2 (B) when they generated steps of their preferred (P) and nonpreferred (NP) step type during the incongruous trials of the PI-LOC, PI-DIR, and MI tasks. Error bars are SE.

Association with MAPIT.

Reaction times during the congruous and incongruous trials of MAPIT were greater than the step PA1 but less than the step LO times (Table 2). In the MAPIT, the difference scores between the incongruous and congruous trials represent the perceptual inhibition (PI-MAPIT) and motor inhibition (MI-MAPIT) scores and were found to be 54 ms and 102 ms, respectively. The equivalent measures for the step inhibition tests are the PI-DIR-PA1 (i.e., the difference in PA1 between PI-DIR-INCON and PI-DIR-CON conditions), PI-DIR-LO (i.e., the difference in LO between PI-DIR-INCON and PI-DIR-CON conditions), MI-PA1, and MI-LO. Qualitatively, PI-DIR-PA1 and MI-PA1 of the step task were similar to PI-MAPIT and MI-MAPIT. However, PI-DIR-LO and MI-LO were greater than analogous scores for MAPIT and PA1, with both being around 130 ms.

Table 2.

Reaction times obtained for congruous and incongruous trials of PI and MI components of MAPIT and step initiation test

	Perceptual Inhibition Task	Motor Inhibition Task
MAPIT
Incongruous	539 (12)	502 (16)
Congruous	486 (12)	400 (9)
	PI-MAPIT	MI-MAPIT
Incongruous − congruous	54 (5)	102 (12)
Step initiation-PA1
Incongruous	346 (13)	398 (13)
Congruous	294 (8)	306 (8)
	PI-DIR-PA1	MI-PA1
Incongruous − congruous	52 (6)	92 (10)
Step initiation-LO
Incongruous	742 (22)	777 (21)
Congruous	614 (18)	647 (19)
	PI-DIR-LO	MI-LO
Incongruous − congruous	128 (8)	130 (12)

Values are mean (SE) reaction times obtained for the congruous and incongruous trials of the perceptual inhibition (PI) and motor inhibition (MI) components of the Motor and Perceptual Inhibition Test (MAPIT) and step initiation test. PA1, onset of first postural adjustment, all step behaviors; LO, onset of liftoff, all step behaviors.

Considering the effect of step behavior group on PI and MI, there were no differences in PI-MAPIT (t₃₀ = 0.86, P = 0.40) or MI-MAPIT (t₃₀ = −0.85, P = 0.40) between groups SB1 and SB2. Nor were there group differences in MI-PA1 (t₂₄ = 1.39, P = 0.18) or MI-LO (t₂₄ = −0.29, P = 0.77). However, there were significant differences between the groups for the step task PI-DIR-PA1 (t₃₀ = 2.09, P = 0.046) and PI-DIR-LO (t₃₀ = −2.61, P = 0.014). PI-DIR-PA1 was 23 ms greater for SB1 compared with SB2. On the other hand, PI-DIR-LO was 47 ms greater for SB2 compared with SB1.

Table 3 details the correlation between the values of the manual MAPIT test and the analogous values obtained from the step initiation measures. PI-MAPIT was not significantly correlated to either PI-DIR-PA1 or PI-DIR-LO. There were strong correlations between the percentage of nonpreferred steps made and PI scores for both PA1 and LO. Subjects who made more errors had less PI in their initial postural response (PI-DIR-PA1, r = −0.64) but had greater PI-DIR-LO times (r = 0.60) because of the time spent making additional postural adjustments (Fig. 10). For the MI tasks, MI-MAPIT was positively correlated with MI-LO as well as the number of nonpreferred steps in the MI-INCON condition.

Table 3.

Pearson correlation and P values between PI and MI scores obtained from MAPIT reaction time measures and step initiation measures: PA1 and LO

	PI-DIR-PA1	PI-DIR-LO	% Nonpreferred Steps PI-DIR-INCON	MI-PA1	MI-LO	% Nonpreferred Steps MI-INCON
PI-MAPIT	r = −0.06 (P = 0.70)	r = 0.06 (P = 0.71)	r = −0.09 (P = 0.61)
PI-PA1		r = 0.15 (P = 0.36)	r = −0.64 (P = 0.0001)
PI-LO			r = 0.60 (P = 0.0002)
MI-MAPIT				r = −0.09 (P = 0.61)	r = 0.34 (P = 0.05)	r = 0.45 (P = 0.020)
MI-PA1					r = 0.55 (P = 0.001)	r = −0.18 (P = 0.39)
MI-LO						r = 0.47 (P = 0.015)

Values are Pearson correlations (P value) between PI and MI scores obtained from the MAPIT reaction time measures and step initiation measures: PA1 and LO. PI-DIR-INCON, perceptual inhibition-direction task, incongruous trials; MI-INCON: motor inhibition task, incongruous trials.

Fig. 10.

Scatterplot of perceptual inhibition scores computed from the initial postural adjustment (PI-DIR-PA1, top) and liftoff (PI-DIR-LO, bottom), as a function of subject's percentage of nonpreferred step trials during the perceptual inhibition-direction task. Each point represents a different subject in group SB1 or group SB2.

DISCUSSION

The purpose of the experiment was to further investigate the influence of inhibition on postural control by having subjects perform a step initiation task involving response inhibition to incongruous visual cues. The experimental paradigm exploited the tendency for motor responses to be directed toward the location of spatial attention (Simon 1969), and thus subjects were required to inhibit step responses toward an irrelevant visual cue. A primary finding was that older adult subjects demonstrated different stepping behaviors, as revealed by the VGRF curves, in response to trials in which they knew the direction to step. Second, when presented with incongruous conditions that presented conflicting visual cues about step direction, subjects were more likely to change their behavior and generate a greater number of postural adjustments than they typically produced during congruous trials. This type of behavior was more likely to be elicited when the step was initiated earlier, before the subjects were able to inhibit the processing of irrelevant visual cues, producing responses that required online correction. Another important result was that even when subjects used their preferred step strategy the inhibition tasks induced delays in the latency of the initial postural adjustments (PA1), demonstrating the interference and increased processing time caused by the conflicting cues. A potentially deleterious functional outcome of both the existence of postural adjustment errors and delays in PA1 for preferred step behavior was an increase in step liftoff times (LO). Finally, we determined that inhibitory function assessed with a manual task is not directly related to inhibitory function assessed in this step initiation task. The present results, which arose from a test in which the inhibition task was integral to the postural control task, complements existing research in which inhibitory function has been related to increased postural sway, errors in step initiation, and an increase in incident falls (Cohen et al. 2011; Herman et al. 2010; Mirelman et al. 2012; Redfern et al. 2009).

Classification of group step behavior.

Most subjects (32/40) generated a stereotypical response during the SRT task, i.e., when they knew a priori in which direction they were going to step. Eighteen of the subjects (group SB1) typically generated only one postural adjustment, executing a step without shifting their center of gravity toward the stance limb. This pattern was previously reported in young adult subjects for lateral steps (Patla et al. 1993). Fourteen of the subjects (group SB2) characteristically used two postural adjustments: first an increase in VGRF of the stepping limb to propel the center of gravity toward the stance limb, then the unloading of the limb to accomplish foot liftoff and step. The remaining eight subjects displayed both behaviors, more or less equally (group SB3). At this time, it is not known whether the type of step behavior that one preferentially makes to known step locations relates to any demographic or physiological factors, such as sex, age, strength, fall risk, etc. For example, factors associated with differences in sex may play a role; although women were equally likely to be in each of the subject groups, men were more likely to demonstrate only one postural adjustment. Biomechanical factors may partially explain these different step behaviors. In the case of SB1, subjects have adequate muscle power to abduct the leg in time to land ahead of the center of mass that is moving toward the stepping leg. It is possible that SB2 subjects do not have adequate muscle power to do this, and thus must initially shift their center of mass toward the stance leg, thereby increasing the required time of muscle contraction to unload and abduct the leg successfully. Thus the subjects in group SB2 may have learned this more conservative strategy over time before making a lateral step. Within the narrow age range of older adults in this study, the exact influence of age cannot be determined. However, Patla et al. (1993) reported that their young adult subjects more frequently generated only a single postural adjustment compared with their older adult subjects (Patla et al. 1993), and preliminary data from our lab confirm this finding. Thus the SB2 characteristics seem to be only performed by some older adults, possibly reflecting reduced motor capabilities.

Changes in step behavior induced by inhibition task.

Deviations from normal SRT stepping behavior that occurred during the inhibition task conditions are believed to represent errors in initially selecting the correct step response. During the inhibition task conditions, a greater number of postural adjustments were performed than the subject typically made during the SRT: two or more postural adjustments for group SB1 and three or more postural adjustments for group SB2. In reference to Fig. 7A, consider a subject who typically generated one postural adjustment during a step to the right (solid line). Generation of two postural adjustments (dashed line, Fig. 7A) could indicate that the subject initially started stepping to the left, before correcting online and stepping to the right. Another possibility is that they had always intended to step right, but this time shifted their center of gravity toward the stance limb. Because it is not clear why a subject would suddenly start using a different technique, we believe that the first scenario is more likely. On the other hand, when a subject generated three or more postural adjustments, as shown in Fig. 7B, it is clear that an initial postural adjustment error was made, even though the subject ultimately stepped with the correct leg. Similar postural adjustment errors have been observed previously during forward stepping when subjects were unsure about which limb to step with during choice reaction time step tasks (Cohen et al. 2011; Jacobs and Horak 2007). Jacobs and Horak (2007) described the phenomenon as a “consequence of requiring online limb selection” during the choice reaction time step task, and Cohen et al. (2011) labeled them as errors in motor programming. We believe that the pattern of demonstrating more postural adjustments than is typical for an individual subject is an indication that there was a failure to initially select the correct motor program. Also, the ability to make subsequent postural adjustments demonstrates that corrections can be made rapidly so that the subject ultimately steps with the correct limb.

The proportion of steps of each subject's preferred behavior depended greatly on the inhibition task condition. The primary effect was a decrease in the proportion of the preferred step behavior and an increase in the resulting proportion of steps with multiple postural adjustments during incongruous conditions of the PI and MI tasks. For the PI tasks, subjects made more errors when required to inhibit the spatial location component and respond to the arrow direction. This was particularly true in group SB2, who produced errors in step behavior 66% of the time in the PI-DIR incongruous condition, compared with group SB1, who generated errors 35% of the time (note that both groups had <7% error rate in the SRT condition). Furthermore, group SB2 had a greater error rate in the MI-INCON task as well (50% for group SB2 v. 25% for group SB1). These results suggest that spatial location was a more potent attractor of visual attention compared with the symbolic meaning of the arrow. Consequently, inhibition of responses toward the spatial location of the arrow was more difficult than inhibition of responses in the direction that the arrow was pointing. The eight subjects in group SB3 were susceptible to the interference conditions in that their frequency of steps with three or more postural adjustments increased. Their mixture of step patterns makes it hard to specifically place their performance on a single continuum with groups SB1 and SB2.

We also analyzed how many additional postural adjustments were made during the nonpreferred steps. During the congruous trials, the overwhelming majority of these nonpreferred steps had only one additional postural adjustment, indicating that the subject initially responded in the wrong direction but successfully inhibited carrying out this erroneous initial response to step in the correct direction. Nonetheless, during 6–10% of the incongruous trials of the PI tasks subjects made two additional postural adjustments, indicating that the initial postural adjustment was in the correct direction but was inhibited. The relatively few occurrences of these initially correct, but inhibited, postural adjustments preclude formal statistical analysis. However, for some subjects, it was apparent that these steps were initiated at a very early time compared with the other steps, perhaps at a time when the subject had not yet processed the true step direction.

This pattern of step performance is consistent with conceptual and empirical work on choice based largely on hand and finger responses (Forstmann et al. 2008; Jahfari et al. 2011; Leite 2012; van Gaal et al. 2011). Typically, a choice response task is considered to initiate a nonspecific activation during preparation prior to and coincident with the choice stimulus. Activation then appears to increase somewhat randomly for each choice, with guidance by stimulus identity increasing over time. When activation for one choice exceeds a threshold, response occurs. Central processes act to inhibit action when interference occurs, alter the threshold, and generally regulate the buildup of activation, whether the response is correct or incorrect.

Effect of step behavior on PA1.

Although it is not clear why there was a group difference in number of steps with multiple postural adjustments, we can ascertain why the errors were made by examining the onset of PA1. During the incongruous trials, steps with additional postural adjustments were more likely when early PA1 were generated, as shown in Fig. 8. A likely interpretation is that subjects generated additional postural adjustments because they prematurely generated an initial step response (most frequently in the wrong direction but also in the correct direction) at a time early in the processing stage when the visual attention was not yet focused on the relevant feature (i.e., direction or location of the arrow), and thus the subjects were not certain in which direction to step (see next paragraph) (Gratton et al. 1988). Subsequently, online adjustments were made to step in the correct direction. Additional evidence for this interpretation is the similarity between the onset of PA1 during the incongruous trials when postural adjustment errors were made and PA1 obtained for the congruous stimuli in the very same block of the PI-DIR task. Across groups SB1 and SB2, the difference in PA1 between the PI-DIR congruous stimuli steps and incongruous stimuli steps with errors was 1 ms on average. Consequently, this finding suggests that on the trials when errors were made, subjects were following an instructional set appropriate for congruous stimuli and thus did not inhibit attention to irrelevant cues.

The finding of earlier activation of inappropriate step responses is consistent with several other studies that have investigated grip force responses to a flanker task with incongruous visual cues (Coles et al. 1985; Gratton et al. 1988). Similar to the present step results, the proportion of trials in which subjects exerted with both hands (i.e., made incorrect adjustments) increased during the presentation of incongruous cues. Furthermore, the onset of the EMG and force response of the incorrect hand during the trials when multiple adjustments were made was earlier than the onset of the EMG and force response of the correct hand during trials when a single adjustment was made. The investigators concluded that errors were more likely to occur during an early activation period when a person is still processing the whole stimulus, which can be dominated by irrelevant features (i.e., noise). On the other hand, errors were less likely to occur during a later activation period when attention was gradually allocated to the relevant feature of the stimulus (Gratton et al. 1988; White et al. 2011). The same hypotheses probably hold for the present study. Combining the findings of the upper extremity flanker task with our lower extremity step task, it appears that preparatory motor responses requiring inhibition are generalized across upper and lower extremities, and that the likelihood of making a correct response depends greatly on when motor program is initiated.

Effect of inhibition task on PA1.

PA1 increased with task condition, with SRT responses being fastest, followed by CRT and then the congruous and incongruous trials of the inhibition task blocks. Across the different tasks, which reflected different levels of inhibitory control, we propose that differences in PA1 indicated differences in the time of central processing, assuming that afferent (visual) and efferent (alpha motoneuron) conduction times were equivalent for all the task conditions. The increase in PA1 from the SRT (1 stimulus and 1 response) to the CRT (2 stimuli and response options) condition follows general models of cognitive processing during reaction time experiments in that reaction time increased as the number of potential stimuli and responses increased (Garner 1962; Hick 1952; Hyman 1953). PA1 was 19 ms greater in the CRT task compared with the SRT task, suggesting that the decision process increased by 19 ms. Previous tests of step initiation that had graduations in task difficulty reported similar findings (Cohen et al. 2011; Jacobs and Horak 2007).The mean increase in PA1 of at least 20 ms from the CRT to the congruous trials of the PI tasks, which were visually identical to the CRT but were displayed during a block that included two additional incongruous stimuli, also is supported by these models of cognitive processing.

With respect to the congruous trials of the three inhibition tasks, PA1 for the MI trials was significantly greater than PA1 for the PI-LOC trials. Given the lower number of potential stimuli during the MI task (i.e., 2) compared with the PI task (i.e., 4), one may think that responses would be faster for the MI task. However, it is possible that the novelty of the centrally located arrow delayed the responses compared with the well-practiced lateralized arrows of all of the preceding tasks (SRT, CRT, PI-LOC, and PI-DIR). As discussed above, it is also possible that the lateralization of the arrows in the PI conditions provided a stronger attractor of visual attention than the symbolic meaning of the central arrows, facilitating faster responses. Counter to this argument, however, are the results of Jennings et al. (2011), who reported MI-congruous reaction times to be faster than the PI-DIR congruous reaction times during the manual version of the inhibition task. It is unclear why the results of these two studies are inconsistent.

The additional increases in PA1 during the incongruous trials compared with the congruous trials were likely caused by interference from the conflicting cues, because the number of potential stimuli and responses were exactly the same for both congruous and incongruous trials. Evidence in support of the interference hypothesis is obtained by examining PA1 for the preferred steps only. The difference between congruous and incongruous preferred steps was 26 ms for the PI-LOC task, 98 ms for the PI-DIR task, and 114 ms for the MI task. The presence of interference in processing the instructed step direction is also supported by the multitude of experimental paradigms that involve congruous and incongruous cues (Coles et al. 1985; Craft and Simon 1970; Jennings et al. 2011; Liu et al. 2004). With the addition of dual tasks such as visuospatial working memory and Stroop tasks that were unrelated to the step task, several other groups found interference effects that caused significant delays in the latency of PA1 that ranged from 150 to 500 ms (Melzer and Oddsson 2004; St George et al. 2007).

Comparison of the two PI tasks revealed PA1 to be delayed more during the task when the subjects were instructed to step toward the side where the arrow was pointing (i.e., PI-DIR) compared with the task when they were instructed to step toward the side where the arrow was located (PI-LOC). This finding of greater interference during the PI-DIR task reinforces the results of the errors in step behavior and indicates that the spatial location of the arrow was a more potent attractor of visual attention than the directional meaning of the arrow. The greater interference in step behavior and latency of PA1 caused by incongruous cues during the PI-DIR condition compared with the PI-LOC condition probably relates to the Simon effect (Simon 1969), which is the tendency to respond more quickly toward the spatial source of a stimulus. In Simon's seminal experiments, subjects responded more quickly when they moved a handle toward a monaural cue compared with away (Simon 1969), or when they pressed a button that was compatible with the location of a visual cue (Craft and Simon 1970). Furthermore, Nassauer and Halperin (2003) found that for the upper extremity responses to the arrow location were faster than responses to arrow direction.

If we assume that the speed of transmission across the peripheral nervous system segments (i.e., retina to the brain and spinal cord to muscles) is consistent for all trials, then the increased reaction times that are due to differences in inhibition task design reflect increased central nervous system (CNS) processing in the widely distributed regions of the brain involved in planning and executing a step. Several fMRI studies of manual button presses in response to congruous and incongruous directional cues (i.e., similar to the PI-DIR task) have demonstrated increased activation of many frontal areas including the dorsolateral prefrontal cortex (DLPFC), anterior cingulate cortex, and supplementary and presupplementary motor areas (Liu et al. 2004; Peterson et al. 2002). Many other areas such as visual association, inferior temporal, and inferior parietal are also involved (Liu et al. 2004; Peterson et al. 2002). Furthermore, using near-infrared spectroscopy, our lab has detailed increased activation of the DLPFC during a similar step task involving incongruous cues (Huppert et al. 2012). Future studies may reveal what structures in the brain are responsible for the increased processing time.

Effect of inhibition task and step behavior on LO.

The LO time represents the final act of the step initiation. Once LO occurs, the goal is final and a step in that direction is determined. For this reason, we consider it to be the functional outcome of the task, similar to the reaction time measured by manual button press tasks. Two primary factors account for the influence of task condition on LO. The first is the dependence of LO on PA1. Obviously, if PA1 is delayed because of the task condition, then LO will be as well. The contribution of delayed PA1 to LO can be estimated by examining the relative increases in both measures across the task conditions for the preferred step behavior. For example, on average the PA1 for preferred steps increased by 93 ms between the PI-DIR congruous trials and incongruous trials. This represented 90% of the change in LO, which differed by 103 ms between the same two conditions.

The second factor concerns the delays in LO caused by making additional postural adjustments during nonpreferred steps (refer to Figs. 7 and 9). Again, using PI-DIR condition as an example, the LO time was delayed by 71 ms when nonpreferred steps were made compared with when the preferred step strategy was produced. Therefore, the consequences of making an error by failing to inhibit an incorrect response are worse than delaying the PA1 in order produce to a correct response.

Association with MAPIT.

The manual MAPIT test responses shed additional light about inhibitory processes during stepping. First, we observed that the manual MAPIT responses were greater than the PA1 times but less than the LO times (Table 2). At first glance, it may not seem possible that upper extremity reaction times (i.e., MAPIT responses) could be slower than lower extremity reaction times (i.e., latency of PA1). However, considering that the manual button press reflects the timing of a binary process (the switch being open or closed), it indicates the functional motor output of the task, similar to the way that the LO reflects the functional task of step initiation (i.e., foot is on the ground or off the ground). A simple switch such as used during the manual task does not allow us to determine whether a subject initially intended to press with the wrong hand and self-corrected.

There was a difference between groups SB1 and SB2 in the stepping equivalents to the MAPIT PI tests (PI-DIR-PA1 and PI-DIR-LO). Interestingly, the differences were in the opposite directions for PI-DIR-PA1 compared with PI-DIR-LO. The value for PI-DIR-PA1 was longer for group SB1 compared with group SB2, while PI-DIR-LO was shorter for SB1 compared with SB2. These seemingly discordant effects of group membership on PI can be explained if the percentage of steps with multiple postural adjustments is included in the models. Recall that group SB2 had a greater percentage of steps with additional postural adjustments compared with group SB1 during the PI-DIR-INCON condition. Figure 10, top, demonstrates that as the percentage of nonpreferred steps increased the PI-DIR-PA1 score decreased (r = −0.64, P = 0.0001), indicating a smaller difference of PA1 between congruous and incongruous trials in subjects who made more errors in their initial response direction. Another way to view this relationship is that subjects who made fewer errors did so because they waited longer, i.e., used inhibition appropriately. An analysis of covariance demonstrated that when the effect of the step errors was accounted for, the relationship between group membership and PI was no longer significant (P = 0.88).

The effect of group membership on the PI-DIR-LO score was also affected by the relative amount of nonpreferred steps each group made. Figure 10, bottom, shows that subjects who made more errors had increased PI-DIR-LO scores (r = 0.61, P = 0.0002), primarily because additional postural adjustments during incongruous trials resulted in greater LO times, relative to congruous trials when there are fewer errors. Again, after accounting for the percentage of nonpreferred steps, PI-DIR-LO was not related to group membership (P = 0.52). The interpretation of this result in the context of the experiment is that group SB2 initiated a step faster when presented with a perceptually incongruent task but, however, took longer to lift off because of more online error corrections. This is consistent with the error analysis between the groups. However, by presenting this in the context of PI, the results may suggest that people who use the SB2 strategy rely less on resolving perceptual conflict before they initiate the step. Rather, they use online correction to make the final decision toward final execution (i.e., liftoff).

There was only one significant correlation between the MAPIT inhibition measures (MI or PI) and the step inhibition scores, MI-MAPIT and MI-LO (r = 0.34, P = 0.05). Thus the measures of PI and MI obtained from the step test are not correlated with the manual MAPIT test to the degree one would expect by using identical test protocols. Consistent with current theories of inhibition (Kramer et al. 1994; Lustig et al. 2007; Nassauer and Halperin 2003), our results have demonstrated that several measures of inhibition used in the study appear to be independent. Furthermore, there appear to be two distinct types of inhibitory processes uncovered by this experiment. The differences between the incongruous and congruous reaction times during the PI and MI tests reflect an aspect of inhibition that is commonly assessed by using the classic Stroop paradigm. Greater differences are generally considered to indicate greater interference in resolving the appropriate response. The other process revealed in this experiment is inhibition failure, evidenced by the earlier onset of PA1 during trials when postural adjustment errors were made. Thus it could be that the step inhibition scores are measuring something different than the MAPIT manual method.

Interestingly, the magnitude of the step inhibition scores for both for PI and MI were similar to those found in the MAPIT test. This suggests that perceptual and motor inhibitory function within the CNS are being measured at approximately the same timescale. Differences arise when measured by LO, potentially for two reasons: First, there appears to be online correction occurring during step initiation between PA1 and LO. This is something that cannot occur in the MAPIT manual test. Thus this is a fundamental difference between the two tests. Second, there can be two strategies that exist in step execution (i.e., SB1 vs. SB2). Given the differences in the step PI scores seen between the two groups as discussed above, the behavioral strategy of the individual will have an effect on the results. This was a healthy cohort of older subjects with relatively intact inhibitory function. Examining a population with reduced inhibitory function would be useful to determine the similarities and differences in MAPIT and step inhibition.

Clinical implications.

There is indication in the literature that increased errors in step performance are related to increased fall risk (St George et al. 2007). In addition, Cohen et al. (2011) reported that when errors were made the subjects' center of mass traveled 50% further, which may signal greater likelihood of a loss of balance and fall. Because of the small sample size, it is not clear from our study whether subjects who made greater errors had an increase in fall risk. However, it is interesting to note that of the seven subjects who reported a fall in the previous year, five were members of group SB2. This is consistent with the previous studies, and could imply that there is also an association of inhibitory capacity in stepping error production related to falls. Further research is warranted to explore this relationship.

As mentioned above, a fast step response is an important determinant of arresting potential falls (van den Bogert et al. 2002). As a result, if conditions dictate uncertainty as in real life, and postural adjustment errors are generated, the probability of falling may increase. For example, postural perturbations such as bumps or stumbles occur frequently. If a person is not able to accurately detect which direction the perturbation came from quickly enough, an erroneous initial postural adjustment may occur that ultimately delays the correct protective stepping response, resulting in a loss of balance or fall.

Conclusion.

Dual-task balance studies have illuminated the critical interaction between balance and attention. This study adds to the literature by examining inhibitory function that inherently involves postural control processes. The results showed that inhibition plays an important role in selecting the appropriate visuospatial cues needed for this step task. Furthermore, lack of inhibition leads to errors in selecting the correct motor program. Further exploration is warranted to investigate how inhibitory function relates to fall risk.

GRANTS

This research was supported by funding from the National Institutes of Health (R01 AG-031118, P30 AG-024827, P30 DC-005205), including the Pittsburgh Claude D. Pepper Older Americans Independence Center (P30 AG-024827), and the Eye and Ear Foundation.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

Author contributions: P.J.S., M.S.R., J.R.J., S.P., R.D.N., and J.M.F. conception and design of research; P.J.S., S.I.F., M.S.R., and S.P. analyzed data; P.J.S., S.I.F., M.S.R., J.R.J., S.P., and J.M.F. interpreted results of experiments; P.J.S. and S.I.F. prepared figures; P.J.S., S.I.F., and M.S.R. drafted manuscript; P.J.S., M.S.R., J.R.J., S.P., and J.M.F. edited and revised manuscript; P.J.S., S.I.F., M.S.R., J.R.J., S.P., R.D.N., and J.M.F. approved final version of manuscript.