Cognitive flexibility is part of the subdomain “shifting” of executive function (Miyake et al., 2000). It is controlled by the frontal lobe and associated areas (Miyake and Friedman, 2012), and influenced by aging (Wecker et al., 2005) and diseases, such as dementia (Stopford et al., 2012) and stroke (Rasquin et al., 2002). Cognitive flexibility is required for adapting behavior to external influences (Gilbert and Burgess, 2008; Klanker et al., 2013). This flexibility is necessary for the unrestricted performance of daily life during waking states, because it helps to make the right decisions in multitasking situations. Cognitive flexibility is often measured with the delta Trail-Making Test (dTMT) under experimental conditions (Ble et al., 2005; Bowie and Harvey, 2006; Coppin et al., 2006; Montero-Odasso et al., 2009; Hirota et al., 2010). Cognitive flexibility is likely associated with gait particularly during difficult walking situations, such as fast walking and walking when performing additional tasks, but this effect was not investigated in detail. This hypothesis arises from the following studies.

An investigation of 926 older community-dwelling persons with a mean age of 75 years using a single tasking (ST) 4 m walk at a convenient speed and a 7 m fast walk with obstacles (Ble et al., 2005) demonstrated an association between gait speed and cognitive flexibility, as measured with the dTMT in the fast walking with obstacles, but not in the convenient walking paradigm. This report was the first study to indicate an association of cognitive flexibility with gait speed in complex walking conditions in older adults (Ble et al., 2005; Coppin et al., 2006).

In another study of 493 Japanese individuals with a mean age of 74 years, the influence of cognitive flexibility on gait speed in particular under challenging walking conditions was also proposed (Hirota et al., 2010). The participants performed walking tasks of different levels of complexity, and the primary result revealed a stronger association of dTMT with the walking performance in more complex tasks compared to convenient walking.

Finally, the most convincing evidence for an interplay of cognitive flexibility and gait under dual tasking (DT) conditions comes from the Irish longitudinal Study on Aging (Killane et al., 2014). In this study, 4431 participants with a mean age of 62 years underwent walking tasks under ST and DT situations, and cognitive flexibility measures (i.e., the Color Trail Test) correlated significantly with gait speed only during DT, but not ST, situations.

Common in all of these studies is that they found associations of cognitive flexibility with gait. People with poor cognitive flexibility seem to have deficits in gait (control), especially in more difficult gait conditions. This might indicate that people with poor cognitive flexibility have a lower capacity to adapt to the demands of the more difficult walking condition. However, none of the above-mentioned studies reported other quantitative gait parameters beyond gait speed. As gait speed seems to be a sensitive but unspecific parameter to assess health in older age (Studenski et al., 2011), we were interested in whether cognitive flexibility is also associated with quantitative gait parameters “beyond” gait speed and with the “strategies” that are used in walking conditions with different levels of difficulty. This assessment can contribute to a mechanistic model of the interplay between cognitive flexibility and walking behavior.

This study evaluated quantitative gait parameters of four walking paradigms in a large cohort of healthy older adults wearing a small sensor unit at their lower back, and investigated the association of gait parameters and walking paradigms with cognitive flexibility. Recognizing and addressing such specific deficits in therapy and training settings may lead to improved gait performance in older adults with poor cognitive flexibility.

Our results partially confirm the results of previous studies (Ble et al., 2005; Coppin et al., 2006; Montero-Odasso et al., 2009; Hirota et al., 2010; Berryman et al., 2013; Killane et al., 2014): More challenging walking paradigms require more cognitive flexibility than simple paradigms. In our study, it is reflected by an increasing number of significantly different quantitative gait parameters with increasing difficulty of the walking condition. Accordingly, our observations suggest that the best walking task for the assessment of cognitive flexibility is the DT fast walk with subtracting serial 7s. This task exhibited the highest number of quantitative gait parameters that were significantly associated with the dTMT, and it also included the gait parameter that correlated strongest with the dTMT (gait speed). This observation may be simply explained by the increased influence of supraspinal control mechanisms on gait under more challenging walking conditions (Maetzler et al., 2013; Hobert et al., 2014). However, the obviously higher influence of cognitive flexibility on walking during subtracting serial 7s compared to walking when checking boxes requires further reflection. The control of a DT situation per se is a cognitive task, which means that a cognitive process is involved in sharing of resources or shifting of attention between the different tasks, and controls the prioritization of the tasks (Hobert et al., 2011). Based on this, we assumed that the following individual situations were present in our experimental setting. A study participant performed two tasks with mainly motor components (walking and checking boxes) during the checking boxes DT and one cognitive task (the above-mentioned cognitive control of the DT situation). A person during the serial subtraction task performed one task with a primarily motor component (walking) and two tasks with primarily cognitive components (serial subtraction and cognitive control of the DT situation). The bottleneck hypothesis proposes that the processing of two tasks using the same (or similar) network(s) create(s) a bottleneck (Ruthruff et al., 2001; Yogev-Seligmann et al., 2008). Therefore, a gait paradigm with two cognitive tasks performed in a (relatively) healthy cohort should exhibit “more” correlation with the dTMT than a gait paradigm with one cognitive task. This presumption was observed in our study.

Indirect support for the above hypothesis comes from studies investigating individuals with motor network deficits, e.g., patients with mild-to-moderate PD (i.e., at a disease stage, where the motor deficits are generally more prominent than the cognitive deficit). These patients might experience the performance of two motor tasks and one cognitive task as more challenging than the performance of one motor task and two cognitive tasks. Our prospective longitudinal study of PD patients with and without falls during an observation period of 3.5 years demonstrated that only DT deficits in the checking boxes task, but not in the subtracting serial 7s task, predicted the first fall in the former group (Heinzel et al., 2016). A recent study did not find any additional value of a cognitive DT paradigm on falls in 263 mild-to-moderate PD patients (Smulders et al., 2012). Unfortunately, the authors did not include a secondary task with a primarily motor component in their study protocol (Smulders et al., 2012).

After the association of cognitive flexibility with walking conditions, we analyzed the association of cognitive flexibility with different gait parameters: The parameter that was most closely associated with the dTMT was gait speed, followed by stride duration and number of steps. This result indicates that pace-associated parameters are more closely correlated with cognitive flexibility than variability-associated parameters, which is basically consistent with the results of a recent study (Martin et al., 2013). However, the authors in the previous study focused on the comparison of gait and executive function in general, and not specifically on cognitive flexibility (Martin et al., 2013). The result is still surprising, because one may associate cognitive flexibility with adaptation, rather than velocity aspects of gait. Regardless of the mechanisms for the “dominance” of gait speed over other quantitative gait parameters for, e.g., detection of gait deficits per se (Lord et al., 2013), motor-cognitive interference deficits (Al-Yahya et al., 2011), and aspects of general health and survival (Studenski et al., 2011), gait speed is a non-specific parameter. This fact is also true for the prediction of cognitive flexibility using this parameter. Gait speed (only) explained 7.4% of the variance of dTMT in the most challenging walking task of our setting.

As a general comment, single quantitative gait parameters may not reach sufficiently high prediction values for any kind of pathology or alterations of human movement that eventually enable an individual diagnosis. The more promising approach to differentiate between specific pathologies / alterations and control states and detect progression and changes due to therapy may be the use of “gait parameter panels” or multivariate regression models. Such models can account for the complex interplay between (dys)function and the compensation mechanisms that are involved in complex movements, such as gait and balance performance in particular under challenging conditions (Maetzler and Hausdorff, 2012; Lord et al., 2013; Maetzler et al., 2013).

This study demonstrated also that cognitive flexibility influences the “type” of walking pattern a person uses in distinct challenging walking situations. Good dTMT performers exhibited significantly increased variability, asymmetry, and irregularity of gait (Hausdorff et al., 2006; Plotnik et al., 2007; Montero-Odasso et al., 2011) in the subtracting serial 7s DT compared to the (easier) checking boxes DT. Notably, poor dTMT performers did not adapt their parameters accordingly, which indicates an impairment of this adaptation strategy. We interpret this finding as follows: Individuals with poor cognitive flexibility reach the maximum adaption capability of their walking pattern earlier than individuals with good cognitive flexibility in a sequence of walking tasks with increasing levels of complexity. A recent study (Lowry et al., 2012) indirectly supports our hypothesis. This study compared the gait parameters of walking along a figure 8 (where changes between walking patterns are necessary) with straight walking (no changes necessary) in 106 old adults and found a significant association between dTMT and the number of steps only when walking on the figure 8 (Lowry et al., 2012).

The present study has some limitations. First, quantitative gait parameters were assessed with a wearable sensor at the lower back. This technique may not be as accurate as more complex gait evaluation systems in measuring at least some of the gait parameters, e.g., double support time. Second, we analyzed approximately 14 m of steady state walking. Although other studies used even shorter distances (Ble et al., 2005; Montero-Odasso et al., 2011; Killane et al., 2014), longer walking distances may deliver more valid gait parameters. It has been shown that the reliability of gait parameters improves with increasing number of steps. This is the case for gait variability parameters (Galna et al., 2013), whereas gait speed reaches already the steady state after 2.5 m (Lindemann et al., 2008). Third, the different tasks were not randomized and may result in a learning effect, but this effect should be comparable in all groups.

This study demonstrates that cognitive flexibility is associated with walking, in particular under challenging walking conditions, in a cohort of older adults without relevant motor and cognitive deficits. We also demonstrated that older individuals with poor cognitive flexibility use a pattern in variability-related gait parameters across walking conditions that differs from individuals with good cognitive flexibility. This difference might indicate a lower capability of the former population to adapt to challenging walking situations with different demands. Our findings add relevant information to our understanding of gait and balance deficits in older adults with poor cognitive flexibility and may give a basis for interventional studies.

MH, DB, and WM made substantial contributions to the acquisition, analysis and interpretation of data for the work. MH, SM, SH, FM, US, and GE made substantial contributions to the acquisition of the data. MH and WM drafted the paper, all remaining authors revised the draft critically for important intellectual content. All authors gave their final approval of the version to be published, and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.