Dual-task walking on real-world surfaces: Adaptive changes in walking speed, step width and step height in young and older adults

Objectives: Age-related changes in dual-task walking are well established, but research in this topic is based on evidence from laboratory rather than real-world studies. We investigated how dual-task walking on real-world surfaces affects young and older adults ’ gait characteristics and cognitive resource allocation. Method: Sixteen young (aged 19 – 35, 12 female) and fifteen older adults (aged 70 – 85, 7 female) with no major neurological or musculoskeletal disorders walked at a self-selected speed on forty-metre outdoor paths that had asphalt or grass surface. They walked with or without a cognitive task (counting backwards). Cognitive task difficulty was individually adjusted at 80 % accuracy. Participants performed the three tasks in Single Task (ST Asphalt, ST Grass, ST Cognitive) and Dual Task context (DT Asphalt-Cognitive, DT Grass-Cognitive). Results: The two groups showed similar dual task effects in cognition and walking speed, both of which were slower when dual-task walking. Older adults ’ steps were wider overall but only young adults widened their step width when dual-task walking on grass compared to asphalt. Similarly, young adults ’ step height increased from single to dual-task walking when on grass, where older adults ’ did not. Discussion: The lack of adaptation of step width and height when dual-task walking may leave older adults vulnerable to tripping or falling in common real-world conditions, such as while walking on grass, gravel, or uneven city sidewalks. Considering this, the built environment should be made more accessible to facilitate older adults ’ safe walking.


Introduction
Walking in the built environment is a key form of physical activity for many urban-dwelling older adults, helping them to maintain health, wellbeing, and independence (Branco et al., 2015;Diehr and Hirsch, 2010). However, walking becomes increasingly challenging with age due to structural and functional changes in the brain, declining muscle strength and sensorimotor control (Seidler et al., 2010;Song & Geyer, 2018). These changes often lead to gait changes such as shorter, narrower strides, and increased gait variability (for review see Skiadopoulos et al., 2020). This pattern of gait characteristics may cause instability and falls (Lockhart et al., 2010) which are detrimental to older adults' health and wellbeing, and costly to health services (Florence et al., 2018;Terroso et al., 2014).
In addition to age-related decline, older adults' walking characteristics can be influenced, positively and negatively, by aspects of the built environment (Ghani et al., 2018;Rosenberg et al., 2013). On the positive side, areas with high residential and intersection density, green spaces, and good quality sidewalks are highly walkable and encourage physical activity (Lee and Dean, 2018; van Cauwenberg et al., 2016). However, on the negative side, areas with poor accessibility, narrow sidewalks, and obstacles can be perceived as challenging by older adults and have been associated with increased gait instability and fall risk (Hennah et al., 2021;Lee, 2021;Rosenberg et al., 2013). Considering the impact of sidewalk quality, irregular or compliant walking surfaces may be another challenging feature of the built environment.
Sidewalks and pathways in the built environment are typically made of either regular, even surfaces such as asphalt or concrete, or irregular and compliant surfaces such as gravel or grass (Gibbons, 1999). Little is known about the impact of irregular real-world surfaces on older adults' gait. Recent studies carried out in a real-world environment have shown that walking on irregular surfaces such as grass and gravel can result in more variable and less stable gait for young adults and those with diabetes (Allet et al., 2009;Nohelova et al., 2021;Ippersiel et al., 2021). In addition to high variability, walking on irregular surfaces such as grass, gravel, and woodchips has been related to higher step height and energy expenditure (Kowalski et al., 2021a(Kowalski et al., , 2021b, which may lead to gait instability and a potential fall risk. These studies focused primarily on younger adults, it is likely that irregular real-world surfaces will affect older adults to a greater extent as they may not be able to adapt their gait as well as younger adults (Hansson et al., 2021).
Extensive research into walking on uneven surfaces has been primarily carried out in a controlled laboratory setting. For example Voloshina et al. (2013) constructed an uneven surface on a treadmill and found that this surface reduced step length, challenged gait variability, and required more energy in young adults. Uneven surfaces were also associated with increased stride-to-stride variability in older adults, and high cadence variability (Menz et al., 2003;Thies et al., 2005). This increase in variability was interpreted as either due to surface compliance or to adjustments made by participants to enhance balance and stability. An increase in variability, however, is often associated with instability and increased risk of falls, and therefore real-world walking on uneven or irregular surfaces such as grass may challenge older adults' gait and present a fall risk.
Any difficulties that do arise from real-world walking on irregular surfaces may be further exacerbated during dual-task walking. Dual-task walking is often required in everyday life, such as when simultaneously walking while talking or following directions. Completing both tasks successfully requires sufficient attentional resources (Kahneman, 1973), however attentional capacity is thought to decline with age, and therefore dual-task walking can be challenging for older adults (Li et al., 2018;Madden, 2007). While young adults generally adapt well to dualtask walking and can allocate attention to both the walking and cognitive tasks, older adults may lack sufficient attentional resources to focus on both tasks, particularly if the tasks are complex (Woollacott and Shumway-Cook, 2002). If the walking task is particularly challenging, such as walking on an irregular surface like grass, task prioritisation may occur (Verghese et al., 2009) leading older adults to adopt a 'Posture First' strategy, ensuring they prioritise walking over the cognitive task (Kelly et al., 2013). This strategy can lead to impaired cognitive performance but ensures that gait remains relatively stable, thereby minimising fall risk (Vervoort et al., 2019). Similarly, task prioritisation is likely to occur when dual-task walking in a complex urban environment including irregular surface paths, where walking requires additional cognitive resources.
The aim of this study was to investigate how real-world surfaces and dual-task walking affect young and older adults' gait. We identified two 40 m flat outdoor path surfaces with different characteristics, an asphalt path (even, regular) and a grass path (irregular, compliant). The characteristics of these two surfaces were different, not only in terms of their physical properties, rigid vs compliant, but also in terms of the profile of gait characteristics when walking on them (Kowalski et al., 2021a). Firstly, based on previous research in young adults (Kowalski et al., 2021b) we predicted that walking on grass will induce slower gait speed, higher step height, and greater gait variability. Secondly, we expected that dual-task walking on these real-world surfaces will affect young and older adults differently. We predicted an age-related dual task deficit, with young adults' gait characteristics and cognitive performance remaining similar between single and dual task conditions, but older adults showing slower speed, increased gait variability, and decreased cognitive performance in the most challenging condition, dual-task walking on grass, which could be indicative of instability and increased fall risk.

Participants
A power analysis showed that a minimum of 11 participants per group would be required (α = 0.05, β = 0.08) based on data from a similar dual-task walking study (Nieborowska, 2018) to detect differences between young and older adults in single-and dual-task conditions. Our sample included 31 participants, 16 young adults aged 19-35 (age: M = 26, SD = 5.2; 12 female, 4 male) and 15 older adults aged 70-85 (age: M = 74, SD = 4.7; 7 female, 8 male). We chose to recruit older adults aged over 70 after failing to see considerable differences in dual-task walking between young and older adults aged 65+ in a previous study (Hennah et al., 2021). Participants were recruited through local advertising using posters and social media, and received an £8 Amazon voucher for participation.
Inclusion criteria were the ability to walk for five minutes unaided, and lack of major neurological conditions (e.g. Parkinson's disease or dementia), or medications that could affect gait or cognitive ability. To ensure eligibility participants completed a health questionnaire and several cognitive tests prior to participation (Table 1) including the Montreal Cognitive Assessment (MoCA) where the minimum score for participation was 26 out of 30 (Nasreddine et al., 2005), Digit Symbol Substitution and Digit Span tasks (Wechsler, 2008). The study protocol and risk assessment were approved by the Faculty of Engineering and Physical Sciences, Queen's University Belfast Research Ethics Committee (EPS 21_325).

Materials
The study took place outdoors, in a real-world setting. To minimise any potential hazards and unexpected distractors, the paths were situated away from major roads and busy pavements. Walking trials were set on two flat, straight paths that were already in existence, one of rigid asphalt and the other of compliant grass. Paths were 40 m long to allow enough steps for gait variability calculation. Throughout the study, gait was measured using Xsens AWINDA wireless IMU sensors (Roetenberg et al., 2009), used as directed by the manufacturer (MVN, 2021) and on the lower body only ( Fig. 1) to reduce the time participants spent in proximity of the researcher, thereby reducing the risk of Covid-19 transmission. Specifically we placed sensors on the feet, lower and upper legs, and pelvis and secured them with Velcro straps. The participant then walked approximately 5 m in a straight line to calibrate these sensors at the beginning of the study.
Participants were asked to perform the cognitive task of counting backwards from randomly selected three-digit numbers (Random.org Integer Generator) in decrements of either five, three, or seven. Based on pilot testing we considered five to be the least challenging task level, three moderately challenging, and seven the most challenging. Task difficulty was titrated to approximately 80 % accuracy based on performance in the counting backwards task performed at the beginning of the testing session. Participant responses were recorded throughout the study using a voice recorder (Sony Mono, BX series) for later transcription.

Procedure
This was a single-session study which lasted 45 min. Upon arrival, participants provided informed consent then completed a general health questionnaire to confirm eligibility, as well as the MoCA (Nasreddine  (Wechsler, 2008). The xsens sensors were then positioned and calibrated. Following this, the cognitive task titration procedure was administered. Participants were asked to count backwards from a random 3digit number, starting with counting backwards in fives, the simplest task, and potentially progressing through to threes, and finally sevens. To ensure a consistent standard across participants and age groups, we chose the most challenging level in which each participant could achieve at least 80 % accuracy and used this level during all cognitive single and dual task trials. In our young adult group, one person counted backwards in fives, seven counted in threes, and eight participants counted in sevens. In the older adult group, again one counted backwards in fives, nine in threes, and five in sevens. Once the individual difficulty level was established, participants completed the cognitive single task pre-test which required them to complete three standing trials at this predetermined level.
Participants then completed the walking trials which were structured in four blocks: Asphalt Single Task, Asphalt Dual Task, Grass Single Task, and Grass Dual Task, with four trials per block. All participants completed all blocks and the order of presentation was counterbalanced to control for any order effects of Surface or Dual Task Context. During dual task trials participants were asked to walk the 40 m asphalt or grass path at a self-selected pace while simultaneously carrying out the cognitive task at their individually pre-established level. During single task trials participants walked the paths at a self-selected pace, but counted from one to one hundred out loud, to control for verbalisation without significantly increasing cognitive load. Between each block of walking trials participants were given a two-minute seated break. After completing the walking trials all participants ended the study with a cognitive post-test, which consisted of a further three standing trials of the cognitive task and allowed us to assess for any improvement or fatigue effects.

Data analysis
We compared cognitive performance and gait characteristics under single and dual task conditions, while walking on both solid asphalt and compliant grass surfaces for 40 m.
We expressed cognitive results as Correct Response Rate (number of correct numbers per second). This was considered preferable to percentage accuracy because the measure of accuracy alone did not capture the rate of responding which was a critical aspect of performance in this task. Paired samples t-tests were used to assess pre-and post-test cognitive performance to check for any improvement or fatigue, and scores from these trials were averaged to give one Cognitive Single Task score. Cognitive scores from the four trials in each dual-task walking block were averaged to create one Cognitive Dual Task score for each block.
Gait characteristics were captured using the xsens AWINDA system and extracted from xsens MVN Studios software to .mat files to be read in MatLab. For analysis, the primary points of interest were the positiontime trajectories from the toes and pelvis. Pelvis position over time was used to calculate gait speed. Toe-off position in the x (anterior-posterior), y (medio-lateral) and z (vertical) axes was used to calculate mean and variability of step length, step width and step height. We analysed cognitive task performance and gait characteristics across conditions using mixed design Analyses of Variance (ANOVA), with age (young vs. older adults) as between-and context (single vs. dual) and surface (asphalt vs. grass) as within-subjects factors. All analyses were conducted using Jasp (v0.9) with an alpha level of 0.05. Interactions were interrogated using pair-wise t-tests with Holm-Bonferroni corrections for multiple comparisons.
Finally, to determine the magnitude of dual task effects in cognition, gait speed and step characteristics (length, width and height), and to adequately compare performance between the young and older adult groups, we calculated dual task costs according to the following equation (Doumas et al., 2008): In these measures a lower value was indicative of a performance decrement. Specifically, in the cognitive task fewer responses per second reflected a performance decrement, as did lower walking speed which is indicative of age-related decline (Weber, 2016;Studenski et al., 2011). Narrower gait was also identified as a performance decrement, because wider gait is considered more stable (Young & Dingwell, 2012;Ko et al., 2007). However, there is no evidence on step height and aging or falls, because most studies assessing step height focus on energy expenditure, which suggests that higher steps are less efficient (Kowalski et al., 2021a;Voloshina et al., 2013). However, we felt that lower steps may increase the possibility of trips and falls, thus, we defined lower step height as a performance decrement.
To calculate dual task costs for gait variability measures we reversed the single and dual task values in this equation. Dual task costs were analysed first using one sample t-tests to assess if they were significantly different from zero and then using Mixed-design ANOVAs.

Cognitive
Cognitive performance did not significantly differ from pre-test (M = 0.73, SD = 0.21) to post-test (M = 0.86, SD = 0.49), t(34) = − 1.89, p = .07. Response rate (correct items per second) was lower in dualcompared with single task performance as shown by a main effect of context F(1.03, 29.9) = 65.37, p < .001, η2 = 0.61 (Fig. 2, left panel). No other main effects or interactions were shown in this measure. Dual-task costs (Fig. 2, right panel) were significantly different from zero in all conditions (p < .001) but no group or condition differences were shown.

Gait characteristics: means
Gait Speed (Fig. 3, upper left panel) was slower when walking on grass compared to asphalt F(1,27) = 9.07, p = .006, η 2 = 0.03, and was also slower in dual-compared with single task performance F(1,27) = 24.119, p < .001, η 2 = 0.05. Dual-task costs for gait speed were significantly different from zero only for young adults on asphalt (p < .05) and for both groups on grass (p < .05). In all cases costs were positive, reflecting slower gait in dual task performance. No group or surface differences were shown for dual task costs in speed.
Step Height (Fig. 3, lower-left panel) was greater overall when walking on grass compared with asphalt F(1,22) = 5.36, p = .03, η 2 = 0.07. Furthermore, a surface by context by age interaction F(1,22) = 5.86, p = .024, η 2 = 0.03, interrogated using post-hoc t-tests showed that young but not older adults increased their step height from asphalt to grass only in the dual task condition t = 3.99, p holm = 0.008. Dual task costs were significantly different from zero only in young adults walking on asphalt (p = .042). An ANOVA on height dual task costs (Fig. 3, lower-right panel) revealed a surface by group interaction F(1,23) = 6.8, p = .016, η 2 = 0.1 showing that older adults exhibit no differences in costs between the two surfaces, but young adults show positive dual task costs when walking on asphalt, reflecting lower step height in dual task performance, but these costs are reduced when walking on grass t = 3.42, p holm = 0.014. Finally, step length showed no significant changes between groups or conditions.
Step length variability was also greater on grass F(1,23) = 6.85, p = .017, η 2 = 0.05 but no other main effects or interactions were shown in this measure. Finally, no differences were observed for step height variability.

Discussion
The aim of this study was to investigate how dual-task walking on asphalt and grass surfaces in the real-world affected young and older adults' gait characteristics and cognitive resource allocation. Based on a previous study (Kowalski et al., 2021a(Kowalski et al., , 2021b we predicted that walking on grass will induce slower gait speed, higher step height, and greater gait variability in young adults. Our results confirmed this prediction, not only in young (Kowalski et al., 2021a(Kowalski et al., , 2021b, but also in older adults. We also predicted that older adults' dual-task walking performance will be affected by our surface manipulation to a greater extent than young adults'. Results showed that surface affected young and older adults' step height differently. Young adults increased step height in dual-task walking on grass compared with asphalt, whereas older did not make any step height adjustments. Smaller age differences were observed for step width, and no age differences were shown for gait speed. Young adults increased their step height to cope with the additional demands of the grass surface when dual-task walking, but older adults did not make this adjustment. This finding could be due to age-related changes in musculoskeletal, metabolic or sensory factors affecting gait characteristics. On the musculoskeletal side, it is possible that this lack of step height adjustment is due to age-related decline in muscle strength and flexibility, with muscle strength loss reaching up to ~50 % in the 8th and 9th decade of life (Wilkinson et al., 2018). This decline may contribute to older adults' inability to lift their foot to take higher steps and adaptively change step height. Furthermore, on the metabolic side, higher steps have a greater energy expenditure cost, and perhaps older adults avoid an adaptive increase in step height to reduce this cost and conserve energy (Schrager et al., 2014). This energy cost would be especially high on a compliant path such as grass (Davies and Mackinnon, 2007), so the lack of adjustment by older adults observed in the present study may be a means of energy conservation. Finally, this finding can also be interpreted as due to a sensory integration deficit. Walking on a compliant surface like grass is likely to alter somatosensory information in a similar manner to the way other compliant surfaces like foam or sway referencing have been shown to alter such information primarily in postural control studies the laboratory (Horak, 1986;Nashner, 1982). Compliant surfaces alter somatosensory information by decreasing the reliability of information from plantar cutaneous mechanoreceptors (Horlings et al., 2008;Jeka et al., 2004;Wu and Chiang, 1997) and from ankle proprioceptive information (Horak et al., 2009;Peterka and Loughlin, 2004). This decrease in the reliability of cutaneous and proprioceptive information is exacerbated by age related decline (Craig and Doumas, 2019;Doumas and Krampe, 2010;Maitre and Paillard, 2016). In the present study, this reduction in the reliability of tactile and proprioceptive information with surface compliance (grass) and age is likely to have affected older adults' ability to determine the appropriate step height necessary for safe walking on a grass surface to a greater extent than young adults. Regardless of its origin, which requires further replication and investigation, this agedifferential adjustment in step height has functional implications for older adults because it is likely to contribute to fall risk, particularly on uneven or compliant paths where tripping is a hazard.
Our results for step width showed similarities but also differences from our findings for step height. On the one hand, similar to step height, young and older adults' results showed a different pattern but this time in dual task costs. Young adults' steps were narrower in dualcompared with single task walking on asphalt but wider in dualcompared to single task performance on grass. In contrast, older adults showed no dual task costs in this measure. The shift to a wider gait in the most challenging condition, dual-task walking on grass, suggests that young adults widened their gait to accommodate task demands. However, the finding of walking with narrower steps in dual-compared with single task performance on asphalt is not as easy to interpret. On the other hand, different from step height, we found that both groups showed an increase in step width when walking on grass, suggesting that both groups were able to adapt their gait by widening steps to increase stability and to accommodate the different characteristics of walking on the grass surface (Aboutorabi et al., 2015).
In contrast to step height and width, our results showed no age differences in gait speed and cognitive task performance. Both groups walked slower in single task walking on grass compared to asphalt, and even slower in dual-task walking, suggesting that older adults can adjust their gait speed in these conditions to the same degree as young adults. Similarly, in the cognitive task, we used a titration procedure to individually adjust task difficulty and to eliminate age differences at single task performance. We showed large dual task costs in cognition suggesting that a large proportion of cognitive resources dedicated to the cognitive task during single task performance were directed to walking during dual tasking. However, contrary to our initial prediction, we did not find any differences in cognitive performance between conditions nor evidence for task prioritisation in older adults. Task prioritisation in older adults would have been present if dual task costs in cognition increased and dual task costs on gait characteristics decreased on grass compared with asphalt, suggesting that an increase in gait difficulty directs cognitive resources to the gait task to support stability (Vervoort et al., 2019). However, cognitive costs were very high in both groups and in all conditions, suggesting that in our study gait was prioritised in all dual task conditions and there was little to no room for further investment of cognitive resources in this task.
This study had several limitations, firstly the participants' typical physical activity levels were not taken into account. According to Li et al. (2006), regular physical activity may help to reduce fall risk, and therefore physical activity levels could have affected these results. Secondly, gender was not matched between groups, with the young adults containing more female participants than the older adult group, which could have impacted attention flexibility and comparisons between groups (Yogev-Seligmann et al., 2010). Generally, studying dual-task walking in the real-world comes with a number of limitations, primarily a lack of control over the testing environment. This lack of control restricted our choice of cognitive task as counting backwards in fives, threes, or sevens was not ideal due to the large jumps in difficulty between levels, making individual titration of the task challenging. We would have preferred to use the auditory n-back task (Doumas and Krampe, 2015;Hennah et al., 2021) but were unable to, as we could not adequately control sound levels in the real-world environment without compromising participant safety. Additionally, disruptive intrusions from other people and wildlife were unavoidable, and adverse weather meant that many sessions had to be rescheduled or curtailed to maintain participant and researcher safety. Conditions on the grass path also varied from day-to-day due to rain or frost, and although no slips, trips or falls were observed, this reduced path consistency. Future research could further explore the effect of path surfaces on cognitive performance and gait stability while taking these limitations into consideration, such as using irregular surfaces such as gravel or woodchips in lieu of grass, as these likely present similar gait challenges but will maintain a consistent structure regardless of weather conditions.
In conclusion, this study shows that young and older adults both adapt their gait sufficiently when dual-task walking on a compliant grass path in the real-world, however older adults may have difficulty adapting their step height and step width. This lack of adaptation may mean that real-world gait becomes less stable with age and could potentially leave older adults vulnerable to tripping or falling when walking on grass, gravel or sand, especially if they are walking and performing a cognitively demanding task like having a conversation. Given the commonality of dual-task walking, and the considerable effect of the built environment on fall risk (Li et al., 2006), it would be useful if policy makers prioritised the accessibility of the built environment to facilitate older adults' walking. Future research should explore the musculoskeletal, metabolic and sensory factors contributing to older adults' inability to adjust gait characteristics on a compliant path, and to assess if task prioritisation is present when dual-task walking in a realworld environment to reduce their risk of falling.

Data availability
Data will be made available on request.