Psychology of Sport and Exercise 14 (2013) 323e328

Contents lists available at SciVerse ScienceDirect

Psychology of Sport and Exercise

journal homepage: www.elsevier.com/locate/psychsport

Moderate intensity exercise facilitates working memory

Amadeu Quelhas Martins, Maria Kavussanu, Adrian Willoughby, Christopher Ring*
School of Sport & Exercise Sciences, University of Birmingham, UK

a r t i c l e i n f o a b s t r a c t

Article history: Objectives: Although the effect of exercise on cognitive functioning has received considerable empirical
Received 28 May 2012 and theoretical attention, the influence of concurrent exercise on complex cognitive function remains
Received in revised form poorly understood. Our research was designed to investigate working memory during a bout of dynamic
3 November 2012
exercise.
Accepted 19 November 2012
Design: An experimental design was used.
Available online 16 December 2012
Methods: In two experiments, we examined the impact of moderate intensity exercise on performance of
a paced auditory serial addition task (Experiment 1, N ¼ 24) and a Sternberg task (Experiment 2,
Keywords:
Exercise
N ¼ 120). The tasks were performed at rest and while cycling at different power outputs.
Memory Results: We found that moderate intensity exercise increased the number of correct responses at
medium-to-fast stimulus presentation rates during the paced auditory serial addition task and lowered
the response latency slopes during the Sternberg task.
Conclusions: Our findings show that working memory is improved by dynamic exercise at moderate
intensities and short duration.
Ó 2012 Elsevier Ltd. All rights reserved.

Introduction (d ¼ 0.14) albeit heterogeneous negative effect on cognitive

functioning assessed during exercise (Lambourne & Tomporowski,
The influence of exercise on cognitive functioning has attracted 2010). Specifically, this negative effect was moderated by exercise
considerable empirical and theoretical attention. The benefits of mode (negative for running but positive for cycling), exercise
habitual physical activity are now well established (for a recent duration (negative for less than and positive for greater than
review see, Smith et al., 2010). For instance, aerobic fitness has been 20 min) and task complexity, with the latter two moderators
linked to improved executive function (e.g., Themanson & Hillman, interacting. This unreliable effect for cognitive functioning during
2006; Themanson, Pontifex, & Hillman, 2008). In addition, a large exercise can be attributed, at least in part, to methodological
number of studies have reported that cognitive performance is weaknesses of the 21 studies reviewed (e.g., lack of a control group,
improved after performing exercise (e.g., Kimura & Hozumi, 2012). low reliability of the cognitive function test), with the low statistical
The issue concerning the effects of acute exercise on concurrent power associated with small sample sizes (median ¼ 14,
cognitive performance has attracted considerable research interest. range ¼ 8e41 participants) a notable limitation.
Two quantitative reviews of the literature have systematically Overall, the findings were interpreted as offering partial support
interrogated the effects of a bout of dynamic exercise on the for a number of theoretical accounts, including both attention-
performance of a variety of cognitive tasks. An early meta-analysis based and arousal-based models. Finally, it was concluded that
concluded that exercise had a small (d ¼ 0.16) positive effect on exercise-induced arousal improves performance on rapid decision-
cognitive functioning; task difficulty emerged as a moderator, with making tasks by impacting basic sensorimotor processing during
larger effects for simple than choice reaction times (Etnier et al., and soon after exercise, and, moreover, that the residual arousal
1997). A more recent meta-analysis concluded that acute during the post-exercise period also facilitates memory processes
dynamic exercise has a small (d ¼ 0.20) homogeneous positive (Lambourne & Tomporowski, 2010). It should be noted, however,
effect on cognitive function assessed following exercise and a small that a rider tempered this conclusion: relatively few studies have
assessed memory in this context. Our research was designed to
address this limitation.
Dietrich’s (2003, 2006) transient hypofrontality model attempts
* Corresponding author. College of Life and Environmental Sciences, University of
to explain why complex cognitive processes are impaired during
Birmingham, Birmingham B15 2TT, UK. Tel.: þ44 121 414 4115; fax: þ44 121 414
4121. exercise compared to rest and recovery. Assuming that the brain’s
E-mail address: [email protected] (C. Ring). metabolic resources are limited, the model proposes that exercise

1469-0292/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.psychsport.2012.11.010
324 A. Quelhas Martins et al. / Psychology of Sport and Exercise 14 (2013) 323e328

shifts resources to areas (i.e., sensorimotor cortex) required to approved by the local research ethics committee and all volunteers
monitor and control movements at the expense of non-essential gave informed consent to participate.
areas (e.g., prefrontal cortex), and, thereby predicts decrements in
complex cognitive functioning during but not after exercise. In Apparatus
support of this model, Dietrich and Sparling (2004) found that Participants sat on a cycle ergometer (814, Monark). An audio-
performance on working memory tasks (paced auditory serial tape player and headphones (Sony) were used to present the
addition task, PASAT and Wisconsin card sorting task) that instructions and auditory stimuli. Heart rate (bpm) was recorded
depended on prefrontal cortex was impaired during prolonged (c. using a heart rate monitor (Vantage NV, Polar). A coded transmitter
50e65 min) moderate intensity (c. 70e80% of maximal heart rate) was strapped to the participant’s chest just below the xiphoid
exercise. They also found that concurrent exercise did not influence process while a coded receiver was held by the experimenter. The
performance on tasks (intelligence and vocabulary tests) that did experiment took place in a temperature controlled room.
not depend on the prefrontal cortex. Lo Bue-Estes et al. (2008) also
found that working memory performance was impaired during Paced auditory serial addition task (PASAT)
20 min of incremental treadmill exercise. Specifically, the ability to A version of the PASAT was used to assess working memory. The
add and/or subtract a series of three digits was worse while running task consisted of four 2-min blocks of trials. Participants were
at 50% than 25%, 75% and 100% of maximum aerobic capacity which instructed to add two sequentially presented single digit numbers,
were worse than before and after exercise. However, Lambourne, while retaining the latter of the two numbers in memory for
Audiffren, and Tomporowski (2010) observed no change in subsequent addition to the next number presented (Gronwall, 1977;
performance on the PASAT measured at rest and repeatedly during Tombaugh, 2006). Numbers, which ranged from 1 to 9, were pre-
and after a 40 min bout of moderate intensity cycling. Overall, the sented via an audiotape player and headphones. Participants were
results of these studies paint a complex picture of working memory instructed to add each number they heard to the previous number
before, during and after exercise. and to state the answer out loud. If performance broke down,
In the present research we conducted two experiments participants were told to continue with the next number presented.
designed to examine the effects of acute bouts of exercise on For the control group, the task consisted of four 2-min blocks of 30,
working memory performance. In Experiment 1, we used 34, 40, and 48 numbers at inter-stimulus intervals of 4.0, 3.5, 3.0, and
a between-subjects design to compare the effects of moderate 2.5 s respectively. These inter-stimulus intervals included the
intensity exercise on a cycle ergometer and rest on performance of duration (c. 500 ms) of each number. For the exercise group, the task
the PASAT (Gronwall, 1977; Tombaugh, 2006). In Experiment 2, we consisted of four 2-min blocks of 28, 33, 37 and 47 numbers; the
used a mixed design to examine the effects of different levels of slight reduction in trials was due to periodic announcements of
exercise intensity on performance of the Sternberg task (Sternberg, required changes in pedalling cadence.
1966). Neuroimaging confirms that prefrontal cortex is activated by
the PASAT (Audoin et al., 2005; Lockwood, Linn, Szymanski, Coad, & Procedure
Wack, 2004) and Sternberg task (Schon, Quiroz, Hasselmo, & Stern, Participants completed a single testing session while sitting on
2009; Wager & Smith, 2003; Wolf, Vasic, & Walter, 2006). The the cycle ergometer. They relaxed during a 5-min formal rest period
transient hypofrontality model might argue that moderate inten- while heart rate was measured and then completed 10 practice
sity exercise would impair working memory performance, and trials. Participants were tested in a mixed multi-factorial experi-
therefore we predicted that concurrent exercise on a cycle mental design, with group (experimental, control) as a between-
ergometer would be associated with poorer working memory subjects factor and trial block (1, 2, 3, 4) as a within-subjects
performance on the PASAT (Experiment 1) and the Sternberg task factor. Participants were randomly assigned to one of two groups.
(Experiment 2). The exercise group (N ¼ 12) completed the PASAT while cycling at
moderate intensity whereas the control group (N ¼ 12) completed
the PASAT while sitting at rest on the cycle ergometer. The exercise
Experiment 1
group was periodically instructed to pedal at a specific number of
revolutions per minute, ranging from 60 to 90 (M ¼ 77) rpm,
With the aim of improving our understanding of the exercise
intended to generate a power output ranging from 60 to 180
and working memory relationship we investigated PASAT
(M ¼ 146) Watts. Specifically, the target power outputs (and
(Gronwall, 1977; Tombaugh, 2006) performance at rest and while
revolutions per minute) for blocks 1, 2, 3 and 4 of the memory task
cycling. Neuroimaging confirms that prefrontal cortex is activated
averaged 95 (63 rpm), 165 (83 rpm), 155 (78 rpm) and 170
by the PASAT (Audoin et al., 2005; Lockwood et al., 2004). Based on
(85 rpm) Watts, respectively. This exercise protocol was designed to
the transient hypofrontality model, we hypothesised that moderate
ecologically simulate the changing demands associated with
exercise would impair working memory performance, reflected in
bicycle races. Heart rate was recorded every minute of the task.
decreased accuracy on the PASAT.
Data reduction
Method The number of errors (omissions, incorrect responses, late
responses) in each block were recorded and used to calculate the
Participants percentage of correct responses per block (Tombaugh, 2006). The
Participants were 24 healthy male students enrolled on a degree heart rates measurements were averaged to yield mean heart rate
course in Sport and Exercise Sciences at a British University with during rest and each block of the task.
a mean age of 20.50 (SD ¼ 0.89) years and body mass index of 23.75
(2.84) kg/m2. On average, they played sport or exercised mildly, Results and discussion
moderately and vigorously for 6.06 (SD ¼ 3.03), 4.27 (SD ¼ 2.14),
and 3.35 (SD ¼ 2.43) hours per week. At rest, their mean pulse rate PASAT
was 76.10 (SD ¼ 9.10) beats per minute, mean systolic blood pres- A 2 Group (exercise, control) by 4 Block (1, 2, 3, 4) analysis of
sure was 128.06 (SD ¼ 12.50) mmHg, and mean diastolic blood variance (ANOVA) was performed on the percentage of correct
pressure was 75.40 (SD ¼ 10.08) mmHg. The study protocol was responses. Overall, the exercise group (M ¼ 89, SD ¼ 9) only tended to
 A. Quelhas Martins et al. / Psychology of Sport and Exercise 14 (2013) 323e328 325

outperform the control group (M ¼ 82, SD ¼ 9), F(1, 22) ¼ 2.83,

p ¼ 0.11, h2 ¼ 0.11. However, the analysis yielded effects for block, F(3,
20) ¼ 8.06, p < 0.001, h2 ¼ 0.55, and group by block, F(3, 20) ¼ 5.06,
p < 0.01, h2 ¼ 0.43. The scores in each task block for each group are
shown in Fig. 1. Polynomial trend analyses confirmed a significant
linear trend for block, F(1, 22) ¼ 20.74, p < 0.001, h2 ¼ 0.49, indicating
that memory function deteriorated with increasing digit presentation
rates. Analyses also revealed a significant group by block quadratic
trend, F(1, 22) ¼ 7.58, p < 0.01, h2 ¼ 0.26, with group differences in
memory function evident only after the first block of the task.

Heart rate
A 2 Group by 5 Period (rest, block 1, block 2, block 3, block 5)
ANOVA yielded effects for group, F(1, 22) ¼ 49.00, p < 0.001,
h2 ¼ 0.69, period, F(4, 19) ¼ 57.49, p < 0.001, h2 ¼ 0.92, and group by
period, F(4, 19) ¼ 27.11, p < 0.001, h2 ¼ 0.85. The heart rates during
rest and while completing each task block for the control and
exercise groups are shown in Fig. 2. Significant group by period
linear, F(1, 22) ¼ 112.86, p < 0.001, h2 ¼ 0.84, and quadratic, F(1, Fig. 2. Mean (SE) heart rates at rest and during each two minute block of the PASAT for
22) ¼ 28.42, p < 0.001, h2 ¼ 0.56, trends indicated that that the the non-exercising control group and the exercise group.
exercise group’s heart rates increased progressively from rest to the
last block whereas the control group’s heart rates increased from
rest to the first block and then remained similarly elevated factors (i.e., time, exertion) may partly explain our findings. Finally,
throughout the remainder of the task. It is worth noting that the the sample size was relatively small and therefore the lack of
exercise group’s heart rates during the four blocks of the task were differences in memory task performance at the slowest stimulus
60 (SD ¼ 7), 69 (SD ¼ 9), 74 (SD ¼ 9), and 77 (SD ¼ 10) percent of presentation rate may be attributable to low statistical power. The
maximum predicted heart rates, calculated as 220 minus age. In next experiment was designed with these issues in mind.
contrast, the control group’s corresponding heart rates were only
47 (SD ¼ 6), 46 (SD ¼ 6), 47 (SD ¼ 6) and 48 (SD ¼ 6) percent of Experiment 2
predicted maximum.
In sum, the results of Experiment 1 indicate that exercise With the aim of further improving our understanding of the
selectively influenced PASAT performance. Specifically, cycling was exercise and working memory relationship we investigated Stern-
associated with improved performance compared to rest when the berg task (Sternberg, 1966) performance at rest and while cycling.
stimuli were presented every two to three seconds whereas this Neuroimaging confirms that prefrontal cortex is activated by the
effect was absent when stimuli were presented at a slower rate. It Sternberg task (Schon et al., 2009; Wager & Smith, 2003; Wolf et al.,
should be noted here that exercise intensity was also greater at the 2006). Based on the transient hypofrontality model, we hypoth-
faster presentation rates when performance was improved. esised that moderate exercise would impair working memory
Potential limitations need to be considered when interpreting the performance, reflected in decreased accuracy on the PASAT and
results of the current experiment. First, the exercise intensity was slower processing of information in memory (i.e., a steeper slope)
varied while participants were cycling and therefore this feature of for the Sternberg task.
the design may have distracted participants from the memory task
performance. Second, the current design confounded presentation Method
rate with elapsed time and level of exertion, and, therefore, these
Participants
Participants were 120 (55 males, 65 females) healthy right-
handed students enrolled on a degree course in Sport and Exer-
cise Sciences at a British University with a mean age of 19.57
(SD ¼ 0.83) years and body mass index of 22.99 (2.57) kg/m2. On
average, they exercised for 7.07 (SD ¼ 3.69) hours per week. At rest,
their mean heart rate was 74.94 (SD ¼ 12.07) beats per minute,
mean systolic blood pressure was 121.68 (SD ¼ 9.90) mmHg, and
mean diastolic blood pressure was 76.23 (SD ¼ 8.87) mmHg. The
study protocol was approved by the local research ethics
committee and all volunteers gave informed consent to participate.

Apparatus
Participants sat on a cycle ergometer (824E, Monark) with
a stimulus box mounted on the front of the ergometer and
a response box under their dominant hand. The stimulus box
contained a single 40 mm wide by 55 mm high dual-colour (green,
red) 7-segment light emitting diode panel that was used for pre-
senting warning, experimental, probe and feedback stimuli. The
Fig. 1. Mean (SE) performance accuracy scores, indexed by the percentage of correct
response box contained two low force microswitch levers (D459-
responses, during each two minute block of the PASAT for the non-exercising control V3LD, Cherry). Heart rate was measured as described in Experiment
group and the exercise group. 1. The experiment took place in a temperature controlled room.
326 A. Quelhas Martins et al. / Psychology of Sport and Exercise 14 (2013) 323e328

Sternberg task condition, F(4, 232) ¼ 3.55, p < 0.01, h2 ¼ 0.06. The performance
A version of the Sternberg task (Sternberg, 1966) was used to scores under control and exercise conditions for each intensity
assess working memory. A computer was programmed in Spike2 to group are shown in Fig. 3. To interrogate these effects, a series of 2
present stimuli and collect responses via a Power1401 (Cambridge Condition (control, exercise) ANOVAs were performed on each
Electronic Design). At the start of each of 96 trials, participants were group. The slopes were shallower during exercise than control in
required to depress the two response levers with the index and the low, F(1, 41) ¼ 11.56, p < 0.002, h2 ¼ 0.22, and medium, F(1,
middle fingers of their dominant hand. The task waited (i.e., no 37) ¼ 9.05, p < 0.005, h2 ¼ 0.20, intensity groups but did not differ
stimuli were presented) until both response levers were depressed. between conditions in the very low intensity group, F(1, 39) ¼ 0.67,
Following a 250 ms delay, the program serially presented a set of p ¼ 0.42, h2 ¼ 0.02. To further explore this effect, we computed the
either two or six green single-digit numbers ranging from 1 to 9, change in slope (exercise minus control) and compared the change
without any repeats. Each number was presented for 750 ms with scores of the low (M ¼ 12.23 ms/digit) and medium
a 250 ms interval between numbers. After a 3000 ms delay, a red (M ¼ 11.13 ms/digit) groups using a 2 Group (low, medium)
probe number was presented for 750 ms. Participants were required ANOVA. This analysis indicated that the effect of exercise on
to decide whether this red number was presented in the previous memory performance was comparable for these two groups, F(1,
set of green numbers. If the red number was a match, participants 78) ¼ 0.05, p ¼ 0.83, h2 ¼ 0.00. In terms of basic sensorimotor
were instructed to lift their middle finger; if it was not a match, processing speed, the zero intercepts did not differ between
participants were instructed to lift their index finger. Participants conditions for any group: very low, F(1, 39) ¼ 0.40, p ¼ 0.84,
were then given performance feedback: they were shown a green U h2 ¼ 0.00, low, F(1, 41) ¼ 0.75, p ¼ 0.39, h2 ¼ 0.02, and medium, F(1,
if the response was correct and a red U if the response was wrong. 37) ¼ 1.20, p ¼ 0.28, h2 ¼ 0.03.
Participants were instructed to respond as rapidly as possible while
keeping errors to a minimum. The task was divided into blocks of 48 Heart rate
trials, each of which lasted approximately eight minutes. Partici- A 3 Exercise Intensity Group by 2 Condition ANOVA on the heart
pants rested for three minutes after each block. rates yielded effects for group, F(2, 114) ¼ 29.21, p < 0.001, h2 ¼ 0.34,

Procedure
Participants completed a single testing session. At the start of the
session they completed 24 practice trials. Participants were tested in
(A)
a mixed multi-factorial experimental design, with condition
(control, exercise) as a within-subjects factor and exercise intensity
(very low, low, medium) as a between-subjects factor. In the control
condition, which was performed first, participants completed the
Sternberg task while sitting on the cycle ergometer. In the exercise
condition, which was performed second, they completed the
Sternberg task while exercising at one of the three randomly
assigned intensities. The very low intensity group (N ¼ 40) was
instructed to pedal at 45 revolutions per minute with no added
brake friction, which corresponded to a power output of approxi-
mately 5 Watts. The low intensity group (N ¼ 42) was instructed to
pedal at 50 (women) and 60 (men) revolutions per minute at
a power output of 50 Watts (women) and 60 Watts (men). The
medium intensity group (N ¼ 38) was instructed to pedal at 50
(women) and 60 (men) revolutions per minute at a power output of
75 Watts (women) and 90 Watts (men). In the exercise condition,
participants pedalled for two minutes to approach steady state
before starting each block of trials of the Sternberg task.
(B)
Data reduction
Response latency (ms) was calculated as the time between the
onset of the probe stimulus and the release of the switch lever.
Responses were discarded if the response latency was less than
100 ms (i.e., anticipation error) or greater than 2250 ms (i.e., inat-
tention error), or if the participant lifted both fingers concurrently
(<100 ms apart). The mean response latencies associated with the
two-number and six-number sets were used to calculate the slope
(ms/digit) and zero intercept (ms) using linear regression
(Sternberg, 1966). The heart rate measurements were averaged to
yield mean heart rate during control and exercise.

Results and discussion

Sternberg task
A 3 Exercise Intensity Group (very low, low, medium) by 2
Condition (control, exercise) multivariate ANOVA was performed Fig. 3. Mean (SE) performance scores on the Sternberg task during control and exer-
on the slopes and zero intercepts. This yielded multivariate effects cise conditions for the very low, low and medium intensity exercise groups: the slope
for condition, F(2, 116) ¼ 20.12, p < 0.001, h2 ¼ 0.26, and group by of the response latencies (A) and the zero intercept of the response latencies (B).
 A. Quelhas Martins et al. / Psychology of Sport and Exercise 14 (2013) 323e328 327

condition, F(1, 114) ¼ 439.77, p < 0.001, h2 ¼ 0.79, and group by more fully characterise the fitness levels of participants and
condition, F(1, 114) ¼ 90.40, p < 0.001, h2 ¼ 0.61. Control and exercise attempt to confirm the current observations.
cardiac activity for each group are shown in Fig. 4. To explore these
effects, a series of 2 Condition ANOVAs were performed on each General discussion
variable for each group. The heart rates of all groups increased from
control to exercise: very low, F(1, 38) ¼ 6.69, p < 0.01, h2 ¼ 0.15, low, We investigated the influence of dynamic exercise on working
F(1, 40) ¼ 263.74, p < 0.001, h2 ¼ 0.87, and medium, F(1, 35) ¼ 184.49, memory. We found that moderate intensity cycling increased the
p < 0.001, h2 ¼ 0.85. A 3 Group ANOVA revealed group differences in number of correct responses in the PASAT at moderate stimulus
the extent of the heart rate reactions to exercise (i.e., exercise presentation rates when the task was not too simple (Experiment 1)
value minus control value), F(2, 114) ¼ 90.40, p < 0.001, h2 ¼ 0.61: and reduced the response latency per additional digit in the Stern-
the cardiac change scores of the medium (M ¼ 47.7 bpm) and low berg task (Experiment 2). Our findings show that working memory
(M ¼ 42.5 bpm) intensity groups were greater than those of the very can be improved by aerobic exercise performed at moderate
low intensity group (M ¼ 3.0 bpm). It is noteworthy that the heart intensities.
rates during the exercise task corresponded to 41 (SD ¼ 7), 61 Experiment 1 found that exercise selectively influenced PASAT
(SD ¼ 10) and 64 (SD ¼ 13) percent of maximum predicted heart performance. Cycling improved performance compared to rest
rates for the very low, low and medium intensity groups, when the stimuli were presented every two to three seconds
respectively. whereas the exercise effect was absent when stimuli were pre-
In brief, the results of Experiment 2 show that exercise selec- sented at slower rates. The finding that overall task performance
tively influenced Sternberg task performance. Cycling at low to did not differ significantly is in line with a recent study by
moderate intensities (c. 120e130 bpm corresponding to 60e65% Lambourne et al. (2010) who observed no changes in overall PASAT
of maximum heart rate) reduced the response latency slopes performance before, during and after 40 min of cycling at a similar
indicative of improve memory retrieval and scanning processes intensity to our experiment (average heart rates were 143 and
(Sternberg, 1966). However, exercise did not influence the zero 140 bpm, respectively). Their null finding may be attributed to
intercept, an index of basic sensorimotor processes. A number of a combination of low power and a ceiling effect due to consistently
potential limitations need to be considered when interpreting the high performance (with correct responses ranging from 92 to
results of the current experiment. First, the control condition 96%) that was associated with extensive practice and despite
always preceded the exercise condition and therefore the perfor- individually-tailored stimulus presentation rates of 1.2, 1.6 or 2.0 s.
mance during exercise may have been facilitated in part by Importantly, both sets of findings are contrary to the impaired
a learning effect despite the participants practising on the task performance reported by Dietrich and Sparling (2004) using the
before the resting control condition performance was formally classic version of the task comprising a series of 50 digits at
assessed. Accordingly, the interpretations should be based more presentation rates of 2.4, 2.0, 1.6 and 1.2 s. Given that exercise
strongly on relative control versus exercise comparisons among the intensity (c. 150 bpm corresponding to 70e80% of maximum heart
three exercise groups. Second, we did not determine each partici- rate) was comparable, the discrepancy may be attributed to
pant’s maximum aerobic capacity and then require them to exer- differences in exercise mode and duration (cf. Lambourne &
cise at a proportion of that maximal. Instead our aim was to create Tomporowski, 2010). Specifically, Dietrich and Sparling’s (2004)
different exercise intensities and to recruit a large sample of fit protocol required participants to run on a treadmill for a pro-
young athletes so that we could first determine whether a brief tracted period (c. 40 min) before commencing the PASAT. Taken
bout of moderate intensity exercise influenced memory task together, these results argue that the transient hypofrontality
performance. Our posterior computations of workload in terms of model be revised to consider the mental and physical demands (i.e.,
relative heart rate, expressed as proportion of maximum predicted complexity, duration, intensity) of the exercise task. This conclusion
heart rate, indicates that the exercise intensities were comparable is similar to the one drawn by Rooks, Thorn, McCully, and Dishman
to previous studies involving moderate intensity exercise. None- (2010) in their review of the effects of acute exercise on cerebral
theless, it should be conceded that future research would do well to oxygenation as measured using near-infrared spectroscopy,
showing that moderate intensity exercise increased whereas very
high intensity exercise decreased the levels of oxygen levels in the
prefrontal cortex. Finally, it may be worth noting that elapsed time
and level of exertion may also have contributed to the group
differences in the memory task performance as a function of block.
Experiment 2 found that exercise selectively influenced Stern-
berg task performance. Cycling at low to moderate intensities
(c. 120e130 bpm corresponding to 60e65% of maximum heart rate)
reduced the response latency slopes indicative of improved
memory retrieval and scanning processes (Sternberg, 1966). Finally,
exercise did not influence the zero intercept, an index of basic
sensorimotor processes. Although this latter finding may appear
contrary to some previous research indicating that exercise facili-
tates reaction times (e.g., Etnier et al., 1997), it is worth noting that
the zero intercepts, which averaged around 700 ms (i.e., much
slower than simple reaction times), indicate that this measure
reflects more complex processing demands than simple reaction
time. This may help to explain the null effect for the zero intercept
observed here. In agreement with the results of experiment 1, our
findings confirm that working memory improved during low to
Fig. 4. Mean (SE) heart rates while performing the Sternberg task during control and moderate intensity exercise. They are also broadly in line with data
exercise conditions for the very low, low and medium intensity exercise groups. showing that performance on the Sternberg task is improved
328 A. Quelhas Martins et al. / Psychology of Sport and Exercise 14 (2013) 323e328

immediately and half an hour after exercise (Pontifex, Hillman, Dietrich, A. (2003). Functional neuroanatomy of altered states of consciousness: the
transient hypofrontality hypothesis. Consciousness & Cognition, 12, 231e256.
Fernhall, Thompson, & Valentini, 2009) and performance on the
http://dx.doi.org/10.1016/S1053-8100(02)00046-6.
PASAT is improved similarly after exercise (Tomporowski et al., Dietrich, A. (2006). Transient hypofrontality as a mechanism for the psychological
2005). Accordingly, this evidence suggests that working memory effects of exercise. Psychiatry Research, 145, 79e83. http://dx.doi.org/10.1016/
can be facilitated by modest increases in physiological arousal, both j.psychres.2005.07.033.
Dietrich, A., & Sparling, P. B. (2004). Endurance exercise selectively impairs
during or after exercise. This pattern of results is compatible with prefrontal-dependent cognition. Brain & Cognition, 55, 516e524. http://
some arousal-based models (and contrary to the transient hypo- dx.doi.org/10.1016/j.bandc.2004.03.002.
frontality model as originally specified) of exercise and cognitive Etnier, J. L., Salazar, W., Landers, D. M., Petruzzello, S. J., Han, M., & Nowell, P. (1997).
The influence of physical fitness and exercise upon cognitive functioning:
functioning (e.g., Kahneman, 1973; Oxendine, 1984; for review see a meta-analysis. Journal of Sport & Exercise Psychology, 19, 249e277.
Lambourne & Tomporowski, 2010). Similar improvements in Gronwall, D. M. (1977). Paced auditory serial-addition task: a measure of recovery
performance with moderate intensity exercise have been noted for from concussion. Perceptual & Motor Skills, 44, 367e373. http://dx.doi.org/
10.2466/pms.1977.44.2.367.
other aspects of performance, such as attention (Pesce, Capranica, Kahneman, D. (1973). Attention and effort. Englewood Cliffs, NJ: Prentice-Hall.
Tessitore, & Figura, 2002), decision making speed (Davranche & Kimura, K., & Hozumi, N. (2012). Investigating the acute effect of an aerobic dance
Audiffren, 2004; McMorris et al., 1999) and response preparation exercise program on neuro-cognitive function in the elderly. Psychology of Sport
and Exercise, 13, 623e629. http://dx.doi.org/10.1016/j.psychsport.2012.04.001.
(Arcelin, Delignieres, & Brisswalter, 1998). Recent research indicates Lambourne, K., Audiffren, M., & Tomporowski, P. D. (2010). Effects of acute exercise
that memory performance (percent correct) improves with on sensory and executive processing tasks. Medicine & Science in Sports &
increased arousal and then declines with further increases in Exercise, 42, 1396e1402. http://dx.doi.org/10.1249/MSS.0b013e3181cbee11.
Lambourne, K., & Tomporowski, P. D. (2010). The effect of exercise-induced arousal
arousal (Choi et al., 2012), however, the differences in arousal in our
on cognitive task performance: a meta-regression analysis. Brain Research, 1341,
low and medium intensity exercise conditions may not have been 12e24. http://dx.doi.org/10.1016/j.brainres.2010.03.091.
sufficiently distinct to test the inverted-U arousal-performance Lo Bue-Estes, C., Willer, B., Burton, H., Leddy, J. J., Wilding, G. E., & Horvath, P. J.
hypothesis (Yerkes & Dodson, 1908). Based on the YerkeseDodson (2008). Short-term exercise to exhaustion and its effects on cognitive function
in young women. Perceptual & Motor Skills, 107, 933e945. http://dx.doi.org/
Law, arousal-based explanations of the exerciseecognition rela- 10.2466/pms.107.3.933-945.
tionship assume that task performance improves initially at low Lockwood, A. H., Linn, R. T., Szymanski, H., Coad, M. L., & Wack, D. S. (2004).
intensity exercise when arousal increases above basal or resting Mapping the neural systems that mediate the paced auditory serial addition
task (PASAT). Journal of the International Neuropsychological Society, 10, 26e34.
levels, performance peaks when the intensity of exercise has http://dx.doi.org/10.1017/S1355617704101045.
increased the level of arousal to some optimal level, and then McMorris, T., Myers, S., MacGillivary, W. W., Sexsmith, J. R., Fallowfield, J.,
performance declines when exercise-induced arousal continues to Graydon, J., et al. (1999). Exercise, plasma catecholamine concentrations and
decision-making performance of soccer players on a soccer-specific test. Journal
rise. In the current research, performance on the memory task of Sports Sciences, 17, 667e676. http://dx.doi.org/10.1080/026404199365687.
improved at low to moderate levels of exercise-induced arousal, Oxendine, J. B. (1984). Psychology of motor learning. Englewood Cliffs, NJ: Prentice-
however, exercise was never sufficiently intense to cause any Hall.
Pesce, C., Capranica, L., Tessitore, A., & Figura, F. (2002). Effects of a sub-maximal
deterioration in performance. Accordingly, it would seem that physical load on the orienting and focusing of visual attention. Journal of
executive functions can be facilitated by relatively low to medium Human Movement Studies, 42, 401e420.
levels of arousal. Pontifex, M. B., Hillman, C. H., Fernhall, B., Thompson, K. M., & Valentini, T. A. (2009).
The effect of acute aerobic and resistance exercise on working memory. Medi-
In conclusion, the current findings, which were based on rela-
cine & Science in Sports & Exercise, 41, 927e934.
tively medium-to-large samples using well validated tasks and Rooks, C. R., Thom, N. J., McCully, K., & Dishman, R. K. (2010). Effects of incremental
sophisticated experimental designs and recording equipment, exercise on cerebral oxygenation measured by near-infrared spectroscopy:
revealed medium-to-large effect sizes (Cohen, 1992) for the influ- a systematic review. Progress in Neurobiology, 92, 134e150. http://dx.doi.org/
10.1016/j.pneurobio.2010.06.002.
ence of exercise on working memory. Nevertheless, they need to be Schon, K., Quiroz, Y. T., Hasselmo, M. E., & Stern, C. E. (2009). Greater working
replicated by other groups and using other tasks, and, ideally using memory load results in greater medial temporal activity at retrieval. Cerebral
a multi-measure approach in which behavioural measures of Cortex, 19, 2561e2571. http://dx.doi.org/10.1093/cercor/bhp006.
Smith, P. J., Blumenthal, J. A., Hoffman, B. M., Cooper, H., Strauman, T. A., Welsh-
performance are supplemented by physiological measures of brain Bohmer, K., et al. (2010). Aerobic exercise and neurocognitive performance:
function. a meta-analytic review of randomized controlled trials. Psychosomatic Medicine,
72, 239e252. http://dx.doi.org/10.1097/PSY.0b013e3181d14633.
Acknowledgements Sternberg, S. (1966). High-speed scanning in human memory. Science, 153, 652e
654. http://dx.doi.org/10.1126/science.153.3736.652.
Themanson, J. R., & Hillman, C. H. (2006). Cardiorespiratory fitness and acute
The first author was supported by a Portuguese Government aerobic exercise effects on neuroelectric and behavioral measures of action
“Fundação para a Ciência e a Tecnologia” Fellowship, SFRH/BD/ monitoring. Neuroscience, 141, 757e767. http://dx.doi.org/10.1016/
j.neuroscience.2006.04.004.
42348/2007.
Themanson, J. R., Pontifex, M. B., & Hillman, C. H. (2008). Fitness and action moni-
toring: evidence for improved cognitive flexibility in young adults. Neuroscience,
References 157, 319e328. http://dx.doi.org/10.1016/j.neuroscience.2008.09.014.
Tombaugh, T. N. (2006). A comprehensive review of the paced auditory serial
Arcelin, R., Delignieres, D., & Brisswalter, J. (1998). Selective effects of physical addition test (PASAT). Archives of Clinical Neuropsychology, 21, 53e76. http://
exercise on choice reaction processes. Perceptual & Motor Skills, 87, 175e185. dx.doi.org/10.1016/j.acn.2005.07.006.
http://dx.doi.org/10.2466/pms.1998.87.1.175. Tomporowski, P. D., Cureton, K., Armstrong, L. E., Kane, G. M., Sparling, P. B., &
Audoin, B., Ibarrola, D., Au Duong, M. V., Pelletier, J., Confort-Gouny, S., Malikova, I., Millard-Stafford, M. (2005). Short-term effects of aerobic exercise on executive
et al. (2005). Functional MRI study of PASAT in normal subjects. Magnetic processes and emotional reactivity. International Journal of Sport & Exercise
Resonance Materials in Physics, Biology & Medicine, 18, 96e102. http://dx.doi.org/ Psychology, 3, 131e146. http://dx.doi.org/10.1080/1612197X.2005.9671763.
10.1007/s10334-004-0098-9. Wager, T. D., & Smith, E. E. (2003). Neuroimaging studies of working memory:
Choi, M.-H., Min, Y.-K., Kim, H.-S., Kim, J.-H., Yeon, H.-W., Choi, J.-S., et al. (2012). a meta-analysis. Cognitive, Affective & Behavioral Neuroscience, 3, 255e274.
Effects of three levels of arousal on 3eback working memory task performance. http://dx.doi.org/10.3758/CABN.3.4.255.
Cognitive Neuroscience. http://dx.doi.org/10.1080/17588928.2011.634064. Wolf, R. C., Vasic, N., & Walter, H. (2006). Differential activation of ventrolateral
Cohen, J. (1992). A power primer. Psychological Bulletin, 112, 155e159. http:// prefrontal cortex during working memory retrieval. Neuropsychologia, 44,
dx.doi.org/10.1037//0033-2909.112.1.155. 2558e2563. http://dx.doi.org/10.1016/j.neuropsychologia.2006.05.015.
Davranche, K., & Audiffren, M. (2004). Facilitating effects of exercise on information Yerkes, R. M., & Dodson, J. D. (1908). The relation of strength of stimulus to
processing. Journal of Sports Sciences, 22, 419e428. http://dx.doi.org/10.1080/ rapidity of habit-formation. Journal of Comparative Neurology and Psychology,
02640410410001675289. 18, 459e482.