The neural correlates of working memory in elite athletes and controls

(1)

Title: The neural correlates of working memory in elite athletes and controls

Date: 22-08-2018

Name: Carlo Rooth Student ID: 10198350

Supervisor: Zai-Fu Yao Examiner 1: dr. Ilja Sligte Examiner 2: dr. Yaïr Pinto

MSc in Brain and Cognitive Sciences, Cognitive Science track, University of Amsterdam

Abstract

Recently, several opinion papers have proposed that working memory (WM) is important for athletic performance, and that elite athletes might outperform controls on general WM tasks. However, the field lacks both behavioural and neural data to support this hypothesis. The purpose of our study was to examine the differences in WM capacity and the neural circuitry responsible for WM between elite athletes and controls. We recruited 21 athletes and 20 controls that performed a change-detection task measuring WM capacity in a fMRI scanner. Our primary behavioural findings were that elite athletes do not have a higher WM capacity than controls. However, our functional data shows that athletes show stronger activation patterns in parietal regions, whereas controls show stronger activations in frontal regions while performing the WM task. This implicates that athletes and controls have different strategies for WM. Athletes rely on stimulus driven processes, whereas controls might rely on goal-directed processes.

(2)

Introduction

Working memory in sports

Imagine a soccer player on the field during a match. He has the ball between his feet, he needs to update where his opponents are, where his teammates are, where the ball is. Also, he must remember the rules of the game and the tactics of the team. The player needs to keep all such information in mind for a limited time, he is then able to make the best decision for the next move. In the example illustrated, a set of processes that allow us to hold and use relevant information and carry-out complex tasks refer to as working memory. Working memory (WM) is a part of executive function that maintains and manipulates information in the mind. Specifically, two types of WM are commonly distinguished: verbal WM and visual-spatial WM (Diamond, 2013). This distinction is based on the content of WM. Verbal WM consists of the maintenance and manipulation of verbal items, such as words, numbers or letters. Visual-spatial WM consists of spatial and non-verbal items, such as locations, orientations or shapes (Zimmer, 2008). In this research, we will only focus on visual-spatial WM since this is more applicable to sports than verbal WM (Furley & Memmert, 2010b). Despite the critical role of working memory to success in sport, little is known whether elite athletes have higher working memory capacity compared to controls.

Until recently, several theoretical papers emerged to speculate the potential role of working memory capacity in sports, it is proposed that individual variances in working memory capacity might be a moderating variable for performance in sports (Furley & Memmert, 2010b; Furley & Wood, 2016). Even though the field lacks experimental paradigms that are externally valid, it is likely that working memory capacity facilitates the development of expertise in certain contexts (Buszard, Masters, & Farrow, 2017; Buszard & Masters, 2017). We therefore speculated that elite athletes have higher working memory capacity than controls.

However, the behavioral evidence is still inconclusive. One study examined the visual-spatial abilities of team-ball athletes using the Corsi block-tapping test. An experimenter tapped a random sequence of blocks and participants had to repeat this sequence. The difficulty was increased by increasing the length of the sequence. Overall performance on the task did not differ between elite basketball players and non-athletes. This indicates that elite basketball players did not differ in their spatial capacity (Furley & Memmert, 2010a). Nonetheless, this might be explained by the sequential nature of the task, which is not a common requirement for basketball. It is better suited to measure working memory with a

(3)

task that simultaneously presents spatial stimuli, since that is more applicable to the nature of sport (Furley & Memmert, 2010a). Accordingly, our study will use a change-detection paradigm. This is more applicable to the nature of sport, as athletes constantly have to detect changes in the environment.

Differences between sports

Moreover, it has been hypothesized that individual differences in working memory capacity might be a moderator for expertise in open skill sports (Furley & Memmert, 2010; Wang et al., 2013). Open skill sports are defined as those in which the players act in a dynamically changing, unpredictable and externally-paced environment, such as football or volleyball. It is separated from closed skill sports; these are sports in which the players act in a highly consistent, predictable and self-paced environment, such as running or rowing (Voss, Kramer, Basak, Prakash, & Roberts, 2010; Wang et al., 2013). Therefore, we recruited both open-skill and closed-skill athletes to account for the differences in working memory between sport types.

Neural correlates

As shown, the behavioral evidences gives no straight answer whether elite athletes have a superior working memory (Furley & Memmert, 2010a). Examining the functional brain activity would provide more insight into the underlying brain structures responsible for working memory. For example, only one study found higher activations in elite archers for areas associated with visuospatial attention and working memory when they performed a line orientation task (Seo et al., 2012). Participants had to seek the number of lines that had the same angle as a test line. This task has been evident to assess visuospatial perception rather than working memory, since it does not require to hold information in mind and manipulate it (Diamond, 2013).

In our study we will use a single-probe version of the change-detection paradigm of Pessoa, Gutierrez, Bandettini, & Ungerleider (2002) to measure visual-spatial working memory (see Figure 1). Subjects need to indicate whether one of six rectangles had changed in orientation or not. This paradigm measures visual-spatial working memory, since both the location and the orientation of the test cue should be in memory to give a correct answer. To our knowledge, this is the first study attempt to study the neural correlates of working memory in elite athletes.

(4)

Figure 1. The change-detection task subjects perform in the MRI scanner. This is a modified version derived from Pessoa et

al. (2002).

Figure 2. Typical working memory network that is active during the change detection paradigm. It consists of activations

from the dorsal occipital cortex (DO) through the superior parietal lobe (SPL) to frontal regions, such as the dorsolateral prefrontal cortex (dlPFC). Picture is taken from Pessoa et al. (2002).

Prior research on working memory evident that frontal and parietal regions to be active, such as the dorsolateral prefrontal cortex, superior parietal lobe, frontal eye field and presupplementary motor area (see Figure 2; Pessoa et al., 2002) form as distributed networks. Local activity within each region depends strongly on the stimulus and task. For orientation and location stimuli, a network is active from the early visual cortex, posterior parietal cortex,

(5)

frontal eye fields to the lateral prefrontal cortex (Christophel, Klink, Spitzer, Roelfsema, & Haynes, 2017). This is consistent with the previous findings from Pessoa et al. (2002). As such, we will focus on activation in the network from the visual cortex, posterior parietal lobe and frontal eye fields into the dorsolateral prefrontal cortex. Since elite athletes have superior expertise in their sport and working memory capacity might facilitate this expertise (Buszard & Masters, 2017), it might be possible that elite athletes have different neural patterns within this network.

Research questions and hypotheses

Our research focuses on two questions: whether elite athletes and controls differ in working memory capacity and whether elite athletes and controls differ in the neural correlates of visual-spatial working memory.

Hence, we will test three hypotheses:

1) Since several theoretical papers speculate the importance of WM in sports (Buszard et al., 2017; Furley & Memmert, 2010b; Furley & Wood, 2016), we expect elite athletes to have better visual-spatial working memory capacity compared to controls.

2) Since elite athletes excel in their sport and working memory might facilitate sport expertise (Buszard et al., 2017), we expect elite athletes to have different neural patterns in regions responsible for working memory compared to controls.

3) Since athletes practicing open skill sports have to act in a dynamic, unpredictable and externally paced environment, and this requires athletes to have a strong working memory, we expect open skill athletes to have stronger activations in regions of the working memory network compared to closed skill athletes.

Methods

Subjects

Total 43 subjects were recruited which consisted of a group of 22 athletes (Mage = 25.3, SDage

= 4.76) and a group of 21 controls (Mage = 23.8, SDage = 2.60). Controls were matched to an

athlete on gender, age and education level. One control subject could not be matched with an athlete. See Table 1 for a summary of their demographic information. All subjects were compensated with €10 or 1 research credit for their participation.

(6)

Table 2. Summary of demographic information. The closed skill athletes (consisting of rowers and synchronized swimmers)

were older, played at a higher level of competition, were professional athletes for a longer time, and participated in more trainings per week compared to the open-skill athletes (consisting of volleyball, soccer, handball and karate players).

Procedure

The experiment took place in the Spinoza Center for Neuroimaging in Amsterdam. All participants performed two cognitive tasks: a change-detection task measuring visual-spatial working memory and a stop-signal task measuring motor inhibition. Results of the stop-signal task are analyzed in another paper. The cognitive tests were administered in the following order for all participants while they were in the MRI scanner:

1) Stop signal practice session (2 minutes); 2) Change detection practice session (1 minute); 3) Change detection task (6 minutes);

3) Stop signal task (12 minutes); 4) Anatomical T1 images (5 minutes); 5) Stop signal task (12 minutes);

6) Working memory task without practice session (6 minutes).

To mitigate potential fatigue/practice effects, subjects were given a 1-minute break between the consecutive tests in the MRI scanning session.

Apparatus

All experiments were programmed and generated with custom software Presentation® on the HP Z620 desktops and visual stimuli were presented on BOLD screen from Cambridge Research Systems® with 32" displays. The participants made responses on the 2x4 Button Bimanual Curved Lines Response Box (HHSC-2x4-C)® used in computer research tasks that

Closed skill athletes (N=15)

Open skill athletes (N=6)

Controls (N=20) p-value ANOVA

Age 27.2 (±4.29) 20.67 (±1.03) 23.61 (±4.85) .00197*

Gender 6 female, 9 male 6 female 12 female, 6 male N.A.

Highest level of competition

Olympic Games National Championship Local club N.A.

Start age professional 18.21 (±4.57) 15.66 (±4.38) N.A. .3

Years of training 9.14 (±3.66) 4.66 (±2.62) N.A. .025*

Training per week 8.71 (±3.73) 4.17 (±0.69) N.A. .007*

Hours of training per session

(7)

are compatible with MRI scanners. The distance between the participant’s eye to the mirror was about 10cm, and from the mirror to the screen was about 148 cm.

MRI Data acquisition

Functional and structural brain images were acquired with a Philips 3T Achieva MRI scanner located in the Spinoza center for Neuroimaging, The Netherlands. A SENSE 32-channel receiver whole-head array coil was used for data collection. T2*-weighted functional images were acquired using an echo planar imaging (EPI) sequence (TR = 2000ms, TE = 27ms, 37 oblique slices acquired in ascending interleaved order, 3.0 × 3.0 × 3.0 mm2 isotropic voxel, 80 × 80 matrices in a 220 mm2 field of view, flip angle 76°, thickness = 3 mm, slice number = 160 during the change detection task, and slice number = 363 during the stop signal task). Structural scans were acquired using a T1-weighted sequence via a MPRAGE (magnetization-prepared rapid acquisition gradient echo) sequence in the same orientation as the functional sequences to provide high-resolution anatomical images for subsequent alignment to the functional scans (TR = 8 ms, TE = 3.7 ms, FOV = 240 × 240 mm2, thickness = 1 mm, slice numbers = 220, flip angle 8°, voxel size 1 ×1× 1 mm ) and spatial normalization.

Figure 3. The four conditions used in the change-detection task that subjects perform in the MRI scanner in order from top to

down: change trial, no-change trial, null trial, blank trial. The arrow represents time. Activations in the change trial and no change trial together are compared with either the blank trial (see contrast 9 in Table 3) or null trial (see contrast 10 in Table 3), because the change condition and no change condition were assumed to have a load on working memory, whereas the null trial and blank trial did not.

(8)

Task procedure of change detection task

Before the experimental trials started, the subjects performed 10 practice trials to get familiar with the task. Subjects then performed 60 formal trials. One trial had a duration of 6000ms and had the following procedure (see Figure 2). A fixation cross was shown on the screen and turned green for 500ms to indicate that visual cues will be shown. Then 6 rectangles were appeared in a circular manner for 1000ms before they disappear. The orientation of these 6 rectangles needed to be remembered. After a delay-period of 2000ms one rectangle was shown on one of the 6 positions. In 16 trials the orientation of the rectangle changed 90 degrees, and in 16 other trials the orientation did not change. Subjects needed to indicate if there was a change (right button) or not (left button).

Two other types of trials consisted of a blank trial (20 trials) and a null trial (8 trials). In null trials the sample and test cues were empty, so there is no load on WM. In the blank trials, only the red fixation cross is shown for a variable time between 1000ms and 8000ms. These trials were implemented to prevent prediction effects and to compare brain activation between experimental trials (change trial/no change trial), visual input in general (null trial) and no visual input (blank trial).

Data analysis – behavior data

The capacity of working memory is measured by the performance on the change detection task. The performance is measured by the accuracy of the response. Cowan’s k is used as a measure of working memory capacity, which is a number between 0 and 6 (Rouder, Morey, Morey, & Cowan, 2011). Higher values of k correspond to a higher performance. It is calculated with the following equation:

𝑘 = 𝑁 (ℎ − 𝑓)

where N corresponds to the number of items, h to the hit rate in the change-trials, and f to the false alarm rate in the change-trials and the no-change trials. For the statistical analyses ANOVA’s were calculated in R to compare the performance on the working memory task between athletes and controls. After analysing the demographic information of the subjects, we found that the athletes in the open-skill group had significantly less training years, number of trainings per week and performed on a lower competition level compared to the closed-skill group (see Table 1). Therefore, we excluded the group comparison between the open and closed skill group and focused only on the comparison between athletes in general and controls, and the comparison between closed skill athletes and matched controls (see Table 2).

(9)

Athletes Controls

Gender Age Education sport expertise Gender Age Education

Female 23 WO Synch. swimming Female 21 WO

Male 31 HBO Rowing Female 23 WO

Male 33 VWO Rowing Male 28 HAVO

Male 23 WO Rowing Male 23 WO

Male 28 HBO Rowing Male 24 WO

Male 22 VWO Rowing Male 22 VWO

Female 32 HBO Rowing Female 26 WO

Female 29 WO Rowing Female 29 WO

Female 35 WO Rowing Female 27 VWO

Table 2. In the second group comparison subjects were matched to closed skill athletes based on gender, age and education.

One athlete had no data for the second run so was excluded in this group comparison.

Data analysis - functional data

The neural correlates of working memory are measured by changes of BOLD signal in voxels in the brain. This data was analyzed with software FSL (Jenkinson, Beckmann, Behrens, Woolrich, & Smith, 2012). Preprocessing steps included motion correction, non-brain removal, spatial smoothing (5mm) and high-pass temporal filtering. Functional scans were registered to high-resolution EPI images, which were registered to T1 images, which were registered to standard MNI152 space. A linear model was fitted to the data. For the first-level analyses ten events were modelled as regressors for the change detection paradigm to compare activation patterns between the 4 different trial conditions. See Table 3 for these contrasts. Contrast 10 was the contrast of interest. It consisted of the difference in activation when subjects had a load on working memory compared (change trials + no change trials) to when they did not have a load on working memory (null trial). With this contrast it is measurable what the executive features are of the WM network, because it compares trials where WM is used with trials where the visual input is similar, but where subjects only expected to use their WM.

The first-level analyses were the input for the two group comparisons. For both group comparisons (athletes vs controls & closed skill athletes vs controls) four events were

(10)

modelled to compare activation patterns between the two groups (see Table 3). Contrast 3 shows which regions athletes have stronger activations than controls, and contrast 4 shows in which regions controls have stronger activations than athletes. This data was then cluster corrected, so that only clusters with a z-score of at least 2.3 and a p-value lower than .05 survived.

First level analysis Blank Change No change Null Group analysis 1 Athletes Controls

C1 Average blank 1 0 0 0 C1 Average athlete 1 0

C2 Average change 0 1 0 0 C2 Average controls 0 1

C3 Average no change 0 0 1 0 C3 Athletes > controls 1 -1

C4 Average null 0 0 0 1 C4 Controls > athletes -1 1

C5 Change > blank -1 1 0 0

C6 No change > blank -1 0 1 0 Group analysis 2 Closed ath. Controls

C7 Change > null 0 1 0 -1 C1 Average closed ath. 1 0

C8 No change > null 0 0 1 -1 C2 Average control 0 1

C9 WM > blank -2 1 1 0 C3 Closed ath. > control 1 -1

C10 WM > null 0 1 1 -2 C4 Control > closed ath. -1 1

Table 3. Contrasts used in the fMRI design. The numbers represent the weight given to this trial/group. A weight of 0 would

mean that activations for this condition are not taken into account. Left table shows the contrasts used in the first-level analyses in which the activations for each subject are measured. These contrasts give insight which regions are stronger activated compared to each condition. Contrast 9 and 10 are of particular interest as these show the regions that are involved in the perceptual features (contrast 9) and executive features (contrast 10) of working memory. The right table shows the contrasts used in the group comparisons in which the results of the first-level analyses was averaged and compared between groups. Contrast 3 and 4 are of particular interest as these show the regions that are stronger activated for athletes than for controls on the first-level contrasts (contrast 3) and vice-versa (contrast 4).

Results

Three subjects did not fully complete the study. One control subject did not complete the second run, and one control subject did not complete both runs due to problems with the fMRI scanner. One closed skill athlete did not have complete data for the second run, due to problems with data export. In total, two group comparisons have been carried out for the behavioural data and the functional data. The first one is between all athletes (N=21) and all controls (N = 21). The second is between the closed skill athletes (N = 14) and controls (N = 14) that are matched on age, gender, and education.

(11)

Figure 4. The behavioural scores in the change-detection paradigm for the first group comparison between athletes and

controls. Error bars represent the standard error.

Behavioural results

Figure 4 shows the scores of the change-detection task for the group comparison between athletes and controls. It was found that there was no difference in WM capacity between athletes (Math = 1.19, SDath = 2.20) and controls (Mcon = 1.97, SDcon = 1.64) for run 1: t(37) = .014, p = .988, and for run 2 (Math = 3.54, SDath = 1.50; Mcon = 3.11, SDcon = 1.22): t(38) = .989, p = .329. Subjects scored significantly better in the second run, as was shown in a paired samples t-test (Mrun1 = 1.97, SDrun1 = 1.92; Mrun2 = 3.33, SDrun2 = 1.37): t(40) = 5.136, p < .001. This might be explained by a practice-effect and that subjects get more used to scanner conditions in the second run.

Figure 5 shows the scores of the change-detection task for the group comparison between closed-skill athletes and matched controls. Also, for this comparison it was found that there was no difference in WM capacity between the closed-skill group (Math = 1.93, SDath = 2.20) and the controls (Mcon = 1.71, SDcon = 1.42) for run1: t(22) = .305, p = .763. For the second run there was also no difference observed (Math = 3.66, SDath = 1.38; Mcon = 2.97, SDcon = 1.16): t(25) = 1.445, p = .160. Similar to the first group comparison, subjects scored better in the second run (Mrun1 = 1.82, SDrun1 = 1.83; Mrun2 = 3.32, SDrun2 = 1.30): t(27) = 3.087, p = .004. Therefore, we only focused on the second run for the functional data for both group comparisons.

(12)

Figure 5. The behavioural scores in the change-detection paradigm for the second group comparison between closed skill

athletes and matched controls. Error bars represent the standard error.

Figure 6. Average activation for the three groups in the WM > null condition. The WM network in the closed-skill group and

(13)

Voxel P value Z score X Y Z Brain region 509 7.75e-07 3.67 -24 -32 64 Somatosensory cortex

313 0.000185 3.69 -10 -58 46 Praecuneus cortex/superior parietal lobe 291 0.000365 3.54 -36 -54 60 Superior parietal lobe

Figure 7. For the first group comparison, these are the regions where athletes have stronger activations than controls. In red is

the postcentral gyrus, in green is the praecuneus cortex and in blue is the superior parietal lobe.

Figure 8. For the first group comparison, these are the regions where controls have stronger activation than athletes. In red is

the dorsal occipital cortex, in green is the cingulate gyrus, and in blue is the orbital frontal cortex.

fMRI results – first group comparison

Figure 6 shows the average activation of the three subject groups. It is similar to the WM network of prior research by Pessoa et al. (2002). The results of the group comparisons are

Voxel P value Z score X Y Z Brain region

469 9.39e-12 4.41 -48 -64 48 Lateral occipital cortex 181 3.14e-05 3.39 -4 38 30 Anterior cingulate gyrus 84 0.0274 3.73 44 60 -12 Orbitofrontal cortex

(14)

based on the first-level contrast WM > null. It measures the executive features of the WM network.

For the first group comparison we compared all athletes with controls. It was found that athletes showed stronger activations in parietal cortices, such as the postcentral gyrus and superior parietal lobe, than controls (see Figure 7). This is in line with the findings of Pessoa et al. (2002). In the opposite contrast, controls showed stronger activations in left lateral occipital lobe and in frontal regions, such as the paracingulate gyrus and orbitofrontal cortex (see Figure 8). These are regions that are associated with executive functions, such as WM (Osaka et al., 2003; Pessoa & Ungerleider, 2004).

2096 3.74e-24 4.13 8 -50 6 Posterior cingulate cortex 1223 2.7e-16 4.24 48 -72 18 Lateral occipital cortex 987 7.13e-14 4.4 70 -4 0 Superior temporal gyrus

Figure 9. For the second group comparison, these regions where closed-skill athletes have stronger activations than controls.

(15)

597 2.14e-15 4.79 -48 -64 48 Lateral occipital cortex 201 2.5e-06 3.55 -4 38 30 Anterior cingulate cortex 119 0.00081 3.94 44 58 -12 Orbitofrontal cortex

Figure 10. For the second group comparison, these regions were the regions where controls have stronger activations than

closed-skill athletes. In red is the lateral occipital cortex, in green the paracingulate gyrus and in blue the orbitofrontal cortex.

fMRI results – second group comparison

For the second group comparison we analysed a subset of the data by comparing the closed-skill athletes with controls that were matched on gender age and education. It was found that the closed-skill athletes showed stronger activations in temporal and parietal regions in the right hemisphere and in posterior cingulate gyrus (see Figure 9). This area is involved in self-reflection (Johnson et al., 2006) and the default mode network (R. Leech, Kamourieh, Beckmann, & Sharp, 2011). In the opposite contrast, controls showed similar activation patterns as in the first group comparison; stronger activations in the left occipital lobe and frontal regions (see Figure 10).

Discussion

The purpose of our study was to examine the differences in working memory capacity and the neural circuitry responsible for working memory between elite athletes and controls. We recruited 21 athletes and 20 controls that performed a change-detection task in a fMRI scanner. Our primary findings were that elite athletes do not have a higher WM capacity than controls. Nevertheless, subjects perform the WM task we found that athletes show stronger activation patterns in parietal regions, whereas controls show stronger activations in frontal regions. This implicates that athletes and controls have different strategies for the change-detection task. Athletes rely on stimulus driven processes, whereas controls might rely on goal-directed processes.

Behavioural data

Both group comparisons showed that athletes and controls do not have a different WM capacity. As such, we reject the first hypothesis that elite athletes have a higher WM capacity than controls. The first group comparison consisted of the difference between all athletes and controls. The second comparison consisted of the difference between the closed skill athletes

(16)

and controls. Though the difference in WM capacity was not significant for the second comparison, it was remarkably higher than in the first comparison: t(25) = 1.445, p = .160.

A potential explanation for failing to observe superior performance in athletes can be attributed to levels of sport expertise. In the second comparison, we excluded the open-skill athletes, that performed on a lower level than the closed-skill group. Thus, it might be that athletes that perform on an elite level have a better WM capacity. However, we were not able to collect enough data, hence this difference can only be investigated properly if the sample size would be increased.

Activations in athlete group

We found different patterns in the athlete group across the two comparisons. In the first comparison we found regions in the parietal lobe to be active. These regions included parts of the superior parietal lobe (SPL) and postcentral gyrus. The SPL is associated with maintaining and manipulation of information in WM (Koenigs, Barbey, Postle, & Grafman, 2009; Wolpert, Goodbody, & Husain, 1998). The ability to manipulate spatial information is essential for athletes. For example, when a football player sees his team member run across the field he should be able to predict his spatial location, so he can give a through ball to his team mate. The postcentral gyrus is part of the somatosensory cortex, an area that is involved in the spatial coding of touch (Tamè et al., 2012). A possible explanation for athletes to have stronger activations in somatosensory areas might be that they are in a state of action readiness, in which they can perform faster motor responses based on the sensations of their body.

In the second comparison we found activation more widely distributed across the parietal, temporal and occipital lobe. These regions are associated with action planning and control (Gallivan & Culham, 2015). Athletes showed activations in posterior cingulate cortex, right lateral occipital cortex and superior temporal gyrus. Activations in the LOC are associated with encoding of visual stimuli within in the change-detection paradigm (Pessoa et al., 2002). This makes sense as athletes are performing in an environment that requires many visually-guided actions.

Activations in the posterior cingulate cortex are associated with both internally-directed cognition as well as externally-internally-directed cognition. It is suggested that this region is important in shifting attention along this internal/external dimension (Leech & Sharp, 2014). This region stronger activations for closed-skill athletes, which might suggest that these athletes have a higher ability to switch between attention. It is interesting that we only found

(17)

this for the second comparison, as closed-skill sports require athletes to be internally focused to perform the sport (Voss et al., 2010).

Activations in control group

The activation patterns in the control group were consistent across the comparisons, showing activations in the frontal lobe (anterior cingulate cortex and orbital frontal cortex) and the dorsal occipital area in the left hemisphere. The activations in the ACC are associated with cognitive control (Shenhav, Cohen, & Botvinick, 2016), whereas activations in OFC are associated with decision making (Wallis, 2007). These functions are important when WM is active. It is suggested that activations in these frontal regions are less active, when one has mastered a task (Diamond, 2013). It might be that athletes learned the task faster, resulting in stronger frontal activations for controls. It seems that athletes have more activation in parietal regions, whereas controls have more activations in frontal regions.

Different cognitive strategies

The main finding is that athletes and controls show different neural patterns during change detection. Athletes show stronger activations in parietal lobe and controls show stronger activations in the frontal lobe. The frontal lobe is important in goal-directed behaviour, in which a goal is selected upon a delayed reward (Hasselmo, 2005). In sports, these behaviours are useful for action selection (Yarrow, Brown, & Krakauer, 2009), as the correct actions lead to the achieved goals.

The parietal lobe in the human brain is important for saving and manipulating representations of sensory stimuli (Koenigs et al., 2009; Wolpert et al., 1998). Regions in parietal lobe such as the SPL are therefore central parts of the WM network (Pessoa et al., 2002). Saving and manipulating stimuli is essential in sports. For example, when a soccer player runs into an opponent with the ball, he has to react very fast, and without thinking deliberately he has to make the right decision, while also keeping in mind other concepts of the game.

Our results suggest athletes and controls rely on different cognitive strategies for WM. We found that athletes have stronger activations in parietal regions, whereas controls have stronger activations in frontal regions. This suggest that athletes are able to rely on perception when performing this WM task, whereas controls need goal-directed strategies. This implicates that athletes rely on a bottom up strategy, and that controls rely on a top down strategy.

(18)

Limitations and implications

The current study had some limitations. First, we only did a whole-brain analysis. Another approach for analysing fMRI data is by a region of interest analysis (Poldrack, 2007). Based on prior research or data exploration, predefined regions are masked and analysed between groups. In that way, it is measurable how different regions in a WM network differ between groups. In the current whole-brain analyses most of these regions did not survive the cluster-correction.

Second, we did not measure the differences between open-skill sports and closed-skill sports. The reason for this was the small and heterogenous sample size of the open-skill group. Therefore, we were unable to test the third hypothesis that open-skill athletes have a higher WM capacity than closed-skill athletes.

Lastly, in the current study we analysed the individual differences of WM capacity between elite athletes and controls. Yet, we did not consider the difference between individual and team sports. It has been hypothesized that elite athletes do not show a difference in WM capacity in team sport athletes, as these individual differences are overruled by team-performance (Hambrick, Macnamara, Campitelli, Ullén, & Mosing, 2016). Though it is challenging to find a well-balanced elite athlete group for scientific studies, future research might take these shortcomings into account by testing for differences between sport types.

Conclusion

The current study is the first to measure the neural correlates of WM in elite athletes. Though no difference was observed in WM capacity between athletes and controls, the functional data suggest athletes have a different cognitive strategy. Since we found that athletes have stronger activations in parietal regions, and controls have stronger activations in frontal regions, this might suggest that athletes rely on a bottom up strategy and controls rely on a top down strategy.

(19)

References

Buszard, T., Masters, R. S., & Farrow, D. (2017). The generalizability of working-memory capacity in the sport domain. Current Opinion in Psychology, 16, 54–57.

https://doi.org/10.1016/j.copsyc.2017.04.018

Buszard, T., & Masters, R. S. W. (2017). Adapting, correcting and sequencing movements: does working-memory capacity play a role? International Review of Sport and Exercise

Psychology, 0(0), 1–21. https://doi.org/10.1080/1750984X.2017.1323940

Christophel, T. B., Klink, P. C., Spitzer, B., Roelfsema, P. R., & Haynes, J. D. (2017). The Distributed Nature of Working Memory. Trends in Cognitive Sciences, 21(2), 111–124. https://doi.org/10.1016/j.tics.2016.12.007

Diamond, A. (2013). Executive Functions. Annual Review of Psychology, 64(1), 135–168. https://doi.org/10.1146/annurev-psych-113011-143750

Furley, P. A., & Memmert, D. (2010a). Differences in spatial working memory as a function of team sports expertise: the Corsi Block-tapping task in sport psychological assessment.

Perceptual and Motor Skills, 110(3), 1–8. https://doi.org/10.2466/PMS.110.3.801-808

Furley, P. A., & Memmert, D. (2010b). The role of working memory in sport. International

Review of Sport and Exercise Psychology, 3(2), 171–194.

https://doi.org/10.1080/1750984X.2010.526238

Furley, P., & Wood, G. (2016). Working Memory, Attentional Control, and Expertise in Sports: A Review of Current Literature and Directions for Future Research. Journal of

Applied Research in Memory and Cognition, 5(4), 415–425.

https://doi.org/10.1016/j.jarmac.2016.05.001

Gallivan, J. P., & Culham, J. C. (2015). Neural coding within human brain areas involved in actions. Current Opinion in Neurobiology, 33, 141–149.

https://doi.org/10.1016/j.conb.2015.03.012

Hambrick, D. Z., Macnamara, B. N., Campitelli, G., Ullén, F., & Mosing, M. A. (2016). Beyond Born versus Made: A New Look at Expertise. Psychology of Learning and

Motivation - Advances in Research and Theory, 64, 1–55.

https://doi.org/10.1016/bs.plm.2015.09.001

Hasselmo, M. (2005). A model of prefrontal cortical mechanisms for goal-directed behavior.

Journal of Cognitive Neuroscience, 17(7), 1115–1129.

https://doi.org/10.1162/0898929054475190.A

Jenkinson, M., Beckmann, C. F., Behrens, T. E. J., Woolrich, M. W., & Smith, S. M. (2012). Fsl. NeuroImage, 62(2), 782–790. https://doi.org/10.1016/j.neuroimage.2011.09.015

(20)

Johnson, M. K., Raye, C. L., Mitchell, K. J., Touryan, S. R., Greene, E. J., &

Nolen-Hoeksema, S. (2006). Dissociating medial frontal and posterior cingulate activity during self-reflection. Social Cognitive and Affective Neuroscience, 1(1), 56–64.

https://doi.org/10.1093/scan/nsl004

Koenigs, M., Barbey, A. K., Postle, B. R., & Grafman, J. (2009). Superior Parietal Cortex Is Critical for the Manipulation of Information in Working Memory. Journal of

Neuroscience, 29(47), 14980–14986. https://doi.org/10.1523/JNEUROSCI.3706-09.2009

Leech, R., Kamourieh, S., Beckmann, C. F., & Sharp, D. J. (2011). Fractionating the Default Mode Network: Distinct Contributions of the Ventral and Dorsal Posterior Cingulate Cortex to Cognitive Control. Journal of Neuroscience, 31(9), 3217–3224.

https://doi.org/10.1523/JNEUROSCI.5626-10.2011

Leech, R., & Sharp, D. J. (2014). The role of the posterior cingulate cortex in cognition and disease. Brain, 137(1), 12–32. https://doi.org/10.1093/brain/awt162

Osaka, M., Osaka, N., Kondo, H., Morishita, M., Fukuyama, H., Aso, T., & Shibasaki, H. (2003). The neural basis of individual differences in working memory capacity: An fMRI study. NeuroImage, 18(3), 789–797. https://doi.org/10.1016/S1053-8119(02)00032-0 Pessoa, L., Gutierrez, E., Bandettini, P., & Ungerleider, L. (2002). Neural correlates of visual

working memory: fMRI amplitude predicts task performance. Neuron, 35(5), 975–987. https://doi.org/10.1016/S0896-6273(02)00817-6

Pessoa, L., & Ungerleider, L. G. (2004). Neural Correlates of Change Detection and Change Blindness in a Working Memory Task. Cerebral Cortex, 14(5), 511–520.

https://doi.org/10.1093/cercor/bhh013

Poldrack, R. A. (2007). Region of interest analysis for fMRI. Social Cognitive and Affective

Neuroscience, 2(1), 67–70. https://doi.org/10.1093/scan/nsm006

Rouder, J. N., Morey, R. D., Morey, C. C., & Cowan, N. (2011). How to measure working memory capacity in the change detection paradigm. Psychonomic Bulletin and Review,

18(2), 324–330. https://doi.org/10.3758/s13423-011-0055-3

Seo, J., Kim, Y. T., Song, H. J., Lee, H. J., Lee, J., Jung, T. Du, … Chang, Y. (2012).

Stronger activation and deactivation in archery experts for differential cognitive strategy in visuospatial working memory processing. Behavioural Brain Research, 229(1), 185– 193. https://doi.org/10.1016/j.bbr.2012.01.019

Shenhav, A., Cohen, J. D., & Botvinick, M. M. (2016). Dorsal anterior cingulate cortex and the value of control. Nature Neuroscience. https://doi.org/10.1038/nn.4384

(21)

F. (2012). The contribution of primary and secondary somatosensory cortices to the representation of body parts and body sides: An fMRI adaptation study. Journal of

Cognitive Neuroscience, 24(12), 2306–2320. https://doi.org/10.1162/jocn_a_00272

Voss, M. W., Kramer, A. F., Basak, C., Prakash, R. S., & Roberts, B. (2010). Are Expert Athletes ‘Expert’ in the Cognitive Laboratory? A Meta-Analytic Review of Cognition and Sport Expertise. Applied Cognitive Psychology, 24(June 2009), 812–826.

Wallis, J. D. (2007). Orbitofrontal Cortex and Its Contribution to Decision-Making. Annual

Review of Neuroscience, 30(1), 31–56.

https://doi.org/10.1146/annurev.neuro.30.051606.094334

Wang, C. H., Chang, C. C., Liang, Y. M., Shih, C. M., Chiu, W. S., Tseng, P., … Juan, C. H. (2013). Open vs. Closed Skill Sports and the Modulation of Inhibitory Control. PLoS

ONE, 8(2), 4–13. https://doi.org/10.1371/journal.pone.0055773

Wolpert, D. M., Goodbody, S. J., & Husain, M. (1998). Maintaining internal representations: the role of the human superior parietal lobe. Nature Neuroscience, 1(6), 529–533. Yarrow, K., Brown, P., & Krakauer, J. W. (2009). Inside the brain of an elite athlete: The

neural processes that support high achievement in sports. Nature Reviews Neuroscience,

10(8), 585–596. https://doi.org/10.1038/nrn2672

Zimmer, H. D. (2008). Visual and spatial working memory: From boxes to networks.

Neuroscience and Biobehavioral Reviews, 32(8), 1373–1395.