Neuroimage: Reports 2 (2022) 100112

Contents lists available at ScienceDirect

Neuroimage: Reports
journal homepage: www.sciencedirect.com/journal/neuroimage-reports

Video game players have improved decision-making abilities and enhanced

brain activities
Timothy Jordan a, Mukesh Dhamala a, b, c, d, e, *
a
Department of Physics and Astronomy, Georgia State University, Atlanta, GA, 30303, USA
b
Neuroscience Institute, Georgia State University, Atlanta, GA, 30303, USA
c
Department of Mathematics and Statistics, Georgia State University, Atlanta, GA, 30303, USA
d
Centers for Behavioral Neuroscience and for Diagnostics and Therapeutics, Georgia State University, Atlanta, GA, 30303, USA
e
Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), Georgia State University, Georgia Institute of Technology, and Emory
University, Atlanta, GA, USA

A B S T R A C T

Video game playing is a popular activity which provides a cognitively engaging, sensory rich environment that can lead to cognitive benefits in those who play
frequently. How exactly they change our brain to achieve these cognitive benefits has yet to be known. In a functional magnetic resonance imaging (fMRI)
experiment, we examined the behavioral and brain responses of video game- players (VGP) and non-video game-players (NVGP) during decision-making tasks. In
behavioral response, VGP were overall faster by approximately 190 ms and more accurate by 2% than NVGP. In brain response, comparing percent signal changes in
commonly activated brain regions between groups, we found that video gamers had increased task-related signal changes in the right lingual gyrus, right supple­
mentary motor area (SMA), and left thalamus associated with improved behavioral response. Directed functional network activities to the right SMA and the left
thalamus were also increased. The regional signal changes and network activities of all participants were found to be negatively correlated with decision response
time, indicating that higher the node and network activities better the performance. These results provide novel insights into the brain mechanisms that underlie
improvements in sensorimotor decision-making abilities due to video game playing.

1. Introduction
In our everyday goal-directed tasks, we focus on the relevant aspects
of incoming sensory information from the external world and utilize
them to transform into motor actions. In each stage of processing in the
brain, such transformations deal with alternative choices for what,
where, when and how-questions. This is a part of the decision-making
process, which involves the integration of information across multiple
sub-processes including sensation, perception, and action carried out by
a large-scale brain network of multiple regions (Gallivan et al., 2018;
Siegel et al., 2011, 2015). Many of these perceptual and sensorimotor
decisions require quick and precise responses. Little is known whether
and how fun, perceptually and cognitively engaging activities like video
game playing change the brain and behavioral responses for
decision-making. Beneficial changes due to such activities have impli­
cations in cognitive training that would need to increase accuracy or
decrease response time without the commonly seen speed-accuracy
trade-off (Derosiere et al., 2021; Drugowitsch et al., 2015; Heitz,
2014; Spieser et al., 2017) in order for the effect to be considered an
improvement.
Since video game playing often demands rapidly utilizing incoming
sensory information and precisely making decisions repeatedly,
decision-making tasks are relevant for probing the effects of video-game
playing. Video game playing has been shown to increase working
memory (Anguera et al., 2013; Basak et al., 2008), better attention
control (Anguera et al., 2013; Bavelier, 2007; Greenand and Bavelier,
2015; Wuand and Spence, 2013), increase spatial resolution (Bavelier,
2007; Li et al., 2009), and make players more efficient at task switching
(Basak et al., 2008; Oeiand and Patterson, 2014). These effects have also
been shown to be trainable in those who do not already play video
games (Anguera et al., 2013; Wuand and Spence, 2013; Lynch et al.,
2010; Rosser, 2007). While these studies show that video game playing
has cognitive benefits for certain tasks and that video game playing can
be used to train a person to perform better, video game playing has also
been associated with negative effects. Video game playing in excess can
lead to behavioral symptoms in players similar to other addictive dis­
orders such as withdrawal, inability to restrain themselves from the
activity and inability to fully engage in everyday activities (Weinstein
et al., 2017; Lejoyeux, 2020). However, when played in moderation,
video games can lead to beneficial effects (Anguera et al., 2013; Basak

* Corresponding author.
E-mail addresses: timjordan.neurophys@gmail.com (T. Jordan), mdhamala@gsu.edu (M. Dhamala).

https://doi.org/10.1016/j.ynirp.2022.100112
Received 3 January 2022; Received in revised form 9 April 2022; Accepted 11 June 2022
Available online 22 June 2022
2666-9560/© 2022 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
T. Jordan and M. Dhamala
et al., 2008; Bavelier, 2007; Greenand and Bavelier, 2015; Wuand and
Spence, 2013; Li et al., 2009; Oeiand and Patterson, 2014). The neural
underpinnings of these beneficial effects in decision-making have yet to
be fully understood.
Understanding the changes in brain and behavior measures and the
brain-behavior relation due to video game playing is also important for
learning about the effects of other daily challenging activities in
decision-making. What activity could a person do regularly that allows
for improved decision-making? How does this improvement present in
the brain to achieve these increased abilities?
Brain imaging studies on video game players (VGP) are much more
limited in what they have examined. Around half the studies have
looked solely at how video game playing correlates to violence, and the
other half have primarily been resting-state studies of VGP in video
game environments via presented pictures or are examining addictive
effects of gaming (Weinstein et al., 2017; Lejoyeux, 2020; Palaus et al.,
2017). These studies have primary looked at brain activations (Granek
et al., 2010; Richlan et al., 2017), resting-state functional connectivity
(Momi et al., 2020) and voxel-based morphometry (Tanaka et al., 2013).
Although these studies have examined neural differences between VGP
and non-video game players (NVGP), network interactions during
decision-making tasks have not been studied.
In this functional magnetic resonance imaging (fMRI) study, we
examined the effects of video game playing in sensorimotor decision-
making by comparing the brain and behavioral responses of VGP and
NVGP. We assessed whether video game playing increases decision-
making capabilities, and, if it does, how video game playing alters the
brain regional and network activities. We hoped to provide new insights
into the brain mechanisms and the brain-behavior relation associated
with the effects of video-game playing. Based on the nature of sensory
information-rich, interactive environment that video game playing
2
Neuroimage: Reports 2 (2022) 100112
Fig. 1. Experimental Design and Stimuli. Organi­
zation of task time line and how decision difficulty
levels are determined. A shows the time line that the
task followed. Each task period was 15 s, containing
three trials of moving dots (MD) stimuli, followed 15
s of rest. Each task period was randomly assigned a
difficulty level and speed setting for all three MD
inside that period. B Color wheel to determine diffi­
culty level. Easy difficulty was assigned for primary
versus primary color. Medium difficulty was assigned
for primary versus adjacent secondary color. Hard
difficulty was assigned for primary or secondary
versus adjacent tertiary color. (For interpretation of
the references to color in this figure legend, the
reader is referred to the Web version of this article.)
offers and type of sensorimotor tasks used, we hypothesized that VGP
would have better task performance and that VGP would have increased
sensorimotor node and network activities compared to NVGP.
2. Materials & methods
2.1. Materials
2.1.1. Participants
For this neuroimaging study, 47 participants (28 VGP and 19 NVGP)
were recruited to participate. Groups were age-matched (VGP = 20.6 ±
2.4, NVGP = 19.9 ± 2.6). All participants were required to fill out a
questionnaire about their video game playing for the past two years to
determine which group they would be placed in. Using the similar
criteria as Green et al. (Bavelier, 2007; Stewart et al., 2020 ), partici­
pants indicating playing 5 h per week or more in one of four types of
video games genres for the last two years were considered to be video
game players. The four types of video game player genres we recruited,
based on industry demographics, were First-Person Shooter (FPS),
Real-Time Strategy (RTS), Multiplayer Online Battle Arena (MOBA), and
Battle Royale (BR) players. Participants indicating they have played less
than 5 h per week in any video games over the last two years were
considered non-video game players. NVGP in this study averaged less
than an hour per week. All participants were also required to complete
and pass the full Ishihara’s Test for Color Deficiency (Clark, 1924).
Participants provided signed written consent forms and health screen­
ings prior to their scheduled scan session. The Institutional Review
Boards of Georgia State University and Georgia Institute of Technology,
Atlanta, Georgia, approved this study. Participants were compensated
for their participation in the experiment.
T. Jordan and M. Dhamala Neuroimage: Reports 2 (2022) 100112

Fig. 2. Brain Activations and Commonly Activated Nodes as ROIs (N ¼ 47, p < 0.01 (corrected for multiple comparisons with Family-wise Error Rate),
cluster extent k > 20) A Activation maps for all participants (VGP and NVGP) to determine commonly activated regions to be used as regions of interest for the
group-level comparative analyses of percent signal change and connectivity. Three regions, left thalamus, right lingual gyrus, and right supplementary motor area,
were found to have significant regional differences in percent signal change between groups are labeled. B A BrainNet view of the commonly activated 25 nodes (the
cerebellum is not displayed as it is not available in the version of BrainNet used). Node numbers follow the same order as Table 1.

2.1.2. Stimuli
A modified left-right moving dots (MD) motion categorization task
was used for this study. The MD task began with a 2 s cue for a specific
color, i.e. red. On the screen following the cue, participants would see
two sets of 600 moving dots going the same speed in opposite directions.
One set of the dots would be the cued color and the other set would be an
interference set that needed to be ignored by participants. Participants
would have 3 s to respond with what direction they thought the cued
dots were moving via a button box controller. This stimulus was
repeated three times for a total of 15 s of task and followed by 15 s of
rest. The direction of the cued set of dots was randomized for each
stimulus. The color of the cued dots and interfering dots were random­
ized for each stimulus based on the color pairings for the current task
block difficulty setting. These difficulty pairings were based on the
standard color wheel and divided into three difficulties; easy, medium,
and hard. Easy difficulty pairings were between two primary colors (red,
yellow, and blue), medium difficulty pairings were between a primary
color and its adjacent secondary colors, and hard difficulty pairings were
between a primary or secondary color and it’s adjacent tertiary color.
The speed setting of the dots was also randomized for each block of 3 MD

3
T. Jordan and M. Dhamala Neuroimage: Reports 2 (2022) 100112

tasks. The speed settings ranged from no-motion (speed setting 0) to the Table 1
highest speed setting (speed setting 4). Speed setting 4, and the other Brain Activations at FWE corrected p < 0.01 and cluster extent threshold k
speeds were determined from finding the max speed of the dots before > 20. Region of Interests selected from the common brain activations for Task-
illusory motion reversal became possible. This maximum speed was then versus-Rest and Motion-versus-No-Motion contrasts.
divided into quarters to created the speed settings. Fig. 1 shows the Contrast Region of Interest MNI T
timescale of the stimulus and the difficulty color pairings used for this Coordinates
study. The task sequence was designed in and displayed by the PsychoPy x, y, z (mm)
stimulus software (Peirce et al., 2019). Task - Rest Left Calcarine − 6,-93, − 10 18.27
Right Middle Cingulate 6, 17, 42 14.67
2.1.3. Experimental design Right Inferior Frontal Orbital 25, 29, − 6 14.53
Before their scheduled MRI scan sessions, participants were shown a Right Inferior Frontal 44, 17, 26 14.26
Triangularis
demonstration of the task they would be completing, and all questions
Right Hippocampus 25, − 28, − 2 14.43
pertaining to the task were answered. Participants were told to respond Left Insula − 28, 21, 2 14.54
as quickly and accurately as possible for all MD tasks. Participants were Right Insula 36, 21, 2 16.13
informed of the total time of the scan session, safety protocols, and given Right Lingual 21, − 82, − 14 19.74
an emergency button to reduce anxiety inside the fMRI. During the scan Left Midbrain − 6, − 28, − 6 13.59
Right Midbrain 6, − 28, − 6 14.21
session, participants were asked to lay still and not to move during the Right Superior Occipital 17, − 93, 6 18.2
scan periods. All motion movements were monitored during scan ses­ Left Inferior Parietal − 44, − 40, 42 14.94
sions, and participants were notified if they began to move too much. Left Inferior Parietal 2 − 28, − 51, 46 14.9
The stimulus was displayed on a monitor at the back of the fMRI that Right Inferior Parietal 32, − 51, 46 13.16
Right Precentral 48, 10, 34 12.96
participants viewed using a mirror placed atop the head coil. To respond
Right Putamen 17, 13, − 2 13.45
inside the fMRI, participants would indicate if the dots were moving left, Left Supplementary Motor Area 2, 10, 50 14.77
right, or not at all by pressing the left or right buttons on the button box Left Thalamus − 13, − 6, − 2 12.94
controller with their thumbs or no button press at all if no-motion. Motion - No- Right Calcarine 13, − 89, − 2 9.07
Motion Left Middle Occipital − 25, − 89, 14 7.49
Right Middle Occipital 40, − 66, 6 5.97
Left Superior Occipital − 21, − 89, 22 7.04
2.2. Data collection and analysis Left PostCentral − 47, − 25, 46 5.93
Right Supplementary Motor 10, 6, 50 6.63
2.2.1. Behavioral data collection Area
Left SupraMarginal − 47, − 25, 34 5.34
Stimuli were presented and participants’ behavioral response data
Vermis 9 2, − 59, − 38 7.55
were collected by using the stimulus software, PsychoPy (Peirce et al.,
2019). Participant’s decisions via button presses and response times
were recorded for each MD task. Participant’s left/right decisions were 2.2.3. Selecting regions of interest (ROIs)
translated into participant accuracy scores. Participants had to respond Task-related activations were computed by using the two-level
correctly to the MD within the stimulus period to be considered correct random-effects analysis approach (Friston, 2010). For the first-level
answers. Responses after the stimulus ended, incorrect responses, and no analysis, data was analyzed using a mass univariate approach based
response at all, for any condition other than no-motion, were considered on general linear models (GLMs). The GLM was defined as y = βX + ε,
incorrect answers. The correct responses would then be divided by the where y is the observed data, β is a matrix of regression parameters, X is
total number of tasks and multiplied by 100 to give each participant a the design matrix constructed from the stimulus and task sequence and
percentage score, which was their task accuracy score. Response times participant’s movements, and ε is the residual error. The GLM design
were taken as the time from the MD onset after the text cue. Participants matrix was specified for each participant based on when a specific task
had to respond within 3 s of the MD onset. condition would have appeared for them over four runs. The specified
task conditions were: Easy Speed 0 (E0), Easy Speed 1 (E1), Easy Speed
2.2.2. fMRI data 2–3 (E23), Easy Speed 4 (E4), Medium Speed 0 (M0), Medium Speed 1
Whole-brain structural and functional MR imaging was conducted on (M1), Medium Speed 2–3 (M23), Medium Speed 4 (M4), Hard Speed
a 3 T S Magnetom Prisma MRI scanner at the joint Georgia State Uni­ 0 (H0), Hard Speed 1 (H1), Hard Speed 2–3 (H23), Hard Speed 4 (H4)
versity and Georgia Institute of Technology Center for Advanced Brain and six motion parameters. Contrast images for Task - Rest and Motion -
Imaging, Atlanta, Georgia. First, high-resolution anatomical images No-Motion were created for each participant. ROIs were defined using
were acquired for voxel-based morphometry and anatomical reference 2nd-level analysis activation results for all participants for the two
using a T1-MEMPRAGE scan sequence (TR = 2530 ms, TE1-4: contrasts. Activation clusters that all participants showed, regardless of
1.69–7.27 ms, TI = 1260 ms, Flip Angle = 7 deg, Voxel size 1 mm × group, were identified with family-wise error (FWE) correction at p <
1 mm x 1 mm). Following, four functional runs used a T2*-weighted 0.01 and cluster extent k > 20, shown in Fig. 2A. The coordinates for
gradient echo-planar imaging sequence (TR = 535 ms, TE = 30 ms, peak cluster activations that both groups commonly had were recorded
Flip Angle = 46◦ , Field of View = 240 mm, Voxel Size = 3.8 mm × 3.8 and used in the PickAtlas (Maldjian et al., 2003) as the regions of interest
mm x 4 mm, Number of slices = 32 collected in an interleaved order, (ROIs). Fig. 2B and Table 1 show all 26 ROIs selected in total and from
Slice thickness = 4 mm) and acquired 3440 brain images while partic­ each contrast they came from.
ipants performed the behavioral tasks. Each of the four functional runs
was 7 min and 30 s long, All fMRI data was preprocessed using the 2.2.4. Granger causality analyses
Statistical Parametric Mapping Matlab software suite (SPM12) (Friston, ROIs were resampled to Talairach space using AFNI’s 3dresample,
2010). Data was first imported from DICOM format into NIFTI, then slice and then voxel-wise BOLD time-series were extracted using AFNI’s
time corrected, realigned for motion correction, and realigned for field 3dmaskdump (Cox, 1996; Coxand and Hyde, 1997). Fig. 2 shows the
distortion using field map corrections. Each participant’s data was then ROIs used. Time series were extracted from each ROI using 6 mm
co-registered to their anatomical image and normalized to a Montreal spherical masks created in the MarsBaR software package (Brett et al.,
Neurological Institute (MNI) template using the unified segmentation. 2002). Time-series data for Granger causality calculations were
The normalized images were then spatially smoothed with an 8 mm normalized, voxel averaged, segmented, and detrended. All partici­
isotropic Gaussian kernel. pants’ data were combined along the second dimension as trials to

4
T. Jordan and M. Dhamala Neuroimage: Reports 2 (2022) 100112

Fig. 3. Decision Response Time and Accuracy.

Moving dots left-right discrimination task perfor­
mance for overall performance and difficulty setting
performance. A & B show the overall response time
and accuracy, respectively, by group. C & D show the
difficulty based accuracy and response time, respec­
tively, by group. Significant values are displayed
above their respective comparisons. VGP were overall
faster by approximately 190 ms in decision response
time (A & C) and more accurate by 2.24% (B) than
NVGP.

calculate the proper model order in parametric modeling for GC calcu­ transfer function matrices in the vector autoregressive model (Dhamala
lations. Once the model order was determined, GC matrices were et al., 2008a, 2008b).
computed for each participant using the pairwise GC (Dhamala et al., The frequency-band specific or time-domain equivalent GC is:
2008a, 2008b) and blockwise GC (WANG et al., 2007; Wen et al., 2013). ∫ f2
1
We used the parametric approach to estimate GC with a model order of 5 F2→1 = I2→1 (f )df , (2)
in all of the analyses. The optimal model order was determined by f2 − f1 f1
examining differences in the power spectra from the nonparametric
where f1 = 0.05 Hz and f2 = 0.9 Hz. The sampling rate TR− 1 was 1/535
approach (Dhamala et al., 2008b) with the parametric approach at
ms ≈ 1.87 Hz, where TR is the MR repetition time in the functional runs.
different model orders and selecting the model order that gave the
This rapid sampling rate at which the fMRI BOLD signals were collected
lowest power difference as in our previous work (Adhikari et al., 2014;
makes GC analyses much more appropriate.
Chand and Dhamala, 2016). Granger causality from region 2 to region 1
The GC threshold for statistical significance was computed by using
in the frequency domain is defined as:
the random permutation technique (Blairand and Karniski, 1993; Bro­
S (f ) velli et al., 2004) under the null model of no statistical interdependence
I2→1 (f ) = ln ( 11 2 )⃒ (1)
Σ ⃒
S11 (f ) − Σ22 − Σ1211 ⃒H12 (f )|2 with 200 random permutation samples. The significance of the group
differences was determined by using the Mann-Whitney U test
(McKnightand and Najab, 2010).
where Σ’s and H’s are the elements of the noise covariance and the
Region (node)-specific outflow and inflow were computed from
these band-specific Granger causality matrices. Outflow from m-node is
Table 2 a sum of the causal influences from that m-node to others divided by the
∑
Decision-making behavioral response. Group task performance accuracy (in number of rest of the nodes, i.e., Outflowm = N−1 1 Ni=1 Im→i , where Im→i is
%) and response time (RT in ms) for all difficulty levels and speed settings. a causal influence from m-node to i-node, the self-causality is zero and N
Condition VGP NVGP p is the total number of nodes. Inflow was taken as the sum of the causal
General Accuracy 95.26 ± 3.93 93.016 ± 5.63 0.0008 influences from the other nodes to the m-node divided by N − 1.
70
RT 925.61 ± 432.18 1117.23 ± 494.78 2.05 × 10−
Hard Accuracy 89.29 ± 7.98 86.25 ± 12.45 0.16
16
RT 1123.07 ± 505.45 1296.59 ± 539.06 2.14 × 10−
Medium Accuracy 98.29 ± 2.58 97.12 ± 4.56 0.13
27
RT 866.34 ± 375.77 1043.51 ± 447.21 4.65 × 10−
14
Easy Accuracy 98.45 ± 2.98 91.33 ± 4.48 1.68 × 10−
43
RT 794.68 ± 329.14 1004.24 ± 418.41 1.68 × 10−

5
T. Jordan and M. Dhamala
Fig. 4. Percent Signal Change Differences between Groups. Average percent signal change for regions with significant differences between VGP (green) and
NVGP (orange) at Bonferroni corrected p < 0.01. Left thalamus, right lingual gyrus, and right supplementary motor area were significantly different in percent signal
between groups. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
3. Results
3.1. Video game players have increased decision accuracy and reduced
response time
Decision task accuracy and response time (the time from MD onset to
the response (maximum of 3 s allowed)) were collected for each
participant. VGP and NVGP were compared in their behavioral response
for overall performance and difficulty setting using a t-test. Both groups
showed increased response time as difficulty level increased. For all
conditions and overall performance, VGP had faster response times and
higher accuracy than NVGP. VGP were significantly more accurate than
NVGP for task performance overall (by 2.24 ± 6.87%) and easy condi­
tion tasks (7.12 ± 5.38%). All response times regardless of condition
were significantly lower in VGP than in NVGP with an average differ­
ence of 190 ms (VGP = 926.75 ± 424.71, NVGP = 1115.93 ± 483.89).
Fig. 3 & Table 2 shows all results for behavioral task performance. For all
plots, dark green and dark orange correspond to VGP and NVGP
respectively.
3.2. VGP show increased activity compared to NVGP correlating with
behavior performance
Participants’ brain response was recorded by a whole-brain 3T MRI
while they completed the MD tasks. Task-related activations were
computed by using the two-level random-effects analysis approach
(Friston, 2010). From commonly task activated regions (see Fig. 2B and
Methods), we examined the activity of regions that both groups utilized
for the MD tasks. Regional average percent signal change (APSC) for
each condition specified in the 1st-Level per run was extracted from each
ROI using 6 mm spherical masks and MarsBaR (Brett et al., 2002). Each
participant had 12 values per ROI that were then averaged to obtain the
overall APSC. Significant differences between group APSC per region
were assessed by using a Mann-Whitney U test. Fig. 4 shows the regions
that were significantly different between groups.
Three regions showed increased activity at a significance of Bon­
ferroni corrected p < 0.01. These regions were the right lingual gyrus
(rLING) (p = 2.93 × 10− 5), left thalamus (lTHA) (p = 1.15 × 10− 4), and
right supplementary motor area (rSMA) (p = 8.84 × 10− 12). VGP
showed higher activity than NVGP in all three regions. We then exam­
ined if any of these regions’ activities correlated back to behavioral
response times and plotted APSC against decision response time (RT).
6
Neuroimage: Reports 2 (2022) 100112
APSC values were correlated with RT using a Spearman’s rank correla­
tion coefficient test thresholded at p < 0.05. Both rLING (r = − 0.34, p =
0.022) and rSMA (r = − 0.3, p = 0.046) showed a negative correlation
with decision response time for all participants. Fig. 5 displays scatter
plots for these two brain regions that had significant correlation.
3.3. Increased connectivity in VGP for decision-making tasks
As shown in the previous section, three regions showed significantly
higher activations for VGP than for NVGP. We also examined whether
and how the network activities of these regions were different between
groups. Utilizing Granger Causality (GC), we calculated the directed
functional connectivity among these three regions of interest (ROIs).
The statistical significance tests of spectral GC results at p < 0.01 were
established by using the random permutation method with a null model
of no interdependence. The permutation method thus enacted the cor­
rections for multiple comparisons across frequency points of spectral GC
values. The group results were then compared for statistical significance
using a Mann-Whitney U test. This analysis left us with four connections
(edges) that were significantly different between VGP and NVGP at p <
0.01 (uncorrected for multiple comparisons for the number of possible
connections (tests)). Fig. 6A shows the connections we found to be sig­
nificant, nodes are labeled with which regions were involved and the
arrow points towards the node being influenced. Two connections to
lTHA were found to be greater for VGP. These connections were from
right inferior frontal gyrus par orbitalis (rIFGo) (p = 0.003; VGP = .075
± 0.001, NVGP = .043 ± 0.002) and rLING (p = 0.003; VGP = .158 ±
0.003, NVGP = .1 ± 0.002). The other two connections were to rSMA,
both greater for VGP, from left Calcarine (lCAL) (p = 0.003; VGP = .153
± 0.003, NVGP = .08 ± 0.002) and left inferior parietal lobule node 2
(lIPL2) (p = 0.008; VGP = .148 ± 0.003, NVGP = .09 ± 0.003).
Pursuing the same question about correlation back to behavior dif­
ferences, GC values were plotted against RT. Spearman’s correlation
coefficients were computed for each connection. One connection from
rIFGo to lTHA was found to have a moderate negative correlation (r =
− 0.32, p = 0.03) with response time as shown in Fig. 7.
Finally, the connections that were different between groups were
from the rest of the nodes to lTHA and to rSMA. We further examined to
see if either rSMA or lTHA showed differences in causal source or sink
activities between groups. Utilizing block Granger causality (WANG
et al., 2007; Wen et al., 2013), we computed source and sink activities.
Source activity is defined as the sum of the influences of a region to all
T. Jordan and M. Dhamala
Fig. 5. Percent Signal Change vs Decision Response Time. Scatter plot
correlating regional percent signal change to response time with VGP points in
green and NVGP points in orange. A-B Participant Region percent signal change
vs response time in ms. Right lingual gyrus and right supplementary motor area
were negatively correlated with response time. Corresponding Spearman’s
correlation coefficients and p-values are displayed in the top right of each plot.
(For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)
other regions, and sink activity is the total influence to a region from all
other regions. We found that lTHA was utilized more as a sink region for
VGP compared to NVGP (corrected p = 0.009). Fig. 6B shows the group
distributions for the thalamus sink activity and how they differed be­
tween groups.
4. Discussion and conclusion
The sensorimotor decision-making task we used started with a color
prompt (cue) immediately followed by a visual display of moving dots
for a participant to make decision on the motion of moving dots and
finally to actually execute a choice (left or right) with a motor response
(Fig. 8A). This sensorimotor decision-making task can be hypothesized
to involve multiple subprocesses beginning with sensory information
relay to the brain, sensory evidence accumulation to perception, and,
finally, mapping choice to action with attention control being present
7
Neuroimage: Reports 2 (2022) 100112
Fig. 6. Directed Functional Connectivity. A BrainNet (Xia et al., 2013)
transparent brain displaying the directed functional connectivity graph of
significantly different connections between VGP and NVGP. All connections
were found to be significantly higher in VGP compared to NVGP at p < 0.05
(uncorrected). B. Group differences for thalamic sink activity (total inflow to­
wards the thalamus from all other nodes) during MD task at Bonferroni cor­
rected p < 0.05.
across the whole process (Siegel et al., 2011, 2015) as illustrated in
Fig. 8B. We used the blood oxygenation level dependent fMRI mea­
surements and investigated the brain node and network activities during
the entire decision-making process in video gamers (VGP) and non-video
gamers (NVGP), and compared their brain and behavioral responses.
This study thus allowed us to examine and compare the decision-making
abilities and underlying behavioral response-related brain mechanisms
in VGP and NVGP to assess the effects of video game playing in
decision-making. These comparative results uncovered that VGP and
NVGP differed both in their brain and behavioral responses for
decision-making.
With our modified MD decision-making tasks, we found that VGP
had overall increased decision accuracy and decreased decision response
time. The response time was overall lower in VGP by about 190 ms (VGP
= 926.75 ± 424.71 ms, NVGP = 1115.93 ± 483.89 ms). VGP were more
accurate for overall accuracy (2.24 ± 6.87%) and easy condition tasks
(7.12 ± 5.38%). Both groups also showed a linear increase in response
time as difficulty increased. This difference in response time over diffi­
culty was also found to be significantly different indicating the higher
difficulty required more from participants in order to execute a choice.
These behavioral results of VGP are consistent with previous behavioral
and neuroimaging studies that reported video game playing related
T. Jordan and M. Dhamala
Fig. 7. Granger causality versus decision response time. Granger causality
from the right inferior frontal pars orbitalis to the left thalamus was negatively
correlated with decision response time (with VGP points in green and NVGP in
orange). Corresponding Spearman’s correlation coefficient and p-value are
displayed in the top right of the plot. (For interpretation of the references to
color in this figure legend, the reader is referred to the Web version of
this article.)
improvements in task performance on various other kinds of tasks
(Bavelier, 2003, 2007; Greenand and Bavelier, 2015; Palaus et al.,
2017).
Based on the known positive effects of video game playing (Palaus
et al., 2017), sensorimotor task improvements in VGP can be associated
with enhancements in various aspects of low-level and high-level brain
processes including sensation, attention, spatial navigation, cognitive
control and workload. The sensorimotor decision-making task with
moving color dots used here activated in VGP and NVGP the brain re­
gions known for visuomotor and cognitive processing. As listed in
Table 1, among these commonly activated regions are the left and right
supplementary motor area (SMA), left and right midbrain, right
Fig. 8. Stimulus and Sensorimotor Decision-Making Task, and Hypothesized Subprocesses. Sensorimotor decision-making requires the processing and inte­
gration of information through multiple subprocesses. A. Stimulus and Task sequence B. A breakdown of the decision-making process into its subprocesses for our
task sequence as shown in A: sensation, attention selection and control, perception and motor action.
8
Neuroimage: Reports 2 (2022) 100112
putamen, right lingual gyrus, left inferior parietal lobe, left middle oc­
cipital gyrus, right inferior frontal pars triangularis, cerebellar vermis,
left Thalamus, right hippocampus. These activations here in sensori­
motor decision-making tasks are consistent with the neural implications
of previously reported behavioral and imaging studies to various aspects
of sensorimotor and cognitive processing (Palaus et al., 2017; Chand and
Dhamala, 2016; Lamichhane et al., 2016).
Among these commonly activated regions, three regions right LING,
left thalamus and right SMA differed in their decision-task related %
signal change between VGP and NVGP as shown in Fig. 4. VGP had
significantly higher levels of % signal change. From the analysis of the
percent signal change versus response time for all participants (as shown
in Fig. 5), we see that rLING and rSMA activation levels are correlated
with decision response time. The higher activity in rLING and rSMA, the
lower the decision time. These regions are known to be associated with
visuomotor processing. Previous studies have shown that LING is uti­
lized for word comprehension (Valdois et al., 2006), visual selective
attention (Hahn et al., 2006) and visual encoding of complex stimuli
(Machielsen et al., 2000). Right LING activations correlated with RT
could be associated with attention control consistent with previous
findings of increased attention control in VGP (Anguera et al., 2013;
Bavelier, 2007; Greenand and Bavelier, 2015; Wuand and Spence,
2013). Supplementary motor area activity has been associated with
motor planning and selection (Makoshi et al., 2011; Shibasaki et al.,
1993; Jenkins et al., 1994; Mueller et al., 2007). The differential acti­
vations in these regions, LING and SMA, together are likely to be
involved in recognizing visual patterns, planning and selecting motor
acts guided by visual sensory information and associated with the
mapping to action as a part of the sensorimotor decision-making process
as illustrated in Fig. 8B.
Directed functional connectivity analyses with GC found four con­
nections significantly higher for VGP, but only one of these connections
was correlated with response time: GC from rIFGo to lTHA. Studies have
associated right IFGo with inhibition of response in No-Go tasks (Aron
et al., 2003), initiation and selection of choice, and integrating past
information with current information (Nogueira et al., 2017) while
linking the thalamus, through lesion studies, to playing a role in guid­
ance of visual attention by working memory (de Bourbon-Teles et al.,
2014). Specific associations between these regions may be crucial for
accessing and switching working memory (D. B. E and Tim B. M. T,
T. Jordan and M. Dhamala
2007). This leads us to think this connection relates to attentional and
working memory processes during the task. Finally, although the left
thalamus (lTHA) activity was not associated with response time, there
were multiple differences between groups. LTHA showed higher acti­
vations in VGP as well as increased sink activity (sum total of incoming
interactions). This increased sink activity was dominantly driven by the
two connections that were significantly increased in VGP, from rIFGo to
lTHA and rLING to lTHA.
As a potential activity for training cognitive skills, we needed to see if
the improvements seen in VGP with one aspect of speed-accuracy did
not come with a speed-accuracy trade off compared to NVGP. Previous
research shows that the faster the speed in which a choice is executed the
less accurate the participant becomes and vice versa (Derosiere et al.,
2021; Drugowitsch et al., 2015; Heitz, 2014; Spieser et al., 2017).
Studies in video games have also shown this to be present in VGP with
VGP being faster but less accurate (Heimler et al., 2014). This study
findings indicate, however, that VGP are not only faster in response, but
equally or more accurate than NVGP. This lack of speed-accuracy trade
off would indicate video game playing as a good candidate for cognitive
training as it pertains to decision-making.
This study being presented has two limitations. First, fMRI is limited
by its time resolution capabilities and indirect measures of brain activity
(Babiloni et al., 2009; Glover, 2011). This temporal restriction puts the
restraint for not being able to capture cognitive processing timescales of
subprocesses, as illustrated in Fig. 8, to be able to fully explain the
behavioral differences. Future studies can utilize EEG data recordings
and their source analyses to resolve time scale limitations and complete
the full scope of brain mechanisms to underpin these response differ­
ences. Studies using EEG could also vary rest-periods between task
phases and the number of tasks per task period to reduce the predictable
time duration of attention-focus. Second, this study was not longitudinal
and the underlying assumption about the participants, drawn from
similar backgrounds, was that they only differed in their video-game
playing experience. Although implications for cognitive training were
discussed, we did not specifically train participants and measure
brain-behavior responses. Future studies should focus on training
non-video game players, establish training time required to achieve
measurable differences in behavioral and brain responses for
decision-making.
To summarize, this study demonstrated that video game players have
improved performance on decision-making tasks and that these differ­
ences correlate to specific changes in the node and network activities in
and across the lingual gyrus, the supplementary motor area and the
thalamus. These results indicate that video game playing potentially
enhances several of the subprocesses for sensation, perception and
mapping to action to improve decision-making skills. These findings
begin to illuminate how video game playing alter the brain in order to
improve task performance and their potential implications for
increasing task specific activity. This study leads to not only a potential
method of cognitive training, but also understanding how the training
will affect the brain.
Conflict of interest statement
We do not have any conflict of interest to declare.
Acknowledgments
We thank B. Adhikari and V. Ahluwalia for various discussions and
suggestions on the data analyses. This work was supported by two in­
ternal funding sources: GSU Brain & Behavior seed grant and GSU-
GaTech Center for Advanced Brain Imaging seed grant to the author MD.
9
Neuroimage: Reports 2 (2022) 100112
References
Adhikari, B.M., Sathian, K., Epstein, C.M., Lamichhane, B., Dhamala, M., 2014.
Neuroimage 91, 300.
Anguera, J.A., Boccanfuso, J., Rintoul, J.L., Al-Hashimi, O., Faraji, F., Janowich, J.,
Kong, E., Larraburo, Y., Rolle, C., Johnston, E., Gazzaley, A., 2013. Nature 501, 97.
Aron, A.R., Fletcher, P.C., Bullmore, E.T., Sahakian, B.J., Robbins, T.W., 2003. Nat.
Neurosci. 6, 115.
Babiloni, C., Pizzella, V., Gratta, C.D., Ferretti, A., Romani, G.L., 2009. In: International
Review of Neurobiology, International Review of Neurobiology, vol. 86. Academic
Press, pp. 67–80.
Basak, C., Boot, W.R., Voss, M.W., Kramer, A.F., 2008. Psychol. Aging 23, 765.
Bavelier, C.S. Greenand D., 2003. Nature 423, 534.
Bavelier, C. Greenand D., 2007. Psychol. Sci. 18, 88.
Blairand, R.C., Karniski, W., 1993. Psychophysiology 30, 518.
Brett, M., Anton, J., Valabregue, R., Poline, J., 2002. Human Brain Mapp. Conf. 16, 497.
Brovelli, A., Ding, M., Ledberg, A., Chen, Y., Nakamura, R., Bressler, S.L., 2004. Proc.
Natl. Acad. Sci. USA 101, 9849.
Chand, G.B., Dhamala, M., 2016. Brain Connect. 6, 249.
Clark, J.H., 1924. Am. J. Physiol. Opt. 5, 269.
Cox, R.W., 1996. Comput. Biomed. Res. 29, 162.
Coxand, R.W., Hyde, J.S., 1997. NMR Biomed. 10, 171.
D. B. E, Tim, C., B. M. T, 2007. Science 317, 215.
de Bourbon-Teles, J., Bentley, P., Koshino, S., Shah, K., Dutta, A., Malhotra, P., Egner, T.,
Husain, M., Soto, D., 2014. Curr. Biol. CB 24, 993.
Derosiere, G., Thura, D., Cisek, P., Duque, J., 2021. J. Neurophysiol. 126, 361. https://
doi.org/10.1152/jn.00038.2021.
Dhamala, M., Rangarajan, G., Ding, M., 2008a. Neuroimage 41, 354.
Dhamala, M., Rangarajan, G., Ding, M., 2008b. Phys. Rev. Lett. 100, 018701.
Drugowitsch, J., DeAngelis, G.C., Angelaki, D.E., Pouget, A., 2015. Elife 4, e06678.
Friston, K.J., 2010. Statistical Parametric Mapping: the Analysis of Functional Brain
Images. Elsevier AP.
Gallivan, J.P., Chapman, C.S., Wolpert, D.M., Flanagan, J.R., 2018. Nat. Rev. Neurosci.
19, 519.
Glover, G.H., 2011. Neurosurg. Clin. 22, 133.
Granek, J.A., Gorbet, D.J., Sergio, L.E., 2010. Cortex 46, 1165.
Greenand, C.S., Bavelier, D., 2015. Curr. Opin. Behav. Sci. 4, 103.
Hahn, B., Ross, T.J., Stein, E.A., 2006. Neuroimage 32, 842.
Heimler, B., Pavani, F., Donk, M., van Zoest, W., 2014. Attention, Perception, &
Psychophysics, 76, p. 2398.
Heitz, R.P., 2014. Front. Neurosci. 8, 150.
Jenkins, I., Brooks, D., Nixon, P., Frackowiak, R., Passingham, R., 1994. J. Neurosci. 14,
3775. https://www.jneurosci.org/content/14/6/3775.full.pdf.
Lamichhane, B., Adhikari, B.M., Dhamala, M., 2016. Brain Connect. 6, 558.
Lejoyeux, A. Weinsteinand M., 2020. Dialogues Clin. Neurosci. 22, 113, 32699511
[pmid].
Li, R., Polat, U., Makous, W., Bavelier, D., 2009. Nat. Neurosci. 12, 549.
Lynch, J., Aughwane, P., Hammond, T.M., 2010. J. Surg. Educ. 67, 184.
Machielsen, W.C., Rombouts, S.A., Barkhof, F., Scheltens, P., Witter, M.P., 2000. Hum.
Brain Mapp. 9, 156.
Makoshi, Z., Kroliczak, G., van Donkelaar, P., 2011. J. Mot. Behav. 43, 303.
Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H., 2003. Neuroimage 19, 1233.
McKnightand, P.E., Najab, J., 2010. Mann-whitney u test. In: The Corsini Encyclopedia of
Psychology. John Wiley & Sons, Ltd, 1–1.
Momi, D., Smeralda, C.L., Lorenzo, G.D., Neri, F., Rossi, S., Rossi, A., Santarnecchi, E.,
2020. Brain Imag. Behav. 15, 1518.
Mueller, V.A., Brass, M., Waszak, F., Prinz, W., 2007. Neuroimage 37, 1354.
Nogueira, R., Abolafia, J.M., Drugowitsch, J., Balaguer-Ballester, E., Sanchez-Vives, M.
V., Moreno-Bote, R., 2017. Nat. Commun. 8, 14823.
Oeiand, A.C., Patterson, M.D., 2014. Comput. Hum. Behav. 37, 216.
Palaus, M., Marron, E.M., Viejo-Sobera, R., Redolar-Ripoll, D., 2017. Front. Hum.
Neurosci. 11 https://doi.org/10.3389/fnhum.2017.00248.
Peirce, J., Gray, J.R., Simpson, S., MacAskill, M., Höchenberger, R., Sogo, H.,
Kastman, E., Lindeløv, J.K., 2019. Behav. Res. Methods 51, 195.
Richlan, F., Schubert, J., Mayer, R., Hutzler, F., Kronbichler, M., 2017. Brain Behav. 8,
e00877.
Rosser, J.C., 2007. Arch. Surg. 142, 181.
Shibasaki, H., Sadato, N., Lyshkow, H., Yonekura, Y., Honda, M., Nagamine, T.,
Suwazono, S., Magata, Y., Ikeda, A., Miyazaki, M., Fukuyama, H., Asato, R.,
Konishi, J., 1993. Brain 116, 1387.
Siegel, M., Engel, A.K., Donner, T.H., 2011. Front. Hum. Neurosci. 5 https://doi.org/
10.3389/fnhum.2011.00021.
Siegel, M., Buschman, T.J., Miller, E.K., 2015. Science 348, 1352.
Spieser, L., Servant, M., Hasbroucq, T., Burle, B., 2017. Psychonomic Bull. Rev. 24, 950.
Stewart, H.J., Martinez, J.L., Perdew, A., Green, C.S., Moore, D.R., 2020. Sci. Rep. 10,
14410.
Tanaka, S., Ikeda, H., Kasahara, K., Kato, R., Tsubomi, H., Sugawara, S.K., Mori, M.,
Hanakawa, T., Sadato, N., Honda, M., Watanabe, K., 2013. PLoS One 8, 1.
Valdois, S., Carbonnel, S., Juphard, A., Baciu, M., Ans, B., Peyrin, C., Segebarth, C., 2006.
Brain Res. 1085, 149.
Wang, X., Chen, Y., Bressler, S.L., Ding, M., 2007. Int. J. Neural Syst. 17, 71. https://doi.
org/10.1142/S0129065707000944.
Weinstein, A., Livny, A., Weizman, A., 2017. Neurosci. Biobehav. Rev. 75, 314.
T. Jordan and M. Dhamala Neuroimage: Reports 2 (2022) 100112

Wen, X., Rangarajan, G., Ding, M., 2013. Phil. Trans. Math. Phys. Eng. Sci. 371, Wuand, S., Spence, I., 2013. Attention, Perception, & Psychophysics, 75, p. 673.
20110610. Xia, M., Wang, J., He, Y., 2013. PLoS One 8, 1.

10