Content uploaded by Ulman Lindenberger
Author content
All content in this area was uploaded by Ulman Lindenberger
Content may be subject to copyright.
Cognitive performance is improved while walking:
Differences in cognitive–sensorimotor couplings between
children and young adults
Sabine Schaefer, Martin Lo
¨vde
´n, Birgit Wieckhorst and
Ulman Lindenberger
Max Planck Institute for Human Development, Berlin, Germany
We investigated how 9-year-olds and young adults performed a working
memory task under different difficulty conditions while walking on a treadmill.
Stride-length and stride-time variability tended to decrease with cognitive load
in young adults, whereas children showed an increase in walking variability
when cognitive load was very high. Participants in both age groups improved
their cognitive performance when walking at their preferred speed as opposed
to sitting or walking at a fixed, non-preferred speed. We conclude that the
interaction of walking and cognitive performance is influenced by sharing
resources between two tasks, and that performance improvements in cognition
may be caused by an exercise-induced activation of resources.
Keywords: Children; Dual-task performance; Gait variability; Physiological
arousal; Resource competition; Young adults.
INTRODUCTION
Dual-task situations, namely the simultaneous execution of two tasks, are
common in everyday life. Examples are reading the newspaper while having
breakfast or going for a walk while being engaged in a lively conversation
Correspondence should be addressed to Sabine Schaefer, Center for Lifespan Psychology,
Max Planck Institute for Human Development, Lentzeallee 94, D-14195 Berlin, Germany.
E-mail: schaefer@mpib-berlin.mpg.de
This study was conducted as part of the Sensorimotor-Cognitive Couplings Project at the
Center for Lifespan Psychology, Max Planck Institute for Human Development, Berlin,
Germany.
We thank Oliver Huxhold, Michael Schellenbach, and Julius Verrel for productive
discussions. Thanks also to Nina Blaskewitz, Gabi Faust, Simone Ku
¨hn, Anna Pohlmeyer,
Simone Ra
¨nker, and Tobias Rentz for their help in data collection.
EUROPEAN JOURNAL OF DEVELOPMENTAL PSYCHOLOGY
2010, 7 (3), 371–389
Ó2009 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/edp DOI: 10.1080/17405620802535666
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
with a friend. Often, one of the two tasks is a sensorimotor task (e.g.,
walking or keeping one’s balance on a bus), and the other task is a cognitive
task (e.g., trying to remember which groceries have to be bought on the way
home). Research concerning people’s ability to perform two tasks
concurrently has referred to these types of situations as dual-task or
multi-task situations, and resource theories have often been used to explain
the resulting behaviour. The underlying assumption is that attentional
resources are limited and have to be shared between the two tasks
(Kahneman, 1973; Wickens, 1991). Therefore, when the demands of the two
tasks exceed the individual’s processing resources, performance decrements
are to be expected, which are often referred to as dual-task costs. Note that
this line of reasoning assumes that the two tasks compete for the same pool
of limited resources. However, assuming that there might be multiple
resource pools instead, and that the amount of specific resources required
for performing certain tasks depends on the task characteristics of the two
component tasks, allows for the explanation of a larger variety of potential
findings in the dual-task research domain (Navon & Gopher, 1979). Under
the assumption of multiple resource pools, the execution of two concurrent
tasks only suffers if some of the resources required to perform one of the
tasks are also necessary to perform the other task.
Concerning age differences in the ability to perform two tasks
concurrently, children have often been reported to show larger
performance decrements in a dual-task situation than young adults (see
Guttentag, 1989, for a review). This pattern has been replicated in studies
with cognitive–sensorimotor dual-tasks (see Huang & Mercer, 2001, for a
review), and such situations have also often led to higher performance
decrements in older as opposed to younger adults (see Schaefer, Huxhold, &
Lindenberger, 2006, for a review). For example, Krampe, Schaefer,
Lindenberger, and Baltes (2008) had 9- and 11-year-olds and young adults
walk on a narrow track while concurrently performing a semantic word-
fluency task. Participants from all three age groups reduced their walking
speed to a comparable extent when performing the word-fluency task, but
only 9-year-olds additionally showed reductions in their word-fluency
performance, while 11-year-olds and young adults were able to maintain
their cognitive performance under dual-task conditions.
Not every cognitive–sensorimotor dual-task study, however, shows
higher overall dual-task costs in children as compared to young adults.
Schaefer, Krampe, Lindenberger, and Baltes (2008) conducted a study in
which 9- and 11-year-olds and young adults were balancing on an ankle-disc
board while concurrently performing either an episodic memory task or a
working memory task. Task difficulties of the cognitive tasks had been
adjusted individually to each participant’s performance level. Under dual-
task conditions, cognitive performance was reduced in all age groups, but
372 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
only adults showed more body sway on the ankle-disc board as compared to
the single-task condition. Children, on the other hand, reduced their body
sway under dual-task conditions, and therefore showed smaller dual-task-
related performance decrements when dual-task costs were aggregated over
the two task domains. This pattern of performance was interpreted as an
incidence of adaptive resource allocation in the children’s groups, since
children were operating closer to their stability boundaries already under
single-task conditions and protected their body’s equilibrium under dual-
task conditions by further reducing their body sway.
In general, resource theories (no matter whether they assume a general
resource or a pool of independent resources) assume that performing two
tasks concurrently will either lead to no changes in performance or to
performance decrements. However, there might be situations in which
performance is even improved in a dual-task situation. These cases have
usually not been interpreted within the dual-task research framework, but
have been explained in terms of an increase in ‘‘arousal’’ or ‘‘activation’’,
which may be caused by performing a sensorimotor task (e.g., exercising on
a bicycle ergometer) that leads to increases in physical arousal (Kahneman,
1973). This increase in arousal or activation, then, has beneficial
consequences for the performance of the secondary cognitive task, at least
as long as the exercise intensity is not approaching the individual’s
maximum capacity.
McMorris and Graydon (2000) reviewed several such studies. According
to their review, increments in speed of cognition with exercise were observed
in several studies, and performance was sometimes further improved when
participants were exercising with maximum intensity, especially when the
cognitive task was rather complex. However, accuracy of cognitive
performance—as opposed to speed—has generally not been improved
significantly by incremental exercise. It is therefore assumed that individuals
are able to allocate resources to task-relevant information, even during
maximal exercise. McMorris and Graydon further mentioned
methodological concerns that questioned the validity of several of the
studies, namely the fact that exercise protocols usually did not take
individual differences in fitness levels into account, and that cognitive tasks
differed considerably in complexity and familiarity.
Other studies not included in the overview article by McMorris and
Graydon (2000) further support the assumption of cognitive performance
enhancements with exercise (Adam, Teeken, Ypelaar, Verstappen, & Paas,
1997; Collardeau et al., 2001). For example, Collardeau et al. (2001)
conducted a study in which trained triathletes were running on a treadmill at
a velocity that corresponded to their ventilatory threshold. In the testing
session, five 10 minute submaximal treadmill runs were separated by four
over-ground runs (10 minutes each). Before, during, and after running on
COGNITIVE PERFORMANCE WHILE WALKING 373
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
the treadmill, the athletes performed a simple reaction time task. Reaction
times were slowed compared to the pre-test performance in the first
treadmill run only. After that run, a significant effect of exercise duration on
simple reaction time was observed, with significantly faster reaction times
after 40 minutes of exercise and thereafter. This effect is unlikely to have
been caused exclusively by a practice effect, since pre- and post-test reaction
times that were recorded at rest did not differ significantly. The authors
suggested that simple cognitive performance can be improved during
exercise, despite the negative effect of dual-tasking, and they assumed that
the underlying reason for this improvement was an increase in arousal
induced by a prolonged exercise.
In addition, it has been investigated whether there are age differences
between children of different ages and young adults in the ability to profit
from an exercise-induced arousal. Deakin et al. (1988) had 8-year-olds,
11-year-olds, and young adults perform a visual detection task at rest and
while walking at 75% of their maximum heart rate. Over and above the
main effect of age group, with 8-year-olds’ visual detection being inferior to
11-year-olds’ and young adults’ performance, responses were quicker at all
levels of distraction when targets were present while participants were
exercising. This effect was most apparent when many distracting items were
presented in the display, and it did not interact with age. This indicates that
children are also able to profit from exercise and can improve their cognitive
performance due to exercise-induced arousal.
The current study investigated whether walking on a treadmill at either a
fixed or a self-chosen speed while concurrently performing a cognitive task
(N-back working memory task) with different levels of task difficulty would
lead to decreases or increases in cognitive performance as compared to a
baseline condition in which the cognitive task was performed while sitting
on a chair. We also investigated whether 9-year-olds and young adults
differed in these dual-task performance patterns. A previous study using the
same experimental paradigm in young and older adults indicated that
cognitive performance did not change between single- and dual-task
conditions when participants were walking at their self-chosen speed
(Lo
¨vde
´n, Schaefer, Pohlmeyer, & Lindenberger, 2008). Decreases in
performance would be predicted by a resource account that assumes that
the two tasks compete for limited resources, while performance increases
would be predicted if one assumes that sensorimotor performance leads to
an increase in arousal or activation, which then can be invested into the
cognitive task. Concerning age differences, a resource account would predict
that children have fewer resources available than young adults, so that their
dual-task-related performance decrements are greater than those observed
in young adults. Under the assumption that cognitive performance profits
from an increase in arousal or activation associated with the sensorimotor
374 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
task, it is an open question whether this effect should be more pronounced in
children or young adults.
Li, Krampe, and Bondar (2005) discussed four methodological concerns
that should be taken into account when conducting cognitive–sensorimotor
dual-task research (see also Lindenberger, Marsiske, & Baltes, 2000). They
argued that the tasks that are used should mimic everyday processing
demands in multi-tasking. We tried to incorporate this concern in the
current study by having participants walk on a treadmill while being
engaged in a working memory task. We used the N-back task, which taps
central cognitive functions, like the monitoring, short-term storage, and
scheduled retrieval of the presented stimuli, and which also involves
executive control processes (Dobbs & Rule, 1989). Second, Li et al. argued
that co-ordination costs should be calculated for all tasks involved. Since we
measured cognitive performance as well as several gait parameters such as
stride time and stride length, we were able to investigate dual-task-related
changes in both task domains. The third methodological concern states that
task difficulties should be varied systematically to challenge individuals’
potentials at an age-appropriate level. We included a difficulty manipulation
of the two tasks in our design (i.e., performance on the N-back was assessed
at four different difficulty levels, and walking on the treadmill was
performed at a fixed speed of 2.5 km/h and at a self-chosen speed).
Finally, Li and colleagues argued that differential-emphasis manipulations
should be used, in which participants are instructed to exert control over
their resource allocation operations, by shifting their attention more to one
task than to the other. We refrained from taking this consideration into
account, since it is unclear whether shifting one’s attention to the
performance of a rather automatized task such as walking has detrimental
or beneficial consequences for task performance (Huxhold, Li, Schmiedek,
& Lindenberger, 2006).
METHOD
Participants
Thirty-two 9-year-olds (M
age
¼9.0 years; SD
age
¼0.18 years; 16 girls and 16
boys) and 32 younger adults (M
age
¼25.3 years; SD
age
¼2.89 years; 16
women and 16 men) were sampled from the participant pool of the Max
Planck Institute for Human Development, Berlin, Germany. Exclusion
criteria for young adults were diagnosed disorders directly affecting gait and
balance, diabetes and diagnosed heart problems. Children with orthopaedic
disorders, balance problems (e.g., acute or chronic ear infections),
meningitis, attention-deficit-hyperactivity syndrome (ADHS) or with
neurological disorders (e.g., cerebral palsy, brain tumours, epilepsy) were
COGNITIVE PERFORMANCE WHILE WALKING 375
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
excluded. All participants had normal or corrected-to-normal vision and
hearing. Parts of the data from the group of young adults have been
previously reported by Lo
¨vde
´n et al. (2008) as comparison data for a sample
of older adults.
Perceptual speed (Digit-Symbol Substitution; Wechsler, 1981) and
vocabulary (MWT-A; Lehrl, Merz, Burkard, & Fischer, 1991) were assessed
to document cognitive age typicality of the sample. For the item-per-second
score of the Digit-Symbol Substitution Test, a t-test revealed that young
adults scored significantly more items per second than 9-year-olds,
t(39.87) ¼15.2, p5.001. Furthermore, young adults had significantly
higher scores than 9-year-olds on the vocabulary test, as indicated by a
t-test for independent samples with t(61) ¼10.0, p5.001. The overall
picture on these cognitive measures is consistent with the developmental
literature showing that cognitive speed and knowledge of vocabulary
improve during development (e.g., Case, Kurland, & Goldberg, 1982; Fry &
Hale, 1996).
Participants provided written consent prior to the experiment and
received e50 for their participation. Parents provided written consent for
their children. The study was approved by the ethics committee of the Max
Planck Institute for Human Development.
Apparatus
A 12-camera (infrared V-cam 100 & 200) Vicon motion capture system
(Vicon 612, Workstation 4.6; Vicon Ltd, Oxford, UK), sampling at 200 Hz,
was used for capturing participants’ motion on the treadmill. Gym shoes in
all sizes were prepared with reflector markers at the big toes and the heels,
and an additional 35 reflector markers were attached to participants’ bodies
for the motion analyses.
Participants walked on a treadmill (Woodway GmbH, Weil am Rhein,
Germany) which had its walking area (200670 cm) at the level of the
surrounding floor. No handrail was present. In order to prevent complete
falls, a safety harness was fastened around the waist of the participant and
to the ceiling.
A2006270 cm screen was mounted in front of the treadmill. A virtual
environment consisting of a straight path was back projected to the
screen (175 6234 cm). The visual flow of the virtual environment was
synchronized to the speed of the treadmill with an empirically established
flow/speed ratio (see also Lo
¨vde
´n, Schellenbach, Grossmann-Hutter,
Kru
¨ger, & Lindenberger, 2005). The reason for projecting the virtual
environment was to reduce the risk that participants unnaturally focused
on the experimental environment while avoiding fixation at one point in
space.
376 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Design and procedure
The experimental design for the cognitive measures was a 2 (Age Group:
children/young adults) 62 (Setting: sitting/walking) 62 (Walking Speed:
fixed speed/preferred speed) 64 (WM Load: 1-back to 4-back) factors
design. For the sensorimotor variables, the design was a 2 (Age Group:
children/young adults) 62 (Walking Speed: fixed speed/preferred speed) 6
5 (WM Load: no load to 1-back to 4-back) factors design. The sensorimotor
variables of interest in the current study were the coefficient of variation for
stride time and stride length. The coefficient of variation (CV) is the
standard deviation divided by the mean—CV ¼(standard deviation/
mean) 6100—and it is often used in cases when variances of distributions
with different means are compared, and when mean value and variance are
dependent on each other (Beauchet, Dubost, Herrmann, & Kressig, 2005;
Ebersbach, Dimitrijevic, & Poewe, 1995). Stride length is defined as the
distance covered in two steps, which is from one heel strike to the next heel
strike of the same foot. Accordingly, stride time is defined as the time to
perform one stride. Stride-to-stride fluctuations in walking are assumed to
be a sensitive marker for gait control and gait stability, with higher
variability representing a less consistent and less stable walking pattern
(Hausdorff, 2005).
Data collection consisted of two sessions. One session involved walking
at a fixed speed (2.5 km/h) and one involved walking at a preferred speed.
The order of sessions was counterbalanced. Anthropometric data (e.g.,
height, weight) was obtained in the beginning of the first session. Cognitive
ability measures were assessed at the beginning of the second session.
Participants were tested individually in sessions that lasted approximately
two hours.
In each session, participants were given at least seven minutes to become
familiar with the treadmill. Different walking velocities were introduced. In
the ‘‘preferred speed’’ condition, the participant selected a tempo allowing
natural and rhythmic walking with which they felt comfortable. Otherwise
the fixed speed was introduced for at least three minutes. The participants
completed the four conditions of the single-task N-back (1-, 2-, 3-, and
4-back) while sitting comfortably on a chair. Next, the single-task walking
(no load condition) was carried out, during which the motion capture
started 30 seconds after the participants had started walking on the
treadmill. Motion capture lasted for about 20 seconds. Then the
participant walked and performed the N-back task simultaneously
(dual-task condition). Depending on the counterbalancing condition, the
dual-task phase started with N-back 1 or N-back 4. Each N-back task was
repeated twice. The auditory N-back started 30 seconds after the participant
had started walking on the treadmill. Motion capture started 39 seconds
COGNITIVE PERFORMANCE WHILE WALKING 377
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
after the auditory N-back had started and lasted for about 21 seconds.
Between the first and second trial there was an interval of 30 seconds. The
subject did not stop walking on the treadmill during that time. Finally, the
second trial of single-task N-back 4, 3, 2, and 1 was performed while sitting
on a chair.
The N-back tasks involved series of 26 digits from 1 to 9 presented via
loudspeakers. Depending on the condition of WM load, participants
verbally responded with ‘‘yes’’ whenever a digit was heard that was the same
as the one presented 1, 2, 3, or 4 positions back in the series. All responses
were noted by the experimenter and later scored as correct or false alarms.
Six ‘‘yes’’ responses were required in each trial for perfect performance, so
the maximum score was 6 points for each trial. One point was subtracted
from the score for each false alarm. The inter-stimulus interval varied
randomly between 2350 and 3150 milliseconds to prevent rhythmical
influences on gait. To minimize articulation during the motion capture, none
of the six target items occurred in the 20-second interval in which motion
was captured (item 13–item 21).
Data processing and statistical analyses
The motion data (heel and toe markers) were post-processed in MATLAB
6.5 (Mathworks, Sherborn, MA). From the 20 seconds of motion data from
each trial, the first and last 2.5 seconds were discarded due to unreliable
data. From the remaining 15 seconds, 9 gait cycles were extracted for all
individuals. Mean and coefficient of variation for stride time and stride
length over the 18 steps were computed separately for the left and right foot
and the two trials of each condition. Subsequently, the estimates for left and
right foot were averaged. A few (53%) outliers (+3SDs) were replaced by
the values of the other trial in a condition before averaging across the two
trials of each condition.
Mixed ANOVAs were performed separately for the dependent variables.
For the sensorimotor variables, a 2 (Age Group: younger/older) 62
(Walking Speed: preferred/fixed) 65 (Task: no load–4-back) ANOVA
addressed the effects of walking with no load as compared to walking under
different difficulty levels of the cognitive task and the influence of walking
speed. WM load and task were within-subject factors. For WM
performance, a 2 (Age Group: younger/older) 62 (Walking Speed:
preferred/fixed) 62 (Setting: sitting/walking) 64 (WM Load: 1-back to
4-back) ANOVA was performed. WM load, walking speed, and setting were
within-subject factors. For within-subject effects, the multivariate F-values
are reported. In cases in which variances were not equal in t-tests or in which
the sphericity assumption was not met in mixed-design ANOVAs, corrected
degrees of freedom have been used. The alpha level was .05.
378 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
RESULTS
Walking behaviour
Young adults preferred a faster walking velocity (M¼3.73 km/h, SD ¼
0.39) than 9-year-old children (M¼3.09 km/h, SD ¼0.41), t(61.86) ¼6.4,
p5.001. In addition, the self-selected walking velocity of both 9-year-old
children and young adults was significantly faster than the given fixed-speed
velocity of 2.5 km/h. Concerning age differences in mean values of gait
parameters, children showed shorter stride lengths, higher cadences, and
briefer stride times than young adults in both speed conditions. This pattern
has often been reported in the literature (Adolph, Vereijken, & Shrout, 2003;
Hausdorff, Zemany, Peng, & Goldberger, 1999). Since participants in the
current study were not able to change their walking speed during a trial due
to the constant speed of the treadmill, analyses of gait parameters will focus
on the variability of stride time and stride length. Stride-to-stride variability
is considered to be a sensitive marker for gait control, and gait instability is
characterized by increased variability from one stride to the next (Hausdorff,
2005).
Figures 1 and 2 depict the results for stride time and stride length as
function of working-memory load under preferred and fixed-speed
conditions. An inspection of these figures suggests that young adults tend
to decrease the variability of their gait with increasing cognitive load,
Figure 1. Age differences in the influence of cognitive load on stride time variability as measured
by the coefficient of variation (CV). Children’s stride time variability is depicted by squares, and
young adults’ stride time variability is depicted by circles. Solid lines represent performance in
the fixed-speed condition, and dashed lines represent performance in the preferred-speed
condition. Error bars depict standard errors of the mean.
COGNITIVE PERFORMANCE WHILE WALKING 379
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
especially in the preferred-speed condition, while the pattern is less
consistent for children, who sometimes tend to show a rather U-shaped
function, with increases in gait variability when cognitive load is very high.
In order to address this pattern of data statistically, a mixed-design
ANOVA with Cognitive Load (5: no load, N-back 1 to 4) and Speed
(2: preferred speed vs. fixed speed) as within-subjects factors and Age Group
(2) as between-subjects factor was conducted for stride time and stride
length. Since preliminary analyses showed no main effect for sex in the
current study, this factor was not considered.
A significant effect for age group was detected for stride time,
F(1, 60) ¼128.9, p5.001, Z
2
¼.68, and stride length, F(1, 60) ¼159.2,
p5.001, Z
2
¼.73. Children showed a higher variability in those gait
parameters as compared to young adults, a pattern that has often been
reported in the literature (Hausdorff et al., 1999; Stolze et al., 1997;
Sutherland, 1997).
A significant Cognitive Load effect was detected for stride length, F(3.43,
206.03) ¼11.5, p5.001, Z
2
¼.16, but this effect was qualified by an
interaction of Cognitive Load and Age Group, F(3.43, 206.03) ¼5.4,
p¼.001, Z
2
¼.16, with children showing more pronounced differences
between the load conditions than young adults. These effects were not
present for stride time.
A main effect for the factor Speed was found for stride time,
F(1, 60) ¼79.6, p5.001, Z
2
¼.57, and stride length, F(1, 60) ¼50.1,
p5.001, Z
2
¼.46. Participants showed a higher variability in those gait
parameters under fixed-speed conditions as compared to preferred-speed
Figure 2. Age differences in the influence of cognitive load on stride length variability as
measured by the coefficient of variation (CV). Children’s stride time variability is depicted by
squares, and young adults’ stride time variability is depicted by circles. Solid lines represent
performance in the fixed-speed condition, and dashed lines represent performance in the
preferred-speed condition. Error bars depict standard errors of the mean.
380 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
conditions, which might have been caused by a decrease in postural stability
at lower speeds (Shumway-Cook & Woollacott, 2001). The main effect of
Speed did not interact with Age Group for any gait parameter. For
stride time, a significant interaction of Cognitive Load and Speed,
F(2.77, 166.22) ¼3.1, p5.05, Z
2
¼.05, and also an interaction of Cognitive
Load, Speed and Age Group, F(2.77, 166.22) ¼3.4, p5.05, Z
2
¼.05, were
detected.
To follow-up these significant interactions involving walking speed and
task condition, separate analyses were conducted for the preferred- and
fixed-speed conditions in children and young adults. To address non-linear
trends we inspected the orthogonal polynomial contrasts. Results showed a
significant linear trend in the preferred-speed condition for stride time,
F(1, 31) ¼30.6, p5.001, Z
2
¼.50, and stride length, F(1, 31) ¼10.7, p5.01,
Z
2
¼.26, in the young adults, and the quadratic trend reached significance
for stride length in children, F(1, 29) ¼6.4, p5.05, Z
2
¼.18. For the fixed-
speed condition, the pattern was less consistent, with both the linear and the
quadratic trend reaching significance for stride time and stride length in the
young adults, linear trend for stride time: F(1, 31) ¼10.8, p5.01, Z
2
¼.26;
quadratic trend for stride time: F(1, 31) ¼5.8, p5.05, Z
2
¼.16; linear trend
for stride length: F(1, 31) ¼6.2, p5.05, Z
2
¼.17; quadratic trend for stride
length: F(1, 31) ¼6.3, p5.05, Z
2
¼.17, and only the quadratic trend for
stride length reaching significance in the children’s sample, F(1, 31) ¼6.3,
p5.05, Z
2
¼.17.
In sum, variability of stride time and stride length was higher in children
than in young adults, and it was higher under fixed-speed conditions than
preferred-speed conditions. Moreover, young adults tended to decrease their
gait variability as a function of cognitive load, whereas children decreased
their gait variability when working on an easy cognitive task (1-back and
2-back), but increased the variability again when cognitive load was very
high (3-back and 4-back). However, the pattern was not consistent across all
walking velocities and for both gait parameters.
Cognition
In order to investigate performance changes in the cognitive task, an overall
mixed-design ANOVA with Cognitive Load Condition (4: N-back 1 to 4),
Task Condition (2: single-task vs. dual-task), and Speed (2: preferred vs.
fixed speed) as within-subjects factors, and Age Group (2) as between-
subjects factor was conducted. Preliminary analyses showed no main effect
of sex, and this factor was not included in the analyses. In addition,
analysing only the rate of false alarms per trial indicated that it was
generally low, with less than one false alarm per trial even in the most
difficult condition for the children’s group.
COGNITIVE PERFORMANCE WHILE WALKING 381
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
A significant difference between the performances in different Cognitive
Load Conditions was detected, F(1.52, 92.64) ¼392.0, p5.001, Z
2
¼.87,
and that effect interacted with Age Group, F(1.52, 92.64) ¼32.6, p5.001,
Z
2
¼.35. Cognitive performance decreased with increasing cognitive load,
and that effect was more pronounced in the children’s group.
Furthermore, the within-subjects factor Task Condition (single- vs. dual-
task) reached significance, F(1, 61) ¼9.3, p5.01, Z
2
¼.13, and that effect
did not interact with Age Group. Participants improved their cognitive
performance under dual-task conditions.
No main effect for the within-subjects factor Speed (preferred vs. fixed
speed) was detected, but the interaction of Speed and Task Condition
(single- vs. dual-task) reached significance, F(1, 61) ¼9.4, p5.01, Z
2
¼.13,
indicating that cognitive improvement was more pronounced in the
preferred-speed condition than in the fixed-speed condition.
In addition, the two age groups differed significantly in their performance
in the working memory task, F(1, 61) ¼63.7, p5.001, Z
2
¼.51, with young
adults outperforming children.
To further follow-up the significant interaction of walking speed and
task condition, separate analyses were conducted for the preferred- and
fixed-speed conditions. Figures 3 and 4 depict the performance patterns
graphically. Figure 3 presents the performance in the preferred-speed
condition, and Figure 4 in the fixed-speed condition. In the fixed-speed
condition, both 9-year-old children and young adults maintained their
cognitive performance under dual-task conditions in comparison to
single-task conditions. In the preferred-speed condition, both age groups
significantly increased their cognitive performance under dual-task
conditions, F(1, 61) ¼21.7, p5.001, Z
2
¼.26, and this effect interacted
with age group, F(1, 61) ¼6.8, p5.05, Z
2
¼.10, indicating that children
benefited to a greater extent from the dual-task situation than young
adults. Moreover, the interaction of Task Condition (single- vs. dual-task)
and Cognitive Load reached significance, F(1.962, 119.664) ¼3.4, p5.05,
Z
2
¼.05. With increasing cognitive load, a more pronounced increase in N-
back working memory performance was found under dual-task conditions,
probably because participants were performing at ceiling in the N-back 1
condition and therefore could not further improve their performance in a
dual-task situation.
1
However, as can be seen in Figures 3 and 4, single-task
performances tended to be slightly higher in children for the preferred-speed
condition as compared to the fixed-speed condition. We therefore calculated
1
If the data of the N-back 1 condition were excluded from all the analyses of cognitive
performances, the same pattern of findings emerged, with just one exception: In the analysis of
the preferred-speed condition, the interaction of Task Condition (single vs. dual task) and
Cognitive Load did not reach significance (p¼.334).
382 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Figure 3. Cognitive performance when sitting and when walking with preferred speed. Single-
task performance is presented by the solid bars, and the corresponding dual-task performance is
presented by the striped bars. Error bars depict standard errors of the mean.
Figure 4. Cognitive performance when sitting and when walking with fixed speed (2.5 km/h).
Single-task performance is presented by the solid bars, and the corresponding dual-task
performance is presented by the striped bars. Error bars depict standard errors of the mean.
COGNITIVE PERFORMANCE WHILE WALKING 383
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
an average single-task score for both sessions for each individual, and re-ran
the ANOVAs reported above. There were no differences in the results for
fixed speed, but the interaction effect of single- vs. dual-task and age group did
not reach significance any more for the ANOVA with preferred speed. In
other words, when averaging the single-task scores for both sessions, children
and young adults benefited to a comparable extent from walking on the
treadmill.
DISCUSSION
This study explored differences between children’s and young adults’
ability to perform more or less challenging cognitive tasks while walking.
Concerning gait variability, the study revealed that variability of stride
time and stride length was higher in 9-year-olds than in young adults,
and it was higher under fixed-speed conditions as compared to preferred-
speed conditions. Furthermore, with increases in cognitive load, young
adults tended to decrease their gait variability, whereas children showed a
tendency to decrease their gait variability when working on an easy
cognitive task (N-back 1 and 2), but to increase the variability again
when cognitive load was very high (N-back 3 and 4). However, the
pattern was not consistent across all walking velocities and the two gait
parameters. Nevertheless, there are similarities to the findings of Huxhold
and colleagues (2006), who reported a decrease in postural sway under
conditions in which participants were focusing their attention on an easy
cognitive task, as compared to focusing their attention exclusively on the
execution of the automatized sensorimotor task of balancing. When the
difficulty of the cognitive task was further increased, older, but not
younger, adults showed an increase in postural sway, indicating that there
is a U-shaped relation between the efficacy of postural control and
concurrent cognitive demands. Children in the current study might be
comparable to older adults in the study by Huxhold and colleagues
(2006), since they had fewer resources available than young adults, and
therefore might have begun to increase their gait variability with a
nominally smaller cognitive load than young adults (see also Lo
¨vde
´n
et al., 2008). An alternative explanation for the reduction in walking
variability from no load to the dual-task conditions with low cognitive
load might be that participants are always engaged in some kind of
cognitive activity (e.g., mind wandering) under single-task walking
conditions, which might be cognitively more demanding than performing
the easy cognitive task while walking (Fraizer & Mitra, 2008). From this
perspective, any ‘‘single-task’’ performance of a rather automatized
sensorimotor task includes an unspecified cognitive task of unknown
load.
384 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Concerning performance in the working-memory task, young adults
outperformed children. In addition, performance decreased in all age groups
when cognitive load was increased, and this effect was more pronounced in
9-year-old children than in young adults. The manipulation of task difficulty
in the cognitive task guaranteed that performance was well below ceiling in
the most challenging task conditions. Both age groups significantly
improved their cognitive performance when walking on the treadmill at
their preferred speed as opposed to sitting on a chair. This effect was more
pronounced in the higher cognitive load conditions. It is unlikely that the
improvement in cognitive performance from single- to dual-task conditions
was due to a practice effect, since (a) single-task trials were assessed at the
beginning and end of each session (ABBA-scheme) and (b) there were no
performance differences between single- and dual-task conditions in the
fixed-speed condition.
The data pattern of the current study therefore lends empirical support to
theoretical assumptions that predict an increase in arousal or activation
associated with physical activity (Adam et al., 1997; Collardeau et al., 2001;
Deakin et al., 1988), which then can be invested into the cognitive task.
2
However, performance improvements only occurred in the preferred-speed
condition, in which participants were able to choose a comfortable walking
speed, and they did not occur when participants were asked to walk at a
fixed, non-preferred speed of 2.5 km/h.
There are two possible explanations for why cognitive performance was
not improved in the fixed-speed condition: Walking at the fixed speed, which
was considerably slower than the preferred speed in both age groups, might
simply not have been fast enough to increase arousal sufficiently to achieve
an effect. Given that most studies that investigated the effect of exercise-
induced arousal increases had participants perform at rather fast speeds (see
McMorris & Graydon, 2000, for a review), usually while exercising on a
bicycle ergometer or while running on a treadmill, ensuring that they were
performing at a certain percentage of their maximum exercise capacity
(VO
2max
), the movement required in the current study (walking on a
treadmill) was definitely physically less demanding. Since we did not
measure exercise-induced physical arousal directly (e.g., increases in heart
rate or oxygen consumption during walking as opposed to sitting), we are
not able to report at which stage participants were performing relative to
their maximum capacity.
2
In comparison to other studies that report no changes in accuracy of task performance, but
a decrease in reaction times to cognitive tasks with increasing physical activity (Collardeau,
Brisswalter, & Audiffren, 2001; Deakin, Starkes, & Elliott, 1988), the current study only focused
on increases in accuracy. Potential changes in reaction times from single- to dual-task
conditions have not been measured, since task instructions focused exclusively on accuracy, and
not on reaction times.
COGNITIVE PERFORMANCE WHILE WALKING 385
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Alternatively, and following a dual-task resource competition approach
(Kahneman, 1973; Wickens, 1991), which assumes that resources have to be
shared between the two tasks, it is conceivable that walking on a treadmill at
a fixed, non-preferred speed is more resource demanding than walking at
one’s preferred speed, since one has to pay some attention to adjusting one’s
walking speed to the speed of the treadmill. Studies on the automatization of
sensorimotor skills (Beilock, Carr, MacMahon, & Starkes, 2002a; Beilock,
Wierenga, & Carr, 2002b) indicate that automatized sensorimotor skills
require very little or no attention for their execution, and that paying
attention to the execution of an automatized skill leads to performance
decrements. If adjusting one’s walking speed to the fixed speed of the
treadmill, one might have to refrain from using the speed that is most
comfortable and therefore also most automatized, and if the constant
adaptation of one’s speed to the fixed speed of the treadmill is resource
demanding, no resources might have been left for investing in the
improvement of cognitive performance.
In order to distinguish between the two accounts, future studies should
include a third walking velocity in the study design, in which people are
instructed to walk faster, rather than slower, than their preferred speed. The
arousal-induced resource activation account would predict that this should
lead to further improvements in cognitive performance (at least as long
as the exercise intensity is not extremely high), while the resource
competition account would predict that any walking velocity that differs
from one’s preferred walking speed should require attentional resources and
therefore lead to a reduction in cognitive performance. In this respect, it
would also be interesting to use some other motor task, for example cycling
on a stationary bicycle, which leads to an increase in arousal while
presumably requiring less cognitive resources than walking on a treadmill.
In order to distinguish between the two accounts, some objective measure of
arousal (e.g., galvanic skin response or heart rate) should also be included in
future studies.
A design limitation of the current study is that the time period in which
the motion capture took place (and in which no target stimuli were
presented for the N-back task) was implemented at the same point in time
across trials, and it might therefore have become predictable for subjects.
However, we consider it rather unlikely that this might have led to
participants temporarily ‘‘giving up’’ working on the N-back task. In order
to exclude that possibility in future studies, the timing of the silent motion
capture period should be varied randomly across trials.
Another important extension of the reported results would be to include
special populations (e.g., patients with cerebral palsy, orthopaedic problems
influencing gait patterns, or dyslexia) in cognitive–sensorimotor dual-task
studies. It is conceivable that those groups show a more pronounced pattern
386 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
of dual-task-related performance decrements than healthy individuals,
even when taking each individual’s single-task performance into account.
An interesting population in this respect are individuals suffering from
attention-deficit hyperactivity disorder (ADHD). Motor restlessness is one
symptom of this disorder. Since the current study demonstrated that healthy
children were able to improve their cognitive performance when walking on
a treadmill at comfortable speed, hyperactive children might also be able to
profit from some type of consistent movement that does not require much
attention, even though it is often argued that those children have more
problems than healthy controls when they have to divide their attention
between two concurrent tasks (Fuggetta, 2006; Wimmer, Mayringer, &
Raberger, 1999). On the other hand, exercise is reported to have positive
effects on cognitive performance and overall behaviour in ADHD children
(Azrin, Ehle, & Beaumont, 2006; Baker, 2005; Tantillo, Kesick, Hynd, &
Dishman, 2002; Wendt, 2000), but no study has, to our knowledge,
investigated the influence of concurrent motor activity on cognitive
performance in that group.
To conclude, stride-to-stride variability decreased with cognitive load in
young adults, whereas children showed an increase in variability when
cognitive load was very high. Both children and adults improved their
cognitive performance when walking at their preferred speed as opposed to
sitting or walking at a fixed, non-preferred speed. This improvement was
more pronounced in children than in young adults. Thus, the interaction of
walking and cognitive performance is influenced by the necessity to share
resources between the tasks, and improvements in cognitive performance
can be induced by exercise through the activation of resources. Future
studies need to identify the brain mechanisms underlying arousal-induced
resource activation.
Manuscript received 4 September 2007
Revised manuscript accepted 24 September 2008
First published online 16 June 2009
REFERENCES
Adam, J. J., Teeken, J. C., Ypelaar, P. J. C., Verstappen, F. T. J., & Paas, F. G. W. (1997).
Exercise-induced arousal and information processing. International Journal of Sport
Psychology,28, 217–226.
Adolph, K. A., Vereijken, B., & Shrout, P. E. (2003). What changes in infant walking and why.
Child Development,74, 475–497.
Azrin, N. H., Ehle, C. T., & Beaumont, A. L. (2006). Physical exercise as a reinforcer to
promote calmness of an ADHD child. Behavior Modification,30, 564–570.
Baker, T. C. (2005). The use of mini-exercise breaks in the classroom management of ADHD-
type behaviors. Dissertation Abstracts International Section A: Humanities and Social
Sciences,66(6-A), 2098.
COGNITIVE PERFORMANCE WHILE WALKING 387
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Beauchet, O., Dubost, V., Herrmann, F. R., & Kressig, R. W. (2005). Stride-to-stride variability
while backward counting among healthy young adults. Journal of NeuroEngineering and
Rehabilitation,2, 1–8.
Beilock, S. L., Carr, T. H., MacMahon, C., & Starkes, J. L. (2002a). When paying attention
becomes counterproductive: Impact of divided versus skill-focused attention on novice and
experienced performance of sensorimotor skills. Journal of Experimental Psychology:
Applied,8, 6–16.
Beilock, S. L., Wierenga, S. A., & Carr, T. H. (2002b). Expertise, attention, and memory
in sensorimotor skill execution: Impact of novel task constraints on dual-task performance
and episodic memory. The Quarterly Journal of Experimental Psychology,55(A), 1211–1240.
Case, R., Kurland, D. M., & Goldberg, J. (1982). Operational efficiency and the growth of
short-term memory span. Journal of Experimental Child Psychology,33, 386–404.
Collardeau, M., Brisswalter, J., & Audiffren, M. (2001). Effects of a prolonged run on simple
reaction time of well trained runners. Perceptual and Motor Skills,93, 679–689.
Deakin, J. M., Starkes, J. L., & Elliott, D. (1988). Feature integration of children during
exercise. Journal of Sport and Exercise Psychology,10, 248–261.
Dobbs, A. R., & Rule, B. G. (1989). Adult age differences in working memory. Psychology and
Aging,4, 500–503.
Ebersbach, G., Dimitrijevic, M. R., & Poewe, W. (1995). Influence of concurrent tasks on gait.
A dual-task approach. Perceptual and Motor Skills,81, 107–113.
Fraizer, E. V., & Mitra, S. (2008). Methodological and interpretive issues in posture-cognition
dual-tasking in upright stance. Gait and Posture,27, 271–279.
Fry, A. F., & Hale, S. (1996). Processing speed, working memory, and fluid intelligence:
Evidence for a developmental cascade. Psychological Science,7, 237–241.
Fuggetta, G. P. (2006). Impairment of executive functions in boys with attention deficit/
hyperactivity disorder. Child Neuropsychology,12, 1–21.
Guttentag, R. E. (1989). Age differences in dual-task performance: Procedures, assumptions,
and results. Developmental Review,9, 146–170.
Hausdorff, J. M. (2005). Gait variability: Methods, modeling and meaning. Journal of
NeuroEngineering and Rehabilitation,2, 1–9.
Hausdorff, J. M., Zemany, L., Peng, C.-K., & Goldberger, A. L. (1999). Maturation of gait
dynamics: Stride-to-stride variability and its temporal organization in children. Journal of
Applied Physiology,86, 1040–1047.
Huang, H.-J., & Mercer, V. S. (2001). Dual-task methodology: Applications in studies of
cognitive and motor performance in adults and children. Pediatric Physical Therapy,13,
133–140.
Huxhold, O., Li, S.-C., Schmiedek, F., & Lindenberger, U. (2006). Dual-tasking postural
control: Aging and the effects of cognitive demand in conjunction with focus of attention.
Brain Research Bulletin,69, 294–305.
Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-Hall.
Krampe, R. T., Schaefer, S., Lindenberger, U., & Baltes, P. B. (2008). Lifespan dynamics of
resource-allocation in multi-task contexts: Concurrent walking and memory retrieval in
children, young, and older adults. Ms submitted for publication.
Lehrl, S., Merz, J., Burkhard, G., & Fischer, S. (1991). Manual zum MWT-A. Erlangen,
Germany: Perimed Fachbuch Verlag.
Li, K. Z. H., Krampe, R. T., & Bondar, A. (2005). An ecological approach to studying aging
and dual-task performance. In R. W. Engle, G. Sedek, U. V. Hecker, & D. N. McIntosh
(Eds.), Cognitive limitations in aging and psychopathology (pp. 190–218). Cambridge, UK:
Cambridge University Press.
Lindenberger, U., Marsiske, M., & Baltes, P. B. (2000). Memorizing while walking: Increase in
dual-task costs from young adulthood to old age. Psychology and Aging,15, 417–436.
388 SCHAEFER ET AL.
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010
Lo
¨vde
´n, M., Schaefer, S., Pohlmeyer, A. E., & Lindenberger, U. (2008). Walking variability and
working-memory load in aging: A dual-process account relating cognitive control to motor
control performance. Journal of Gerontology: Psychological Science,63B, 121–128.
Lo
¨vde
´n, M., Schellenbach, M., Grossman-Hutter, B., Kru
¨ger, A., & Lindenberger, U. (2005).
Environmental topography and postural control demands shape aging-associated decre-
ments in spatial navigation performance. Psychology and Aging,20, 683–694.
McMorris, T., & Graydon, J. (2000). The effect of incremental exercise on cognitive per-
formance. International Journal of Sport Psychology,31, 66–81.
Navon, D., & Gopher, D. (1979). On the economy of the human-processing system.
Psychological Review,86, 214–255.
Schaefer, S., Huxhold, O., & Lindenberger, U. (2006). Healthy mind in healthy body? A review
of sensorimotor-cognitive interdependencies in old age. European Review of Aging and
Physical Activity,3, 45–54.
Schaefer, S., Krampe, R. T., Lindenberger, U., & Baltes, P. B. (2008). Age differences between
children and young adults in the dynamics of dual-task prioritization: Body (balance) vs.
mind (memory). Developmental Psychology,44, 747–757.
Shumway-Cook, A., & Woollacott, M. H. (2001). Motor control: Theory and practical
applications (2nd ed.). Baltimore, Lippincott Williams & Wilkins.
Stolze, H., Kuhtz-Buschbeck, J. P., Mondwurf, C., Boczek-Funcke, A., Jo
¨hnk, K., Deuschl, G.,
et al. (1997). Gait analysis during treadmill and overground locomotion in children and
adults. Electroencephalography and Clinical Neurophysiology,105, 490–497.
Sutherland, D. (1997). The development of mature gait. Gait and Posture,6, 163–170.
Tantillo, M., Kesick, C. M., Hynd, G. W., & Dishman, R. K. (2002). The effects of exercise on
children with attention-deficit hyperactivity disorder. Medicine and Science in Sports and
Exercise,34, 203–212.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale – Revised (WAIS-R). New York:
Psychological Corporation.
Wendt, M. S. (2000). The effect of an activity program designed with intense physical exercise
on the behavior of attention deficit hyperactivity disorder (ADHD) children. Dissertation
Abstracts International Section A: Humanities and Social Sciences,61(2-A), 500.
Wickens, C. D. (1991). Processing resources and attention. In D. Damos (Ed.), Multiple-task
performance (pp. 3–34). London: Taylor & Francis.
Wimmer, H., Mayringer, H., & Raberger, T. (1999). Reading and dual-task balancing: Evidence
against the automatisation deficit explanation of developmental dyslexia. Journal of
Learning Disabilities,32, 473–478.
COGNITIVE PERFORMANCE WHILE WALKING 389
Downloaded By: [MPI Max-Planck-Institute fur Bildungsforschung] At: 06:44 28 April 2010