Experimental Brain Research

, Volume 201, Issue 3, pp 467–478

Age-related differences in visual sampling requirements during adaptive locomotion

Authors

  • Graham John Chapman
    • Bradford School of Optometry and Vision ScienceUniversity of Bradford
    • Human Movement Laboratory, School of Sport and Exercise SciencesThe University of Birmingham
Research Article

DOI: 10.1007/s00221-009-2058-0

Cite this article as:
Chapman, G.J. & Hollands, M.A. Exp Brain Res (2010) 201: 467. doi:10.1007/s00221-009-2058-0
  • 251 Views

Abstract

This study investigates if there are age- and falls-risk related differences in the length of time individuals need following fixation of a stepping target in order to step accurately onto it. This aim was achieved by manipulating the timing and location of stepping target presentation and comparing the effects on stepping performance between young adults, older adults characterised as having a low risk of falling and older adults characterised as having increased risk of falling (N = 10 in each group). Eye and lower limb kinematics were recorded using an eye tracker interfaced with a 3D motion analysis system. Temporal and spatial characteristics of eye and stepping movements were analysed and compared between groups and conditions in which participants had either <1, 2 or 3 s, following target fixation, in order to view and respond to target presentations. Comparisons were made between steps to centrally or laterally positioned targets (125% of individual participant’s normal step width). The results showed that high-risk older adults required significantly more time than low-risk older and younger adults in order to plan and execute medio-lateral stepping adjustments. A reduced ability to make rapid sideways stepping adjustments to avoid obstacles or step on safe areas may contribute towards trips and falls in these individuals. Possible neural mechanisms underlying this group-related decline in performance are discussed.

Keywords

Gaze behaviourStepping accuracyWalkingFallsElderly

Introduction

Safe and efficient walking through our cluttered world relies on proactive sampling of information from the visual scene in order to identify obstacles and goals. Therefore, identifying where and when walking individuals look is crucial if we are to gain a full understanding of the visual control of locomotion and how it changes over the lifespan. Previous studies have demonstrated that walking individuals invariably direct their gaze towards stepping targets prior to initiating the step onto that target (Hollands et al. 1995; Hollands and Marple-Horvat 2001) and that there is a consistent timing relationship between looking and stepping. Coordinated eye and stepping movements occur even in complete darkness (Hollands and Marple-Horvat 2001) suggesting that eye movements likely serve to provide extra-retinal information to guide the stepping movement in addition to the visual information that normally results. These findings provide evidence that there is substantial coordination between movements of the eyes and lower limb movements during precision stepping tasks.

Recent evidence suggests that there are age-related changes in characteristics of the eye and stepping coordination. For example, Di Fabio et al. (2003a, b) demonstrated that when required to step over an obstacle or onto a raised platform, older adults made a downward saccadic eye movement significantly earlier than young adults prior to initiating the step onto or over the object. More recently, Chapman and Hollands (2006, 2007) demonstrated that older adults fixated future footfall targets significantly sooner than younger adults when required to step onto small targets and that step accuracy was reduced despite the increased fixation duration. These findings suggest that older adults may need more time than younger adults to process visual information and transform it into successful stepping movements during walking.

The primary aim of the current study was to investigate if there are age- and falls-risk related differences in the length of time individuals need to plan and execute visually guided locomotor adjustments in order to step accurately onto a target. This aim was achieved by manipulating the location and timing of stepping target presentation and comparing stepping performance between young adults, low-risk and high-risk older adults. We included conditions in which participants were required to make lateral stepping adjustments to targets so as to increase task difficulty and reduce target predictability. Our working hypothesis was that older adults require more time than younger adults following initial fixation of a stepping target in order to step accurately onto the target. Our prediction based on this hypothesis was that there would be a “threshold” duration below which individuals would start making systematic stepping errors and that, on average, this threshold would be higher for older adults than young adults and larger still for high-risk older adults. We also predicted any group differences in our outcome measures would be greater during steps to lateral target locations.

Methods

Participants

Ten healthy young adults (mean age 25.1 ± 2.68, range 22–30 years), ten older adults determined to be at high risk of falling (high-risk) (mean age 76.1 ± 5.32, range 68–85 years) and ten older adults determined to be at low-risk of falling (low-risk) (mean age 69.9 ± 4.53, range 65–77 years) participated in the study. All older adults were community-dwelling and lived independently (see Table 1 for participant details). The experimental protocol was approved by The University of Birmingham Ethics Committee and all participants gave written informed consent prior to data collection. Participants were asked to report any neurological or musculoskeletal impairment and excluded from the study if any problems were reported. All participants were examined for cognitive functioning using the Mini Mental State questionnaire (Folstein et al. 1975) and all attained a score of 26 or over indicative of intact cognitive function. Participant’s binocular visual acuity was examined using a Snellen eye chart and all participants were found to have 20/40 vision or better. No participants wore spectacles during the visual screening or during the experimental task. Four participants wore contact lenses during the experiment. Participants were categorised at being low- (scored 45 or over) or high-risk of falling (scored below 45) based on their scores obtained from the Berg Balance Scale (Berg et al. 1992). Only four of the older adult participants (all in the HROA group) group had a score of <80% the Activities Balance Confidence Scale suggesting the majority of participants had minimal fear of falling (Powell and Myers 1992). Cognitive functioning was examined using the Trail Making B test, and ANOVA showed there was a significant main effect of participant group [F(2,27) = 16.40, P < 0.001]. Post hoc analysis revealed that high-risk older adults took significantly longer to complete the Trail Making B test than low-risk older adults and younger adults.
Table 1

Participant characteristics and scores on mobility, sensory and cognitive screening tests averaged across each group

Characteristics

Younger adults

Low-risk older adults

High-risk older adults

Height (m)

 Mean

1.74 ± 0.09

1.66 ± 0.10

1.59 ± 0.08

 Range

1.55–1.85

1.52–1.79

1.50–1.65

Weight (kg)

 Mean

73.62 ± 12.67

64.96 ± 9.59

70.42 ± 23.84

 Range

55.00–95.00

55.00–80.00

40.00–108.00

Mini mental state

 Mean

30 ± 0.00

29.7 ± 0.48

29.5 ± 0.97

 Range

29–30

27–30

Berg balance scale

 Mean

56 ± 0

54.7 ± 2.40

42.6 ± 1.77

 Range

49–56

39–44

Activities balance confidence scale (%)

 Mean

97.26 ± 3.70

94.47 ± 4.57

84.97 ± 10.16

 Range

88.43–99.87

86.87–99.37

70.00–97.87

Trail making B (s)

 Mean

52.28 ± 9.96

72.90 ± 21.06

103.10 ± 25.51*

 Range

35.00–69.00

54.00–127.00

71.00–150.00

Lower visual field size (°)

 Mean

69.50 ± 2.84

65.00 ± 6.24

66.00 ± 3.94

 Range

65–70

55–70

60–70

Contrast sensitivity (log units)

 Mean

1.74 ± 0.08

1.58 ± 0.20

1.58 ± 0.15

 Range

1.65–1.80

1.20–1.80

1.35–1.80

Older adult fallersa

0

0

70

Three or more medications (% of group)

0

0

40

Cardiovascular medications

0

40

50

Anti-anxiety or sedatives

0

Mediations for dizziness

0

General aches and pains

0

50

70

Minor lower limb problems

0

30

30

Bone problems

0

* Significantly different to young adults and low-risk older adults

aPercentage of older adult participants that reported at least one fall in the last year

Participants’ lower visual field was assessed monocularly by kinetic perimetry with a Goldman perimeter (Haag-Streit, UK) using the V/4e target (1.75° test spot at 320 cd/m2) on a background luminance of 10 cd/m2 (Zadnik 1997). Scores from each eye were then averaged to provide a more accurate binocular score for each participant. A one-way ANOVA revealed there were no differences between groups’ binocular lower visual field size [F(2,27) = 2.68, P > 0.05]. Contrast sensitivity (CS) was assessed binocularly using the Pelli–Robson letter sensitivity test (Pelli–Robson CS chart 4K, Metropia Ltd, UK). The test was administered at a viewing distance of 1 m under normal lighting conditions. CS was scored as the number of letters correctly read and was converted to a log CS. A one-way ANOVA revealed there were no differences between groups’ binocular CS [F(2,27) = 3.27, P > 0.05].

All participants self-reported right-leg dominance when asked their preferred leg for kicking a ball.

Data collection

Participants were instrumented with four spherical reflective markers (9 mm diameter) placed on each foot on the following anatomical landmarks: head of the first metatarsal, the dorsum (toe marker between second and third metatarsal heads), the head of the fifth metatarsal and the calcaneus. There was an additional reflective marker placed on the head of the third metatarsal on the right foot and a reflective marker placed on the sternum. These reflective markers were tracked using a Vicon motion analysis system (Oxford Metrics, England) at a sampling frequency of 120 Hz. Gaze behaviour was recorded using an ASL 500 mobile gaze tracking system at a sampling frequency of 120 Hz.

Protocol

Participants walked at a self-selected pace along an 8-m straight travel path and stopped at the end of the walkway. The walkway contained an array of light emitting diodes (LEDs) (10 mm in diameter) placed under a clear Perspex sheet that was flush to the floor, placed approximately 5 m from the starting position. At the beginning of each trial, participants stood at the start position and initiated walking on hearing a “Go” command. The instructions given to the participants were to walk at a self-selected pace along the straight travel path trying to keep at the same speed. They were further instructed that “if a light comes on, try to place your right foot as accurately as possible onto it (placing the marker on the third metatarsal on top of the illuminated LED) and continue walking until told to stop.” Participants were generally told to stop approximately 2 m past the end of the LEDs. Participants were also informed that during some trials an LED would not illuminate in which case they were required to continue walking to the end of the walkway.

Two sets of optoelectric timing gates (Metrodyne, Taiwan) were placed 1.5 m apart on the outside of the travel path and raised 30 cm from the ground. The timing gates were connected to a personal computer to control the timings of the LEDs. A LabView programme used signals from the timing gates to make online estimations of participants’ walking velocity and this information used to predict when the A/P position of the sternum marker would coincide with that of a particular LED. These timing estimates were then used to control the timing of LED presentation such that the targets illuminated 2,500, 1,500, 1,000 or 500 ms prior to participant arrival at the target area. Control trials in which no LED was illuminated were included to encourage participants to walk at a constant speed and deter them from attempting to guess which LED would illuminate.

In total, there were 48 LEDS placed under the Perspex sheet: 24 were placed in a central strip (central targets) and 24 were placed in a more lateral location approximately 125% of normal step width (lateral targets). LEDS were placed 10 cm apart in the anterior–posterior plane (see Fig. 1 for more detail). This procedure was implemented to ensure that the target location was unpredictable and therefore that successful completion of the task required visual guidance. All trials were randomised and a total of 70 trials were performed by all participants (10 no target, 30 central target and 30 lateral target trials)
https://static-content.springer.com/image/art%3A10.1007%2Fs00221-009-2058-0/MediaObjects/221_2009_2058_Fig1_HTML.gif
Fig. 1

Schematic diagram of experimental task

A helper was used during data collection to walk alongside each participant to provide physical support and prevent participants from falling if they became unstable. However, it should be noted that all participants were able to complete the task without the need of support from the helper. Prior to any data collection, all participants familiarised themselves with wearing the gaze tracker until they felt comfortable. Five practice trials were performed to allow the participants to familiarise themselves with the testing conditions.

Data analysis

All body kinematic raw data were passed through a low-pass FIR digital filter at a cut-off frequency of 6 Hz before displacement and velocities were calculated. The times of heel contact (HC) and toe off (TO) were determined using a technique described by Hreljac and Marshall (2000). HC was defined as a local maximum in calcaneus marker vertical acceleration determined by a zero crossing in the third derivative (jerk). TO was identified as a local minimum in toe vertical displacement, determined by zero crossings in the vertical component of toe velocity.

Data from all measured variables were grouped according to the interval between first fixation on the illuminated LED and HC at illuminated LED location, i.e. the total amount of time a foveal image of the illuminated LED was available prior to the foot landing on the target. The groups used were −1.0 to 0, −2.0 to −1.0 and −3.0 to −2.0 s. Fixation was defined as the stabilisation of gaze on a location in the environment for three frames (99.9 ms) or longer. Target fixation was assumed when the gaze cursor on the video image provided by the eye tracking system was positioned within one cursor width of the target image representing 2° accuracy (Vickers 1996). All spatial variables were calculated as a percentage of participant height prior to data analysis to control for the confounding effects of group differences in mean height. Mixed ANOVA with a 3 × 2 × 3 design (LED timing × laterality × participant group) were run on the following dependent variables:

(1) Saccadic reaction time measured as the interval between LED illumination and onset of the saccade towards the target, (2) number of gaze fixations following target illumination, (3) total target fixation time expressed as a percentage of time interval between first fixation and HC, (4) Mean walking velocity : prior to (pre-illumination) and following (post-illumination) LED illumination. (5) M/L and A/P distance between right foot and the LED target at targeting step onset (right TO) as an indication of participants’ path of progression with relation to the target and amplitude of required stepping action, (6) mean stepping error defined as the distance between the reflective marker placed on the third metatarsal on the right foot and the illuminated LED location (separate measures in medio-lateral and the anterior–posterior planes), (7) stepping error variability defined as the mean standard deviations of individual mean stepping errors across all trials for each condition, and (8) step width measured as the distance between the head of the first metatarsal on left foot and head of third metatarsal on right foot. All confidence levels were set at P < 0.05, and the statistical package, SPSS for Windows (release 14.0; SPSS, Chicago, IL, USA), was used for ANOVA and post hoc analysis where appropriate. Pair-wise comparisons post hoc tests (Bonferroni) were performed on main effects and interactions found to be significant. Only statistically significant results are detailed in the results section.

In order to visually describe any group differences in the fixation duration threshold beyond which errors begin to accumulate, raw stepping error scores from all participants and all trials were rank-ordered according to fixation duration (longest to shortest) and the data averaged into 100 ms timing bins, e.g. −3 to −2.9 and −2.9 to −2.8 s, etc. The cumulative sum (CUSUM) was calculated as the summation of the data for all of preceding data points, i.e. sum of the errors across all trials where the fixation duration was greater than or equal to the corresponding time value.
$$ S_{t} = S_{t - 1} + \bar{X}_{t} $$
where St represents the CUSUM for any given time bin (t) and \( \bar{X}_{t} \) the mean error scores across all trials and participants within a group for that time bin. For each group, CUSUM data series were low-pass filtered (1 Hz, dual-pass Butterworth), and plotted against fixation duration. CUSUM charts are useful for visually identifying the timing of changes in trends (Basseville and Nikiforov 1993). The principle behind a CUSUM chart is that it displays a running total of the differences from a target value (in this case zero error). Interpreting the chart depends on an examination of the direction and amplitude of the slope of the lines. If there are no systematic errors (i.e. zero bias) in foot placement then the slope of the line will be zero (flat). The fixation time at which the slope of the CUSUM plot starts to deviate from zero indicates the threshold of fixation duration below which systematic errors begin to accumulate. The direction of the slope reflects the direction of the error, e.g. lateral or medial.

Results

Gaze behaviour

All participants invariably made a saccadic eye movement to fixate the target following illumination. Mixed ANOVA revealed a main effect of laterality [F(2,54) = 21.1, P = 0.0001] and a main effect of group on saccadic latency [F(2,27) = 4.94, P = 0.009]. Saccadic latency was significantly longer in response to lateral target illumination than in response to central target illumination (central: 0.34 s, SD 0.12 s, lateral 0.42, SD 0.18 s). The mean saccadic latency of HROA was significantly greater than that of younger adults (younger adults: mean 0.33 s, SD 0.12, LROA mean 0.38 s, SD 0.16, HROA: mean 0.44 s, SD 0.17).

Mixed ANOVA revealed a significant interaction between LED timing condition and target laterality in the number of fixations participants made following target illumination [F(2,54) = 8.89, P = 0.0001]. In all cases the mean number of fixations was very close to one indicating that in the large majority of trials participants made a single saccade to fixate the stepping target following illumination and maintained fixation until HC (Table 2). However, post hoc analysis revealed that participants made significantly more fixations when stepping to lateral targets compared to when stepping to central targets (1.27 vs. 1.04 fixations) but only when the interval between initial fixation and HC with the target was longer than 2 s.
Table 2

Group averages describing gaze and stepping characteristics under the different LED timing conditions

Time between first fixation and heel contact (s):

−1 to 0 s

−2 to −1 s

−3 to −2 s

Target position:

Central

Lateral

Central

Lateral

Central

Lateral

Mean number of fixations

Mean

1.00

1.00

1.04

1.08

1.04

1.27

Standard deviation

0.00

0.02

0.14

0.13

0.14

0.37

N

30

30

30

30

30

30

% Time spent fixating target

Mean

100

99

99

97

99

92

Standard deviation

0

1

3

4

2

14

N

30

30

30

30

30

30

M/L adjustment required to step on target (%)

Mean

−0.3

18.6

0.1

17.0

1.2

13.8

Standard deviation

5.5

6.1

5.2

5.8

4.2

5.3

N

30

30

30

30

30

30

A/P adjustment required to step on target (%)

Mean

53.3

46.7

54.2

54.9

53.5

53.8

Standard deviation

11.3

10.5

11.2

11.4

12.3

15.0

N

30

30

30

30

30

30

There was a corresponding significant interaction between time condition and target laterality in the mean total time participants spent fixating the target, calculated as a percentage of time between initial target fixation and HC with the target [F(2,54) = 5.54, P = 0.006]. The percentage of time participants spent fixating the target was significantly reduced when stepping to lateral targets compared to when stepping to central targets (92 vs. 99%) but only when the interval between initial fixation and HC with the target was longer than 2 s (Table 2). There were no group main effects or interactions.

General walking behaviour

Walking velocity

Prior to LED illumination there were no group main effects or interactions in mean walking velocity (young 1.20 ms−1, SD 0.08, N = 60, LROA 1.21 ms−1, SD 0.17, N = 60, HROA 1.10 ms−1, SD 0.14, N = 60). Following LED illumination there was a small but significant main effect of participant group on walking velocity F(2,27) = 5.06, P = 0.014 (young 1.13 ms−1, SD 0.13, N = 60, LROA 1.07 ms−1, SD 0.18, N = 60, HROA 0.94 ms−1, SD 0.16, N = 60). Post hoc analysis revealed that, following LED illumination, on average, HROA group walked significantly slower than the younger adult group by circa 17%. There were no significant correlations between post-LED illumination walking velocity and any of our dependent measures.

Path of progression

There was a significant interaction effect between LED timing and target laterality [F(2,54) = 19.55, P = 0.0001] on the M/L distance between the right foot and the target at targeting step toe-off expressed as a percentage of participant height, i.e. the required M/L displacement varied with target laterality and LED timing due to participants changing their path of progression relative to the lateral targets. Post hoc analysis revealed that there were significant differences between LED timing conditions but only during steps towards lateral targets. In general, participants moved progressively closer to the lateral targets when given more time to respond to the LED targets (Table 2).

There was also a significant interaction between LED timing and laterality in A/P distance between the right foot and the target at targeting step onset (right toe-off) [F(2,54) = 19.55, P = 0.0001]. Post hoc analysis reveals that there were significant differences between LED timing conditions but only during steps towards lateral targets. When stepping to lateral targets, participants were significantly closer during the 0 to −1 condition than during the other two LED timing conditions (Table 2). There were no significant correlations between M/L or A/P stepping errors and required stepping adjustment (M/L: R = 0.18, P = 0.09, A/P: R = 0.16, P = 0.12). Therefore, any variation in stepping errors attributable to LED timing cannot be explained by the variation in the required stepping adjustments caused by differences in path progression.

Step width

There were no significant differences between groups in mean step width during control walks, i.e. when no target was presented (young 27.45 cm, SD 1.50 cm, LROA 27.50 cm, SD 2.53 cm, HROA 27.80 cm, SD 15.67 cm).

There was a significant interaction between LED timing and target laterality [F(2,54) = 28.57, P = 0.0001] on step width (expressed as a percentage of participant height) during test walks. Post hoc analysis revealed significant differences between the mean step width measured under each of the three LED timing conditions indicating that step width progressively increased as the fixation time decreased (Fig. 2b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00221-009-2058-0/MediaObjects/221_2009_2058_Fig2_HTML.gif
Fig. 2

Mean step width following the targeting step (expressed as a percentage of participant height) for steps to central (a) and lateral (b) targets. Error bars represent standard error of the mean (SEM). Dashed horizontal lines represent mean step width during control trials in which no targets were illuminated

There was a significant interaction effect between participant group and target location [F(2,54) = 5.7, P = 0.008] on step width. Post hoc analysis revealed differences between high-risk older adults and younger adults in step width only during steps to the lateral target location (Fig. 2b).

Number of steps taken

There were significant main effects of laterality [F(1,27) = 7.50, P = 0.011] and LED timing [F(2,54) = 718.37, P = 0.0001] on the mean number of steps taken following target illumination in order to reach the target. On average, participants took significantly more steps to reach targets that were presented later (i.e. were closer) than to targets presented earlier (0 to −1: mean 2.02, SD 0.36, N = 60; −1 to −2: mean 3.30, SD 0.40, N = 60; −2 to −3: mean 4.83, SD 0.63, N = 0). Participants also took significantly more steps to reach lateral targets (central: mean 3.33, SD 1.27, N = 90, lateral: mean 3.44, SD 1.22, N = 90). There were no group main effects or interactions.

Mean medio-lateral stepping error

Mixed ANOVA revealed there was a significant interaction effect between time condition and participant group on M/L stepping error [F(4,54) = 2.70, P = 0.04].

Post hoc analysis revealed significant differences in the size of Mean M/L stepping error between high-risk older adults and young adults under conditions in which they were able to fixate the target for longer than 1-s prior to target contact. There was a clear trend for participants, and particularly the high-risk older adult group, to step medially with respect to lateral targets (i.e. steps were too narrow). However, when only able to fixate the target for <1 s, on average, high-risk older adults made lateral stepping errors (i.e. steps were too wide) when stepping to central targets.

There was also a significant interaction effect between target laterality and participant group [F(2,27) = 5.1, P = 0.014]. Post hoc analysis revealed that both high- and low-risk older adults stepped significantly more medially with respect to the lateral target locations than young adults, i.e. older adults showed a systematic bias towards making steps that were too narrow when stepping to lateral targets (Fig. 3b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00221-009-2058-0/MediaObjects/221_2009_2058_Fig3_HTML.gif
Fig. 3

aBar charts showing effects of our experimental conditions on mean stepping error in the medio-lateral plane when stepping to central targets. Positive y-axis values correspond to lateral foot placement error (i.e. difference between displacement of a marker on the head of the third metatarsal on the right foot and the target) expressed as a percentage of participant height. Negative values correspond to medial stepping errors. Error bars represent standard error of the mean (SEM). b Shows the same data as Fig. 2a but for steps to lateral targets

There was also a significant interaction effect between time condition and target laterality on M/L stepping error [F(2,54) = 8.23, P = 0.001]. Post hoc analysis revealed significant differences between data obtained from the different target locations in mean M/L stepping error but only in the −1 to 0 condition (Fig. 3c). On average, when experiencing <1 s of target fixation, participants made medial stepping errors when stepping to lateral targets and lateral errors when steeping to central targets.

Cumulative sum of M/L stepping errors

In order to further investigate the interaction effect between group and fixation duration as revealed by ANOVA, we performed an analysis of the CUSUM of stepping errors (see “Methods” for details) for each group during steps to both central (Fig. 3a) and lateral targets (Fig. 3b). The purpose of this analysis was to determine the minimum fixation period below which stepping errors (i.e. foot placement bias) started to occur for each group.

When stepping to central targets (Fig. 4a) high-risk older adults show stepping bias to the medial side of the target (i.e. make narrow steps) even when able to fixate the target for 3 s. Medial stepping errors continue to accumulate at a consistent rate as fixation time decreases. When fixation time falls below 1 s high-risk older adult participants begin to make large lateral stepping errors (i.e. step wide). Low-risk older adult participants do not begin to make consistent stepping errors until fixation time falls to around 2 s at which point errors start to accumulate at a similar rate to that of high-risk older adults. When fixation time falls to around 1 s low-risk older adults also begin to make lateral stepping errors albeit at a slower rate than high-risk older adults. Finally, younger adult participants do not show consistent stepping errors until fixation time falls below around 1 s at which point they start to accumulate lateral stepping errors, but again at a lower rate than high-risk older adults.
https://static-content.springer.com/image/art%3A10.1007%2Fs00221-009-2058-0/MediaObjects/221_2009_2058_Fig4_HTML.gif
Fig. 4

The cumulative sum (CUSUM) of M/L stepping errors for each group during steps to both central (Fig. 3a) and lateral targets (Fig. 3b). The graphs show the minimum fixation period below which mean stepping errors (i.e. foot placement bias) started to occur for each group. The direction and amplitude of the slope of the lines denotes the direction and rate at which errors are accumulating as participants receive progressively less time to view the target prior to stepping. No systematic errors (i.e. zero bias) in foot placement error is represented by a flat line (slope of the line = zero). The fixation time at which the slope of the CUSUM plot starts to deviate from zero indicates the threshold of fixation duration below which systematic errors begin to accumulate. The direction of the slope reflects the direction of the error, e.g. positive represents lateral and negative represents medial errors

When stepping to lateral targets (Fig. 4b) high-risk older adults start to accumulate medial stepping errors when fixation times drop below around 2.7 s. The rate of error accumulation remains fairly consistent as fixation times fall, although there is a trend for error accumulation rate to increase as fixation time drop below 1 s. For both young and low-risk older adults, although there is a trend for some errors to accumulate below fixation durations of 2.5 s, the rate of accumulation is not consistent and the gradient of the CUSUM lines are relatively flat until fixation times fall below 0.8 s, beyond which point there is a clear increase in the accumulation rate of medial stepping errors. Therefore, there are clear differences between high-risk older adults and the other groups in the fixation duration threshold below which stepping errors start to accumulate. However, it is also clear that all participants show increased errors in stepping when receiving <1 s of target fixation prior to foot contact.

M/L variable error

There was a significant main effect of time condition [F(2,54) = 43.23, P = 0.0001] on M/L variable foot placement error expressed as a percentage of participant height. Post hoc analysis revealed that M/L variable foot placement error was significantly greater under the –1 s to 0 s time condition than under the other time conditions. Furthermore, all participants demonstrated significantly greater foot placement variability error under the –2.0 to –1.0 s time condition than under the –3.0 to –2.0 s time condition (–1 to 0 s: mean 0.077%, SD 0.020, N = 60; –2 to –1 s mean 0.063%, SD 0.014, N = 60, –3 to –2 s: mean 0.046%, SD 0.026 mm, N = 60).

There was a significant main effect of participant group [F(2,27) = 8.34, P = 0.002] and a significant interaction effect between target laterality and participant group [F(2,27) = 4.85, P = 0.016] on M/L variable foot placement error.

Post hoc analysis showed that high-risk older adults demonstrated significantly greater foot placement variability than younger adults when stepping to central targets and low-risk older adults demonstrated significantly greater foot placement variability than younger adults when stepping to lateral targets (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs00221-009-2058-0/MediaObjects/221_2009_2058_Fig5_HTML.gif
Fig. 5

Bar charts to show participant group differences in mean variable foot placement error in the medio-lateral plane for steps to central (Fig. 4a) and lateral (Fig. 4b) target locations, i.e. the standard deviation of M/L foot placement (expressed as % participant height) with respect to the target. Error bars represent standard error of the mean (SEM)

Mean anterior–posterior stepping error

There was a significant main effect of time condition on mean A/P stepping error, expressed as percentage of participant height [F(2,54) = 5.64, P = 0.006]. Post hoc analysis revealed that overall, participants demonstrated significantly greater A/P stepping error under the −1 to 0 condition than under the −2 to −1 condition (−1 to 0 s: mean 1.51%, SD 1.53; −2 to −1 s: mean 0.91%, SD 1.59; −3 to −2 s: mean 1.14%, SD 1.62). In all cases participants tended to step long with respect to the target LED.

Anterior–posterior stepping variability

There were significant main effects of both time condition [F(2,54) = 28.41, P = 0.0001] and target laterality [F(1,27) = 9.92, P = 0.004] on A–P variable foot placement error (expressed as percentage of participant height). Overall, participants were significantly more variable when stepping on central target locations (mean 0.083%, SD 0.025) than when stepping on lateral target locations (mean 0.075%, SD 0.029). Post hoc analysis revealed that participants demonstrated significantly higher A–P variable foot placement error under the −1 to 0 s time condition than the other time conditions. Furthermore, all participants showed significantly greater variability under the −2.0 to −1.0 s time condition than under the −3.0 to −2.0 s time condition (−1 to 0 s: mean 0.092%, SD 0.018; −2 to −1 s: mean 0.082% mm, SD 0.014; −3 to −2 s: mean 0.062%, SD 0.036).

Discussion

This study is the first to quantitatively and systematically assess how much time walking individuals need following initial fixation of a target in order to step accurately and also the first to identify differences in this measure between groups of young, high-risk and low-risk older adults.

Gaze behaviour

All participants invariably fixated a stepping target, typically a few hundred milliseconds after its illumination. Furthermore, participants rarely fixated anything else other than the target until after they stepped on it (see Table 2). This suggests that a foveal image of the target is important for the planning and execution of these visually guided stepping movements and that following initial target fixation any role of peripheral visual information describing the target will likely have been minimal. It is noteworthy that HROA show significantly longer saccadic reaction time than younger adults by circa 100 ms. These differences cannot be explained by general visual deficits since there were no differences between groups in our measures of acuity, CS or peripheral visual field sizes. Although there is some evidence that ageing has an adverse effect on saccadic reaction times the observed differences are usually small (Abrams et al. 1998; Kaneko et al. 2004; Munoz et al. 1998; Pratt et al. 2006). The current study is the first to identify a substantial increase in saccadic reaction time in older adults at increased risk of falling. This finding suggests that the neural structures in the oculomotor system responsible for generating visually triggered saccades such as visual occipital cortex, parietal cortex, the brainstem, burst generator, reticular formation, and superior colliculus (Munoz and Everling 2004) may be compromised in older adults and lend support to the notion that decline in visuomotor control may contribute towards falls risk.

Minimum fixation duration for accurate stepping

Our results demonstrated a statistically significant decline in stepping accuracy and precision for all participants when they were only able to fixate the target for <1 s characterised as a decrease in both M/L and A/P stepping accuracy and increased stepping variability.

When target fixation duration was <1 s participants made large medial errors when stepping to lateral targets and lateral errors when stepping to central targets (Fig. 3). Since the average step cycle duration (mean stance plus swing duration) is just over 1 s (1.07 s), these results suggest that walking individuals, irrespective of age, need at least one targeting foot step-cycle phase in order to step accurately. This constraint likely reflects biomechanical limitations, e.g. dynamic postural stability requires appropriate control of the centre of mass which is itself reliant on appropriate placement of the contra-lateral foot prior to the targeting step (see Patla et al. 1999; Hollands et al. 2001).

Age-related differences in visual sampling requirements

Previous research has shown that, when stepping onto targets or over obstacles, older adults choose to fixate stepping targets significantly earlier than younger adults (Di Fabio et al. 2003a, b; Chapman and Hollands 2006, 2007). The primary aim of this study was to investigate if this behaviour is necessitated by age- and falls-risk related differences in the length of time individuals need to look at a stepping target in order to step accurately onto it. Our working hypothesis was that older adults would require more time than younger adults visually fixating a stepping target in order to step accurately. Our findings clearly show that there are significant differences between our participant groups in the effects of manipulating the timing of visual cue appearance.

High-risk older adults were consistently less accurate than low-risk older and younger adults irrespective of fixation duration and tended to make medial stepping errors (Figs. 3, 4). One could argue that this participant group were simply less accurate than the other groups due to general motor or balance deficits or because they adopted a more cautious strategy due to fear of falling. However, this argument is not supported by our data which clearly shows that high-risk older adults are only marginally slower, take no more steps than the other groups (see Table 2), and are still able and willing to alter their step width by around 10% of their height (compared to step width during control walks) with <1-s notice (Fig. 2). In addition, the mean Berg Balance score of 43 obtained from HROA is only just below the cut-off score of 45 recommended as a predictor of falls risk (Berg et al. 1992). For these reasons, we do not believe that the increased errors exhibited by HROA are due to an inability to make the required magnitude of stepping adjustments or due to fear-related unwillingness to attempt the task. Furthermore, although all participant groups showed similar trends for increased stepping errors when receiving <1 s (see previous section) Fig. 3a clearly shows that mean M/L stepping error of high-risk older adults stepping to central targets is significantly greater than that of low-risk and younger adults during this experimental condition. This clear result is also evident in the CUSUM analysis which reveals that the rate of accumulation of lateral stepping errors when stepping to central targets (slope of CUSUM plots in Fig. 4) was much greater for high-risk older adults than for low-risk older or young adults.

In addition, our low-risk older adult group, whose scores on Berg Balance and ABC scales were similar to young adults (Table 1), only started to accumulate stepping errors when they were only able to fixate a stepping target for longer than 2 s prior to arrival at target (Fig. 4a). Poor balance or fear of falling cannot explain the differences between young and LROA. We interpret these findings as evidence that both high- and low-risk older adults need to fixate a stepping target earlier than younger adults in order to step as accurately. It is perhaps no coincidence that, given free reign, low and high-risk older adults choose to fixate a target for around 1.5 and 2 s, respectively, before stepping on it, whereas younger adults choose to look at the target significantly later resulting in a total mean fixation time of only 800 ms (taken from Chapman and Hollands 2006). The finding that low-risk older adults (who performed as well as younger adults on our functional mobility tests) require visual information about a target earlier than young adults in order to step as accurately suggest that the differences are due to age-related changes in visual or visuomotor processes rather than changes to motor or musculoskeletal function. This hypothesis is supported by the increased saccadic reaction times to target appearance as discussed above.

Age-related decline in cognitive function

An interesting result from our participant functional ability tests was that our high-risk older adult group scored significantly higher on the trail making B test than low-risk and older adults reflecting a decline in executive functioning, and/or psychomotor speed and visual scanning in this participant group (Lezak et al. 2004). Decline in cognitive and executive function associated with ageing is a strong predictor of falls (Bergland and Wyller 2004; Chen et al. 2005; Di Fabio et al. 2005) and associated with increased difficulty in making fast postural adjustments (Hauer et al. 2003). There is evidence that obstacle stepping, (Di Fabio et al. 2005) and even routine walking relies on cognitive and executive function to ensure appropriate body posture and consistent walking patterns (Hausdorff 2005). Therefore, it is possible that the reduced ability to make rapid stepping adjustments exhibited by our high-risk participants may be attributable to age-related decline in any number of CNS functions including, visuomotor control, executive functioning, psychomotor speed or visual scanning ability. Further experimentation is needed to elucidate the mechanisms underlying the behavioural changes we observed in the current study.

Age-dependent differences in stepping performance isolated to M/L plane

It is noteworthy that the differences between participant groups in the effects of our visual conditions were not seen in the A/P plane. The A/P foot placements of older adults (both high and low-risk) were not statistically different from younger adults. It would seem that the control of foot placement in the A/P plane is relatively unaffected by age or falls risk. This interpretation is consistent with the findings of previous research demonstrating that older adults generally show significantly greater step width variability, rather than step length variability, than younger adults during unobstructed walking (Owings and Grabiner 2004; Grabiner et al. 2001; Chapman and Hollands 2006—although there is evidence that risk of falling is associated with increased step length variability, e.g. Maki 1997; Verghese et al. 2009). Our results support the notion that age- and falls-risk related differences in the time needed to adjust stepping are mainly constrained to the M/L plane.

The findings that high-risk older adults made greater medial stepping errors than the other groups when stepping to lateral targets support the notion that high-risk older adults have problems controlling the medio-lateral components of their steps. This is not surprising since the large mass of the trunk and its high location relative to the feet during gait results in the need for substantial active dynamic stabilization in the lateral direction (Bauby and Kuo 2000); a process which is adversely affected by the ageing process. For example, it has been reported that when subjected to a lateral tilt of a platform on which they stood, the trunk of young participants moved in the direction opposite to that of the tilt within 30 ms (Allum et al. 2002). In contrast, the response latency of the older adults was >150 ms and the subsequent trunk motion was in the same direction of the impending fall. These age-related changes in lateral balance control are particularly relevant to the problem of falls since risk for hip fracture in the elderly is six times greater during sideways than forward or backward falls, and 30 times greater if the fall results in direct impact to the hip region (Nevitt and Cummings 1993). The results of the current study support the notion that a reduced ability to control the medio-lateral component of foot trajectories is a contributing factor to increased risk of falling in older adults.

Limitations

Ageing is associated with general decline in most physiological systems, the time-course and nature of which is subject to individual variation (Lord et al. 2007). Investigating CNS mechanisms underlying movement deficits in older adults in the presence of confounding decline in associated musculoskeletal and physiological function is a formidable challenge. Although we attempted to control for as many confounding variables as possible, there were a few potential confounds which were not addressed by our experimental design. The first relates to the fact that individuals walk with idiosyncratic gait characteristics such as walking speed, step length, etc. One consequence of differences in gait characteristics would have been that target illumination at any particular time prior to target arrival could have occurred at different stages of the gait cycle for each participant (e.g. single right stance, double support, etc.) and therefore the complexity for making adjustments to any single target illumination may have varied between participants. However, it should be noted that, on average, there were no significant differences between groups in our general measures of gait characteristics, i.e. number of steps taken, walking velocity prior to target illumination, stance and swing durations. Therefore, the confounding influence of differences in task requirements due to gait metrics will have been similar for all groups and is, therefore, unlikely to have affected our experimental findings.

Another potential confound is the age difference between our HROA and LROA groups which was around 6 years. One could argue that any differences between these groups are related to age rather than risk of falling. However, it is clear that age and falls-risk correlate very highly (Lord et al. 2007) and therefore the reason that the HROA group are at higher risk of falling than the LROA group may well relate to more advanced decline in general physiological due to their increased age. This does not affect our interpretation of experimental findings or conclusions.

Finally, it is relevant to discuss whether our results are generalisable to the general elderly population. Since our eye tracking system was not compatible with use of spectacles due to problems caused by reflections of near IR light from our motion analysis camera strobes, participants who needed spectacles for walking could not participate. There were four participants that wore corrective contact lenses during the experiment and all other participants wore spectacles, but only for reading. Only around 30% of over 65 s do not wear eye glasses at all and therefore our participants had unusually good eyesight for their age as confirmed by our visual screening tests. Although the visual function of participants was not necessarily representative of the general public, a useful byproduct of our methodological limitation was that we were able to control for the confound of general visual or optical problems which would have undoubtedly affected our results. Despite this limitation we were able to demonstrate significant differences between our participant groups which are unrelated to general decline in visual function.

Summary and conclusions

Older adults, and in particular high-risk older adults, require more time to initiate eye movements towards, and more time looking at, stepping targets than younger adults in order to plan and execute medio-lateral stepping adjustments. A reduced ability to make rapid stepping adjustments to avoid obstacles or step on safe areas may contribute towards trips and falls in these individuals. Possible mechanisms underlying this decline in performance include decline in visuomotor pathways in the CNS, reduced executive functioning and increased lateral instability caused by a decline in upper body postural control. Further biomechanical investigations of lateral stepping performance are required to gain a better understanding of the effects of ageing on neuromechanical function during locomotion.

Acknowledgments

This project was funded by a Strategic Promotion of Ageing Research Capacity (SPARC) Award awarded to Mark Hollands.

Copyright information

© Springer-Verlag 2009