In everyday life, people perform complex cognitive tasks while walking through various environments. This has led to a variety of studies on how humans dual-task cognitive and locomotor movements. However, the results from these studies are mixed. Recent reviews highlight the conflicting results in regard to cognitive and motor performance during dual-tasking (Al-Yahya et al., 2011; Schaefer and Schumacher, 2011; Kelly et al., 2012b). Elderly individuals (Lindenberger et al., 2000; Simoni et al., 2013; Venema et al., 2013) and individuals with neurological deficits (Sheridan et al., 2003; Camicioli et al., 2006; Panyakaew and Bhidayasiri, 2013) tend to exhibit increased gait variability and decreased mental performance when dual-tasking walking and a cognitive task. Data from young, healthy subjects do not follow such a clear pattern. Some studies have shown cognitive-motor dual task cost in young, healthy adults (Grabiner and Troy, 2005; Al-Yahya et al., 2009; Verrel et al., 2009; Szturm et al., 2013), but other studies have found that young, healthy subjects either have no dual-task effect (Grubaugh and Rhea, 2014) or a reduced magnitude effect (Melzer and Oddsson, 2004; Siu et al., 2008; Srygley et al., 2009; Yogev-Seligmann et al., 2010). Furthermore, the observed dual-task effect on kinematic or kinetic variables often differs across studies.

Some of the discrepancies in dual-tasking results across studies may be related to the type of cognitive task chosen. The cognitive tasks that have been used for gait dual-tasking studies include an N-back task (Verrel et al., 2009), Stroop tasks (Melzer and Oddsson, 2004; Grabiner and Troy, 2005; Siu et al., 2008; Kelly et al., 2012a), a serial subtraction and a phoneme-monitoring task (Al-Yahya et al., 2009; Srygley et al., 2009), an N-back task and a spatial attention task (Nadkarni et al., 2010), a verbal fluency task (Yogev-Seligmann et al., 2010), and an automated operation span task (Grubaugh and Rhea, 2014). All of these tasks except the automated operation span task (Grubaugh and Rhea, 2014) had some effect on gait parameters. The N-back, Stroop, serial subtraction, verbal fluency, and automatic operation span tasks are all non-spatial working memory tasks. If walking is similar to upper limb motor tasks, it will engage cortical areas that are also involved in spatial working memory.

Work by Anguera et al. (2010) supports the idea that spatial working memory tasks overlap with sensorimotor brain areas. Their subjects performed a visuomotor adaptation task that involved manipulating a joystick to hit a visual target, and they also separately performed a spatial working memory task that involved mental rotation. Activation in the dorsolateral prefrontal cortex and the bilateral inferior parietal lobule overlapped between the motor and mental tasks. Moreover, spatial working memory performance predicted the rate of visuomotor adaptation. This suggests that upper limb visuomotor adaptation and spatial working memory likely share mental resources (Anguera et al., 2010).

Another factor that may result in discrepancies in dual-task findings across studies is walking speed. Walking speed fundamentally changes the dynamics of human gait. At faster speeds there is less time to make changes in limb movement for each step, and the body’s inertia is greater compared with slower speeds. These biomechanical changes are related to an increase in mechanical energy and passive dynamics at faster walking speeds compared with slower walking speeds (McGeer, 1993; Collins et al., 2005; Kuo, 2007). In addition, evidence from individuals with spinal cord injuries suggests that faster walking speeds rely more on spinal reflex pathways and spinal neural networks compared with slower walking speeds (Maegele et al., 2002; Beres-Jones and Harkema, 2004; Ferris et al., 2004; Behrman et al., 2005). Results from functional near-infrared spectroscopy indicate that humans have increased frontal brain activity at faster walking speeds compared with slower walking speeds (Harada et al., 2009). All of these data suggest that walking speed is likely to have an effect on brain activity during walking. By studying our subjects across a range of speeds, we could examine speed dependent differences in how the spatial working memory task affected brain activation and behavioral performance measures.

The purpose of this study was twofold. First, we wanted to determine if concurrent performance of a spatial working memory task affects human walking dynamics in young, healthy subjects across a range of walking speeds. Second, we wanted to determine if walking speed affects cognitive task performance and electrocortical activity during a working memory task across those same walking speeds. Reported differences between young and older subjects and across studies might be due to differences in the specific walking speeds studied or the fact that there was no control of walking speed across the cognitive task and no cognitive task conditions. We studied young, healthy subjects because they generally use a wide range of walking speeds in everyday life, and they are more cognitively capable compared with older subjects. We chose a Brooks spatial working memory task, because a balance task during standing disrupted performance on the spatial but not the non-spatial Brooks task (Kerr et al., 1985), and because there seems to be an overlap in spatial working memory brain regions and sensorimotor brain regions (Anguera et al., 2010).

Specifically, we had two hypotheses about cognitive and motor dynamics. We hypothesized that walking kinematics would show signs of stability challenges when subjects performed the spatial working memory task compared with when they had no cognitive task. Increases in gait variability or wider step widths are both signs that humans are adapting to stability challenges. We also hypothesized that walking at higher speeds would decrease the performance of the spatial working memory task and decrease the related electrocortical activity. If faster walking speeds require greater cortical attention for control, then cognitive performance and the related electrocortical activity should both decrease with walking speed. In this study, we measured event-related spectral power synchronized to the presentation of a stimulus as our metric of electrocortical activity across conditions. To address these questions, our subjects stood (0.0 m/s) and walked at a range of speeds (0.4–1.6 m/s) on a treadmill, with and without performing the Brooks spatial working memory task.

Our results indicate that walking speed does not affect spatial working memory task performance or task-related electrocortical activity in young, healthy subjects within the range of speeds measured here. Our subjects showed no change in response accuracy or reaction time for the Brooks spatial working memory task across walking speeds. Their success rate for the task was about 50% regardless of whether they were standing, walking slowly, or walking quickly. We also found no evidence of change in the electrocortical activity devoted to the cognitive task across walking speeds. The ERSP- graphs in Figures 6–11 reveal similar patterns for standing and all walking speeds (with the exception of the premotor and supplementary motor cortex, Figure 9). The mixed effects model showed no significant trend in spectral power change across speeds. If we had found a speed effect in either task performance or electrocortical activity without the other having a speed effect, our results would have been harder to interpret.

Previous studies have shown that performance on some types of cognitive tasks is affected by walking speed. Tomporowski and Audiffren (2013) showed that in older adults performing an executive processing task, response times decreased and response errors increased when walking at faster speeds compared with walking at slower speeds. When walking freely, young and older adults tend to decrease their walking speed when given cognitive tasks to perform during locomotion (Beauchet et al., 2002; Patel et al., 2014). This decrease in motor performance with dual-task walking and cognition suggests that there are common mental resources devoted to both tasks. Thus, we expected to see a change in cognitive performance and electrocortical activity as walking speed increased. However, the majority of studies reporting dual-task performance costs during locomotion have examined elderly or impaired populations. It may be that young, healthy subjects have an increased ability to walk at high speeds and perform a challenging cognitive task with no performance decrease compared with older subjects. In addition, the type of cognitive task may alter the relative amount of dual-task performance cost, as different cognitive tasks rely on different cortical substrates (Smith et al., 1996). When young, healthy subjects have demonstrated a dual-task performance cost (Melzer and Oddsson, 2004; Grabiner and Troy, 2005; Al-Yahya et al., 2009; Verrel et al., 2009; Szturm et al., 2013; Patel et al., 2014), the cognitive task was not a spatial working memory task.

Across all speeds, the addition of the spatial working memory task caused our subjects to step more widely compared to walking without a cognitive task. Wider steps have been associated with more stable gait (Bauby and Kuo, 2000; McAndrew Young and Dingwell, 2012). Past studies on young, healthy subjects performing cognitive tasks during walking have been inconsistent in finding step width changes (Grabiner and Troy, 2005; Siu et al., 2008; Al-Yahya et al., 2009). Our spatial working memory task’s significant effect on step width in young, healthy subjects could be related to its high difficulty level. Subjects correctly placed only about half of the digits at each speed, indicating that the task was very challenging. Patel et al. (2014) found that the complexity of the cognitive task influences the relative dual-task cost during gait. The lack of dual-task effects on gait kinematics for young, healthy subjects in past studies may be related to this difficulty/complexity effect or to the type of cognitive task, as we discussed above.

When we broke down the kinematic results into periods of encoding vs. retrieval for the cognitive task, we found that the retrieval period showed changes in gait indicative of decreased stability. Step width and step width variability were both greater in the phase of the cognitive task that required the subjects to push buttons on the hand-held device (retrieval) compared with the memorization period (encoding). The subjects also took shorter steps during the retrieval period compared to the encoding period. All of these changes in gait kinematics are traditionally seen by individuals that are older and have reduced gait stability (Dubost et al., 2006; Siu et al., 2008; Yogev-Seligmann et al., 2008; Granacher et al., 2011). The encoding task is not unlike many everyday walking tasks such as walking and texting, or walking and dialing a phone number. This suggests that in the real world, performing dual manual and locomotor tasks likely results in humans responding to stability challenges as well.

We found consistent electrocortical activity around stimulus encoding at all walking speeds in the somatosensory association cortex. ERSPs showed substantial alpha (8–13 Hz) synchronization preceding stimulus presentation and alpha desynchronization following stimulus presentation during the Brooks task. This area of the parietal lobe is involved in locating objects in space and is engaged during spatial working memory tasks (Carlesimo et al., 2001). We presented a stimulus every 4 s during the Brooks task. Alpha power in the left, right, and central portions of the somatosensory association cortex increased approximately 1 s before the presentation of the stimulus (a digit and its corresponding position in the grid) and remained elevated until shortly after stimulus presentation. Alpha power in the somatosensory association cortex decreased approximately 0.5 s after stimulus presentation. This could represent neural encoding of the digit and its position in working memory. The alpha power fluctuation was stronger in the right than in the left somatosensory association cortex, which may indicate that this brain area shows a right hemisphere dominance for locating objects in space. Eighteen of the twenty subjects were right-hand dominant.

We found small, significant theta power decreases around stimulus encoding in two brain areas: the right superior parietal lobule and the central posterior cingulate. Theta power modulations during memory tasks have been associated with memory encoding and maintenance by previous electrophysiology studies (Bastiaansen et al., 2002; Friese et al., 2012; Itthipuripat et al., 2013; Lenartowicz et al., 2014). Bastiaansen et al. (2002) found sustained theta power decreases in frontal electrodes during the retention phase of a spatial working memory task. Burke et al. (2013) also found decreases in theta phase synchrony during memory encoding. Many studies have shown theta increases in frontal brain regions during memory encoding and maintenance (Itthipuripat et al., 2013; Lenartowicz et al., 2014). However, we did not find any theta power increases in the frontal brain areas during performance of our task. It could be that during walking, these areas are engaged in gait-related processing, so cognitive task-related electrocortical shifts are undetectable within the gait related electrocortical activity. A host of functional near-infrared spectroscopy studies have found that the frontal cortical areas are highly engaged during human walking, even without a secondary task (Miyai et al., 2001; Suzuki et al., 2004; Harada et al., 2009). In addition, in a previous study from our lab, we also found electrocortical evidence of frontal cortical involvement in human walking without a cognitive task (Gwin et al., 2011). Our analysis of spectral perturbations in electrocortical activity was specifically limited to changes synchronized to stimulus presentation.

On a similar note, activity in the premotor and supplementary motor cortex may demonstrate some overlap between cognitive and motor processing (Figure 9). During standing, power in the lower alpha band increased around stimulus presentation, and alpha and beta power decreased shortly thereafter. During walking, this effect was not visible. Studies have shown that the premotor and supplementary motor cortex is engaged during walking (Gwin et al., 2011). During walking this area was likely engaged in motor processing, and cognitive-related activity was either not present or undetectable.

Some researchers have questioned how much mechanical artifact is present in walking EEG data synchronized to the gait cycle (Castermans et al., 2014). Previous studies have shown that EEG during walking is viable for non-gait synchronized data, as the approach of synchronizing EEG data to cognitive events not coupled to gait allows mechanical artifacts to wash out (Gramann et al., 2010). We observed the same pattern of neural activity at all speeds from standing (0.0 m/s) to very fast walking (1.6 m/s). This result indicates that EEG and ICA can reveal neural activity from a continuous cognitive task even when a subject is moving quickly. To our knowledge, our results are the first to show that high-density EEG and ICA can allow for the study of continuous cognitive dynamics at high walking speeds. This has important implications for the field of mobile brain imaging, as it suggests EEG could be used even in real-world environments that include walking at normal speeds.

Our study had some limitations that prevent us from drawing broad conclusions about the effects of speed on spatial working memory. First, we only examined young, healthy subjects. It is possible that we would have seen a larger effect in gait parameters and task performance if we had tested elderly subjects. Future studies should examine the neural and performance responses of elderly and neurologically impaired individuals. Second, we only examined a single cognitive task and a single cognitive task difficulty level. While unlikely, it is possible that an easier task or a harder task may have resulted in a different outcome. Third, the majority of the subjects in this study were males. There is some evidence that males may walk with increased variability under dual-task conditions. Hollman et al. (2011) had older adults walk and perform a backward spelling task. Both genders increased their variability when walking with the cognitive task, however, the variability increase during dual-task walking was greater in men than women. Finally, we did not assess our subjects’ preferred speed. A study by Schaefer et al. (2010) showed that children and young adults improved their cognitive performance on a working memory task while walking at their preferred speed on a treadmill. The ideal dual-task cost study would include a range of cognitive tasks and difficulty levels, to separate out overlap in neural substrates between walking and various cognitive tasks.

Our findings raise some interesting questions about dual-tasking. A popular view suggests that because walking requires mental effort, performing both walking and thinking can decrease thinking ability compared with thinking while standing still (Kahneman, 2011). Our results do not support that conclusion. For our spatial working memory task, mental performance did not change for walking vs. standing. Similarly, a recent study on the use of a treadmill desk found that human subjects performed equally well on a range of cognitive tasks for slow walking compared to standing (Alderman et al., 2013). John et al. (2009) also reported that subjects performing slow walking at a treadmill desk had no significant differences in selective attention, processing speed, or reading comprehension compared to sitting at a desk. Our study only examined a single cognitive task, but we found no detrimental effects over a wide range of walking speeds on cognitive ability.

Other researchers have suggested that moving your body may actually improve creativity. Slepian and Ambady (2012) found that subjects performed better on cognitive tasks related to creativity when moving their arms in fluid movements compared with not moving their arms. Longer term studies have found that chronic exercise also seems to increase creativity, but that is on the time scale of weeks or months (Flaherty, 2011). Putting together all of these observations with our own results, we suggest that people should not be concerned that walking while thinking will impair their cognitive performance. All in all, our study demonstrates that young, healthy subjects perform equally well on a challenging spatial working memory task during walking at a range of speeds as during standing. Given the many benefits of walking (Rippe et al., 1988; Hardman and Morris, 1998), people should be encouraged to walk and think whenever possible.

Daniel P. Ferris and Julia E. Kline designed the experiment. Julia E. Kline and Katherine Poggensee carried out all experimental procedures. Julia E. Kline performed the data analysis. Katherine Poggensee performed the majority of the statistical analysis. Daniel P. Ferris, Julia E. Kline, and Katherine Poggensee wrote and edited the manuscript.

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.