Co-activation patterns during viewing of different video game genres

Past research has revealed cognitive improvements resulting from engagement with both traditional action video games and newer action-like video games, such as action real-time strategy games (ARSG). However, the cortical dynamics elicited by different video gaming genres remain unclear. This study explored the temporal dynamics of cortical networks in response to different gaming genres. Functional magnetic resonance imaging (fMRI) data were obtained during eye-closed resting and passive viewing of gameplay videos of three genres: life simulation games (LSG), first-person shooter games (FPS), and ARSG. Data analysis used a seed-free Co-Activation Pattern (CAP) based on Regions of Interest (ROIs). When comparing the viewing of action-like video games (FPS and ARSG) to LSG viewing, significant dynamic distinctions were observed in both primary and higher-order networks. Within action-like video games, compared to FPS viewing, ARSG viewing elicited a more pronounced increase in the Fraction of Time and Counts of attentional control-related CAPs, along with an increased Transition Probability from sensorimotor-related CAPs to attentional control-related CAPs. Compared to ARSG viewing, FPS viewing elicited a significant increase in the Fraction of Time of sensorimotor-related CAPs, when gaming experience was considered as a covariate. Thus, different video gaming genres, including distinct action-like video gaming genres, elicited unique dynamic patterns in whole-brain CAPs, potentially influencing the development of various cognitive processes.


Introduction
In recent decades, video gaming experience has been a focal point in research on learning-related cognitive plasticity.Prior studies have shown that certain genres, such as action video games, have a stronger impact on cognitive improvement compared to other genres, such as life simulation games (LSG) (Bavelier and Green, 2019;Bediou et al., 2018).However, with the ongoing evolution of video gaming genres, certain video games, which were not traditionally classified as action video games, now incorporate elements typically associated with action video games.This evolution includes genres like real-time strategy games and action real-time strategy games (ARSG) (Dale et al., 2020a).There is a growing interest in determining whether these video gaming genres exert comparable cognitive effects to traditional action video games.
Traditional action video games, commonly known as first-or third-person shooter games (FPS and TPS), are characterized by action features, including perceiving swiftly moving targets, executing rapid actions under time constraints, managing high perceptual and workload demands, and flexibly shifting between focused and distracted attention (Bavelier and Green, 2019).Targets in these games are often scattered amidst cluttered stimuli, further intensifying the cognitive demands placed on players.Research suggests that playing action video games can lead to cognitive enhancement across various domains, including perception, attention, visuospatial cognition, working memory, and sensorimotor skills (Bediou et al., 2018;Kuehn et al., 2011).
Neuroscience evidence from cross-sectional studies suggests that compared to non-video gaming players, individuals who frequently engage in action video games tend to have larger gray matter volume in the right posterior parietal region, indicative of enhancements in visual working memory (Tanaka et al., 2013), as well as more flexible dynamic reconfiguration of bottom-up and top-down attentional networks (Bavelier et al., 2012).In addition, long-term action video gaming training is associated with increased cortical thickness in areas such as the somatosensory cortex and superior parietal lobule, reflecting improvements in sensorimotor skills (Momi et al., 2018), as well as altered grey matter volume in the hippocampus, which related to spatial cognition (West et al., 2018).
As video game genres expand and evolve, action elements are now integral to various genres.This has led to gaming experiences that combine action elements and other gameplay features, offering players a broader set of cognitive challenges and experiences.Newly developed video games, such as real-time strategy games and ARSG, integrate action and strategy elements, requiring not only quick reflexes and agile motor skills typical of traditional action video games but also strategic planning and collaborative teamwork similar to strategy video games (Dale and Green, 2017b).Thus, real-time strategy games and ARSG have comparable effects on cognitive development as traditional action video games.Indeed, ARSG players exhibit enhancements in the interconnection among insular subregions, attentional network, and sensorimotor network (SMN) (Gong et al., 2015), the inter-regional integration between the ventral attention network (VAN) and fronto-parietal network (FPN) (Gong et al., 2016), and the large-scale white matter networks involving prefrontal and sensorimotor networks (Gong et al., 2017).Real-time strategy gaming experience can also induce changes in the structural connections of occipital-parietal loops (Kowalczyk et al., 2018).By comparing static characteristics during resting, movie viewing, LSG playing, and ARSG playing, prior research found that ARSG playing elicited exceptionally high theta and beta power, which is associated with working memory and attention (Gong et al., 2019a).In addition, as an ARSG session started, resource allocation in the frontal lobe was progressively prioritized, revealing stronger information integration (Gong et al., 2019a).
Although playing ARSG is associated with cognitive enhancements like playing traditional action video games, certain distinctions still exist.Cross-sectional behavioral studies indicate that the performance of real-time strategy gaming players often parallels that of action video gaming players rather than non-action video gaming players in various cognitive tasks (Dale et al., 2020b); additionally, it typically falls between the performance levels of action video gaming players and non-action video gaming players (Dale and Green, 2017a).A recent study, examining dynamic differences between ARSG and FPS genres, using electroencephalogram (EEG) resting-state microstates, demonstrated that FPS emphasizes audition and vision, along with the interaction between audition and attention, whereas ARSG emphasizes vision and the interaction between vision and attention (Cui et al., 2021).
However, the specific dynamic cortical structures elicited by different video games remain unknown.This issue is critical for understanding the unique effect of different video games on brain and cognitive plasticity.Thus, this study explored the dynamic structures of cortical networks immediately elicited by passive viewing of gameplay videos of three genres (LSG, FPS, and ARSG), using a seed-free Co-Activation Pattern (CAP) method based on Regions of Interest (ROIs).The ROIs-based seed-free CAP method has been widely used to identify repetitive CAPs across the entire brain over a temporal scalea method that has robust reproducibility and generalizability (Sun et al., 2023;Yang et al., 2021Yang et al., , 2022)).Building on previous research (Bediou et al., 2018;Gong et al., 2019a), we hypothesized that action-like video games (FPS and ARSG) and LSG should elicit significant, extensive dynamic differences across whole brain networks.This prediction arose from the emphasis on action elements in both FPS and ARSG, distinguishing them from LSG, as well as from past findings that FPS and ARSG may differ in their effect on primary networks and the interaction between primary networks leading to higher-order networks (Cui et al., 2021).

Participants
Seventy-eight healthy adults were recruited from the University of Electronic Science and Technology of China (UESTC).Prior to the experiment, all participants completed an online questionnaire providing demographic information, including age, sex, visual ability, handedness, history of psychiatric or neurological illnesses, and video gaming experience.Additional assessments included the Beck Depression Inventory (BDI), State-Trait Anxiety Inventory (STAI), and a sevenquestion Game Addiction Questionnaire (GAQ).Inclusion criteria required participants to be right-handed, have normal or corrected-tonormal vision and hearing, have no history of neurological illnesses or contraindications for magnetic resonance imaging (MRI)-scanning, BDI scores < 30, S-AI/T-AI scores < 70, and have no symptoms of Internet gaming disorders.Seven participants were excluded due to excessive head movement (> 3 mm) during scanning, resulting in a total of 71 participants in the final sample (mean age = 21.18years, SD = 2.21 years; one female).See Table S1 for detailed demographic information.
Informed consent was obtained from each participant before the experiment.The experimental procedures adhered to the standards outlined in the Declaration of Helsinki.Furthermore, the study protocol received approval from the university Ethics Board.

Stimuli and processes
This study used three games: The Sims 4 (a life simulation game [LSG]), the 5-player vs. 5-player ranking mode in League of Legends (an action real-time strategy game [ARSG]), and the Call of Duty Online (a first-person shooter game [FPS]).In this study, FPS and ARSG were classified as action-like video games.The Sims 4 (an LSG) provided an active control condition, as it involved socializing and open-ended gameplay simulating real-life activities (Blacker et al., 2014;Zhang et al., 2021), thereby demanding a lower workload of perception, attention, and motor execution than both FPS and ARSG.League of Legends (an ARSG) featured clearly defined objectives, requiring players to form a 5-player team against another team with a focus on high attentional control, strategic planning, and teamwork (Gong et al., 2019a).Call of Duty Online (an FPS) involved rapid target searching and shooting at moving targets across varying distances with potential blockages by interfering objects, emphasizing attention and sensorimotor demands.
This study consisted of an active playing task followed by a passive viewing task, separated by a ten-day interval.The purpose of the active task was to obtain stimuli for the subsequent passive viewing task, where functional MRI (fMRI) data were collected.During fMRI data acquisition, as participants were instructed to passively watch gaming videos without engaging in active responses, mitigating drowsiness and enhancing participants' motivation and engagement became crucial.Thus, this study used each participant's recorded gameplay sessions as stimuli for their passive viewing task.
The active playing task.Participants visited the laboratory, where they first completed a 10-min tutorial for each game (The Sims 4, League of Legends, and Call of Duty Online), provided by the gaming software, to learn the basic rules of each game (Gong et al., 2019a).Then, they played the three games while their gaming sessions were recorded.Each gaming session lasted approximately 30 min, and the order of the three games was counterbalanced across participants.Following this, 510-s highlights from each video gaming recording were evaluated and selected as stimuli for the subsequent passive viewing task.The evaluation and selection of the highlights were completed by 18 video gaming experts.Detailed information regarding the selection of experimental stimuli is provided in Supplementary materials.
The passive viewing task.In the fMRI experiment (ten days after the completion of the active playing task), participants began with a 510-s closed-eye resting period, followed by passive viewing of three 510-s video gaming sessions, each separated by a 5-min break.The presentation order of the recorded gaming sessions was counterbalanced across participants.

Data acquisition and preprocessing
MRI data were obtained during the resting state and three video game viewing sessions using a 3T MRI scanner (GE MR750, General Electric Medical Systems, Milwaukee, WI, USA).High-resolution T1weighted structural images were obtained using a three-dimensional fast spoiled gradient echo (T1-3D FSPGR) sequence (parameter setting: Preprocessing steps were conducted using SPM12 (Statistical Parametric Mapping, https://www.fil.ion.ucl.ac.uk/spm/software/spm12/ ), following procedures similar to past research (Gong et al., 2019b).These steps included discarding the first five volumes in each session, correcting for slice timing and head movements, normalizing the fMRI images using a Tissue Probability Map (TPM) template, spatially smoothing with a Gaussian kernel of 8 mm full-width at half-maximum (FWHM), and applying temporal band-pass filtering between 0.01 and 0.1 Hz.Nuisance regression analysis was then performed, incorporating 24 head movement parameters, signals from white matter and cerebrospinal fluid, global signals, and linear drift signals.

Co-activation pattern calculation
The methodology for the seed-free Co-Activation Pattern (CAP) analysis based on Regions of Interest (ROIs) closely followed established protocols from prior studies (Sun et al., 2023;Yang et al., 2021).This included the following steps (Fig. 1): (1) Extraction of Blood Oxygenation Level Dependent (BOLD) signals from 400 ROIs (Schaefer et al., 2018) categorized into Yeo's 7 networks (Yeo et al., 2011).( 2) Z-score normalization for each time series, followed by data concatenation across all participants.(3) Implementation of the k-means cluster algorithm during the resting state (cluster number = 2-21, repeating times = 100) according to prior research (Sun et al., 2023;Yang et al., 2021Yang et al., , 2022)).Spatial similarity between volumes was gauged by computing the difference in Pearson correlation coefficients.The selection of results was based on the largest sums within clusters of spatial similarity between the initial map and fMRI volumes.( 4) Averaging volume distributions within a cluster and dividing them by the standard deviation to generate CAP z-maps.( 5) Application of the CAPs from the resting state to each participant during each video session using spatial similarity (Sun et al., 2023;Yang et al., 2021Yang et al., , 2022)).( 6) Utilization of eight clusters for subsequent analysis, determined by the silhouette score (Rousseeuw, 1987;Yang et al., 2021).
To explore the cognitive functions associated with these CAPs, spatial correlations were computed between the spatial distribution of these CAPs and the association maps of Neurosynth topics (Coronel-Oliveros et al., 2023;van der Meer et al., 2020).Neurosynth provides a collection of whole-brain item-to-activation topic maps related to various cognitive functions, extracted from tens of thousands of studies through meta-analyses (https://www.neurosynth.org/).Fifty-six topics were selected from the Neurosynth database, including ten topics extracted from prior literature on real-time strategy games R. Cui et al. (Coronel-Oliveros et al., 2023) and 46 topics derived from previous cognitive lists (Poldrack et al., 2011;Yarkoni et al., 2011), encompassing diverse aspects of the gameplay process.The special Neurosynth topics used in this study can be found in Supplementary materials.This approach created the cognitive profile for each CAP based on the 56 Neurosynth topics.Additionally, the significance of correlations was assessed using a spin-based spatial permutation test (Baller et al., 2022).
Four dynamic features were calculated for each CAP, following the approach established in previous research (Yang et al., 2021).The dynamic features included Fraction of Time (i.e., the percentage of volumes in one CAP relative to the total volumes in the entire time series), Persistence (i.e., the duration one CAP was maintained before transitioning to another), Counts (i.e., the frequency of occurrence of one CAP throughout the scanning session), and Transition Probability (i.e., the probability that a volume within CAP A transitions to the next volume belonging to CAP B).First, paired samples t-tests were conducted to compare the resting state with each gaming viewing.Then, to analyze the dynamic features induced by gaming viewing, differences between the dynamic features of each gaming viewing and those during the resting state were computed for each participant.A one-way repeated measures ANOVA analysis was then conducted to examine variations among the three video games for each dynamic feature.Multiple comparison corrections were applied using the false discovery rate method (FDR, p < 0.05).In addition, similar statistical procedures were repeated with age, sex, and gaming experience serving as covariates.

The relationships between CAPs and association maps of Neurosynth topics
The eight Co-Activation Patterns (CAPs) are shown in Fig. 2 and organized into four pairs with reversed spatial patterns based on spatial similarity (Fig. S1), with the spatial similarity exceeding 0.9 between the two CAPs in each pair.The absolute Z-value of each Region of Interest (ROI) indicated its deviation from the baseline level (i.e., the average time series; Z = 0), termed activation deviation.Larger absolute Z-values indicated stronger activation (depicted in red color in Fig. 2) or deactivation (depicted in blue color in Fig. 2).Specifically, CAP 1 exhibited a predominant association with the activated Sensorimotor Network (SMN) and Ventral Attention Network (VAN), alongside deactivation of the Default Mode Network (DMN), compared to other brain networks; in contrast, CAP 2 displayed the opposite pattern of results.In addition, CAP 3 exhibited a predominant association with the activated VAN, alongside deactivation of the Visual Network (VN) and SMN; in contrast, CAP 6, characterized by the dominance of SMN, VN, and Fronto-parietal Network (FPN), exhibited the opposite pattern of results.Furthermore, CAP 4 showed a predominant association with the activated FPN and Dorsal Attention Network (DAN) and the deactivated VN, whereas CAP 5, characterized by the dominance of VN, FPN, and VAN, exhibited the opposite pattern of results.Finally, CAP 7 exhibited a predominant association with the activated DAN and FPN, alongside the deactivated SMN and VAN; in contrast, CAP 8 displayed the opposite pattern of results (Table 1).
The correlations between the CAPs and the association maps of Neurosynth topics revealed distinctive functional profiles.CAP 1 exhibited a positive correlation with sensorimotor processing, including somatosensory, sensorimotor, motor imagery, coordination, body, timing, and multisensory aspects, while CAP 2 exhibited a negative association in these domains.The spatial distribution of CAP 3 exhibited positive associations with risk-taking, emotion, rewards, and social cognition, whereas CAP 6 displayed negative associations in these domains.CAP 4 exhibited a robust positive correlation with attentional control, covering attention, working memory, action observation, intention, switching, and executive function, whereas CAP 5 exhibited a negative correlation in these aspects.CAP 7 and CAP 8 were linked to association maps indicating visuospatial cognition, including navigation, visuospatial, and spatial attention (Fig. 3 and Table 1).

CAP dynamic features of different video games
First, dynamic features were compared between resting and each gaming viewing using paired samples t-tests.Regardless of the inclusion of covariates, all gaming viewing elicited significant, extensive changes in dynamic features across primary and higher-order networks.Further details can be found in Supplementary materials.
Next, the differences in the increase or decrease of dynamic features among the three games relative to resting were computed.to life simulation games (LSG), first-person shooter games (FPS) elicited a larger increase in CAPs 1 and 2, alongside a larger decrease in CAP 3 and a smaller increase in CAP 8. Furthermore, compared to LSG, action real-time strategy games (ARSG) elicited a larger increase in CAPs 1, 2, and 5, alongside a larger decrease in CAPs 3 and 6.Additionally, ARSG elicited a larger increase in CAP 5 compared to FPS (Fig. 4a).For Persistence, CAP 3 exhibited a significant main effect (F = 9.84, p < 0.001).Post-hoc analysis showed that both FPS and ARSG elicited a larger decrease in CAP 3 compared to LSG, with no significant difference between ARSG and FPS (Fig. 4a).For Counts, significant main effects were found in CAPs 1 (F = 7.89, p = 0.002), 5 (F = 10.39,p < 0.001), and 6 (F = 4.36, p = 0.039).Post-hoc analyses revealed that compared to LSG, FPS elicited a larger increase in CAP 1, while ARSG exhibited a larger increase in CAP 5, alongside a larger decrease in CAP 6.Moreover, ARSG elicited a larger increase in CAP 5 compared to FPS (Fig. 4a).
A similar set of data analyses was then done, with age, sex, and gaming experience included as covariates, which yielded consistent results.For Fraction of Time, significant main effects were observed in almost all CAPs except CAPs 4, 7, and 8 (CAP 1: F = 13.66,p < 0.001; CAP 2: F = 7.33, p = 0.003; CAP 3: F = 10.95,p < 0.001; CAP 5: F = 4.51, p = 0.021; CAP 6: F = 5.33, p = 0.014).Post-hoc analyses revealed that compared with LSG, FPS elicited a larger increase in CAPs 1 and 2, alongside a larger decrease in CAP 3. Furthermore, compared to LSG, ARSG elicited a larger increase in CAPs 1, 2, and 5, alongside a larger decrease in CAPs 3 and 6.Additionally, compared to FPS, ARSG elicited a larger increase in CAP 5 and a smaller increase in CAP 1 (Fig. 4b).For Persistence, CAP 3 exhibited a significant main effect (F = 9.89, p < 0.001).Post-hoc analyses showed a pattern of results similar to the analyses where the covariates were not included (Fig. 4b).For Counts, significant main effects were found in CAPs 1 (F = 8.07, p = 0.002) and 5 (F = 9.80, p < 0.001).Post-hoc analyses revealed that both FPS and ARSG elicited larger increases in CAP 1 compared to LSG, while ARSG elicited a larger increase in CAP 5 compared to both LSG and FPS (Fig. 4b).
Significant main effects were observed in Transition Probability between CAP 1 and CAP 4 (from CAP 1 to CAP 4: F = 9.60, p = 0.004; from CAP 4 to CAP 1: F = 6.18, p = 0.046), within CAP 3 (F = 10.83,p = 0.004), and from CAP 8 to CAP 5 (F = 6.75, p = 0.036).Post-hoc analyses revealed that compared with LSG, FPS elicited a larger increase in Transition Probability from CAP 4 to CAP 1 and from CAP 8 to CAP 5, alongside a larger decrease in Transition Probability within CAP 3. Furthermore, compared to LSG, ARSG elicited a larger increase in Transition Probability from CAP 1 to CAP 4 and from CAP 8 to CAP 5, alongside a larger decrease in Transition Probability within CAP 3. Finally, ARSG elicited a larger increase in Transition Probability from CAP 1 to CAP 4 compared to FPS (Fig. 5).Data analyses, with age, sex, and gaming experience included as covariates, yielded consistent results.Significant main effects were observed in Transition Probability between CAP 1 and CAP 4 (from CAP 1 to CAP 4: F = 8.35, p = 0.013; from CAP 4 to CAP 1: F = 6.13, p = 0.048), within CAP 3 (F = 10.62,p = 0.006), and from CAP 8 to CAP 5 (F = 6.15, p = 0.048).Post-hoc analyses revealed a pattern of results similar to the analyses where covariates were not included.

Discussions
This study explored the distinctive dynamic structures immediately elicited by passive viewing recordings of three gameplay sessions (life simulation games [LSG], first-person shooter games [FPS], and action real-time strategy games [ARSG]), using a seed-free Co-Activation  Patterns (CAP) method based on Regions of Interest (ROIs).The results revealed significant, extensive dynamic disparities in both primary and higher-order networks when comparing action-like video gaming (FPS and ARSG) viewing to LSG viewing.In addition, compared to FPS, ARSG elicited a significant increase in Fraction of Time and Counts in CAP 5, which is linked to attentional control.Compared to ARSG, FPS elicited a significant increase in Fraction of Time in the CAP 1, which is linked to sensorimotor processing, when gaming experience was included as a covariate.Furthermore, ARSG elicited a heightened Transition Probability from CAP 1 to CAP 4 compared to FPS, indicating a more pronounced shift from sensorimotor processing to attentional control during ARSG viewing compared to FPS viewing.

CAPs indicated distinct functional correlations
This study used the ROIs-based seed-free CAP analysis to identify recurrent spatial activation patterns over various time points, thereby elucidating their dynamics (Yang et al., 2021).The reliability and consistency of this analytical approach in both spatial patterns and temporal dynamics have been rigorously validated across diverse preprocessing steps, task states, cluster numbers, ROI numbers, and multiple independent datasets (Yang et al., 2021).In addition, this analytical approach has been widely used in studies involving naturalistic tasks, such as movie viewing (Freitas et al., 2020;Sun et al., 2023) and investigations into psychiatric disorders including schizophrenia and depression disorders (Kaiser et al., 2019;Yang et al., 2021).
The current study identified four pairs of CAPs characterized by contrasting spatial activations, including both primary and higher-order networks, with each pair predominantly associated with specific functional networks and cognitive functions.For example, CAPs 1 and 2 were dominated by the [SMN & VAN] -DMN pattern, associating with the sensorimotor processing according to Neurosynth topics.This is consistent with previous findings that (a) SMN is linked to primary motor and somatosensory processes, (b) VAN indicates bottom-up attention, and (c) DMN is often referred to as a task-negative network (Uddin et al., 2019;Yang et al., 2021;Yeo et al., 2011).Additionally, CAPs 3 and 6 The CAP 4-5 pair is associated with attentional control according to Neurosynth topics (e.g., attention, working memory, switching, and executive functions), which is consistent with previous findings (Uddin et al., 2019;Yeo et al., 2011).Furthermore, the CAP 7-8 pair is primarily associated with visuospatial cognition, according to Neurosynth topics, reflecting the functions of FPN, DAN, SMN, and VAN (Uddin et al., 2019;Yeo et al., 2011).These results highlight the distinct network configurations and cognitive profiles observed within the identified pairs of CAPs.

The extensive dynamic differences between action-like video games and life simulation games
The current findings revealed dynamic differences across various CAPs elicited by action-like video games (FPS and ARSG) and life simulation games (LSG).Notably, the Fraction of Time and Counts related to sensorimotor CAPs exhibited a larger increase during the viewing of FPS and ARSG compared to that of LSG.ARSG also elicited a larger increase in Fraction of Time and Counts of attentional controlrelated CAPs than LSG.Additionally, compared to LSG, action-like video games elicited a larger increase in Transition Probability from visuospatial cognition-related CAPs to attentional control-related CAPs, as well as between attentional control-related CAPs and sensorimotorrelated CAPs (FPS: from attentional control-related CAPs to sensorimotor-related CAPs; ARSG: from sensorimotor-related CAPs to attentional control-related CAPs).These results align with previous findings that based on static characteristics, action-like video games can elicit higher workload, attention, and information integration than resting, movie viewing, and LSG playing (Gong et al., 2019a).The immediate elicitation observed in this study may contribute to the cognitive enhancement associated with action-like video gaming training.This perspective is consistent with research evidence indicating that action-like video games can influence the development of cognition, brain networks for various cognitive functions (e.g., perception, attention, visuospatial cognition, working memory, sensorimotor skills), and the relevant neural plasticity (Bediou et al., 2018;Kuehn et al., 2011).
This study found a smaller increase in the Fraction of Time of visuospatial CAPs during FPS viewing compared to LSG viewinga result that did not remain significant when gaming experience was considered as a covariate.It is important to note that the existing literature on the impact of action video gaming experience on spatial navigation is mixed.While some research has indicated that action video gaming training can result in increased cortical thickness in the bilateral parahippocampal cortex (Momi et al., 2018) and heightened functional integration between the parahippocampal cortex and thalamus (Momi et al., 2021), indicating enhanced spatial cognition, other studies have shown that action video gaming players have reduced gray matter volume in the hippocampus than non-video gaming players (West et al., 2018).Moreover, the effect of action video gaming experience on gray matter volume in the hippocampus may be dependent on the navigational strategies individuals employ (West et al., 2018).Specifically, the use of response learning strategies during action video games has been associated with a decrease in the grey matter volume of the hippocampus, whereas the use of spatial strategies has been associated with an increase.These findings suggest that individual differences in navigation strategies may influence the impact of action video gaming experience on brain development.Thus, consistent dynamic structures related to visuospatial cognition may not be observed across participants in a study, including the current one.Nevertheless, this explanation needs to be tested by future research.In contrast, The Sims 4 tasked players with navigating through various spatial environments, including rooms, pathways, and malls, potentially engaging visuospatial cognition and thereby prompting dynamic changes in associated neural networks.However, action-like video games (FPS and ARSG) elicited a heightened Transition Probability from CAP 8 to CAP 5 than LSG, suggesting that action-like video games required more interaction between visuospatial cognition and attentional control.
We also found that video gaming viewing led to a decrease in the Fraction of Time and Persistence of the CAPs associated with risk-taking, emotion, rewards, and social cognition.Furthermore, compared to LSG viewing, FPS and ARSG viewing elicited a more pronounced decrease in the Fraction of Time and Persistence of these CAPs.This decrease was also evident in the Transition Probability within these CAPs, with ARSG and FPS eliciting a larger decrease in Counts compared to LSG.However, it is important to note that past research has consistently reported activation of reward-related regions during active video gaming play (e. g., the striatum, caudate, and putamen) (Bavelier and Green, 2019;Koepp et al., 1998).The inconsistency may stem from the differences in experimental procedures across studies.This study tasked participants with passively viewing video gameplay, which lacked the proactive interaction with the game and other players necessary for activating reward-related neural regions during gameplay (Cole et al., 2012;  Katsyri et al., 2013).Additionally, past research on the effect of action-like video gaming experiences on emotional perception has yielded mixed findings (Pichon et al., 2021;Yan et al., 2021).The dynamic perspective provided by whole-brain CAPs in this study further highlights the complexity of the effect of action-like video games on emotional perception and reward processing.Nevertheless, this study demonstrated that viewing action-like video games (FPS and ARSG) and LSG induced distinct dynamic features across extensive brain regions, including both primary and higher-order networks.

The distinct dynamic structures between first-person shooter games and action real-time strategy games
Compared to FPS viewing, ARSG viewing elicited a significant increase in the Fraction of Time and Counts of the attentional controlrelated CAPs ([FPN & VAN] -VN).Conversely, compared to ARSG viewing, FPS viewing elicited a significant increase in the Fraction of Time of the sensorimotor-related CAP ([SMN & VAN] -DMN), when gaming experience was considered as a covariate.These findings suggest that ARSG may engage more attentional resources, whereas FPS may place a greater emphasis on sensorimotor processes.Moreover, compared to FPS viewing, ARSG viewing elicited a heightened Transition Probability from the CAP associated with primary sensorimotor processing to the CAP associated with attentional control.This finding suggests that ARSG may prompt a more pronounced shift from fundamental sensorimotor functions to attentional control.These results highlight differences in the interaction dynamics between primary and higher-order networks, consistent with earlier research findings (Cui et al., 2021).
This study revealed distinct dynamic structures induced by the viewing of different gameplays, thereby improving our knowledge of the underlying mechanisms through which different video games influence brain development.However, several limitations should be acknowledged.First, although video gaming experience was included as a covariate in this study, its potential confounding effect cannot be eliminated due to variations in participants' video gaming experience, including differences in genre preference, experience level, and expertise level.Future research could benefit from recruiting a more homogeneous participant population, possibly by exclusively targeting video gaming amateurs.Second, in line with prior research (Sun et al., 2023;Yang et al., 2021Yang et al., , 2022)), our approach involved conducting k-means cluster analysis solely on the resting state data and subsequently applying the resulting CAPs to the gaming video sessions.Although the spatial similarity of CAPs between resting and each gaming viewing (Fig. S1) exceeded 0.9, suggesting the validity of this approach, it is important to note that the spatial pattern of the gaming state and the resting state may differ.Thus, future research should explore alternative approaches, such as clustering all data (including gaming and resting states) using k-means cluster analysis, to provide further insights into the differences between these clustering methods.
Given the diverse behavioral and neural mechanisms associated with action-like video gaming experience documented in previous studies, the gaming viewing used in this study may induce complex behavioral and neural dynamics related to visual working memory, attention, somatosensory, and emotion, thereby posing potential challenges to data analysis (Bavelier et al., 2012;Gong et al., 2016Gong et al., , 2019a;;Kowalczyk et al., 2018;Momi et al., 2018;Pichon et al., 2021;Tanaka et al., 2013;Yan et al., 2021).Thus, this study used association maps of Neurosynth topics to elucidate the relevant cognitive profile for each CAP (Coronel-Oliveros et al., 2023;Ju, 2022;van der Meer et al., 2020;Yarkoni et al., 2011).However, it is important to note the lack of behavioral performance data in this study prevented us from verifying these cognitive profiles.In addition, since this study aimed to elucidate the cortical dynamic distinctions elicited by different video gaming genres from a global perspective, it did not focus on the fine coding of the temporal information regarding gaming videos (e.g., emotional states during gameplay) due to the inter-individual differences in the experimental content and events.Future research should explore the associations among gaming event coding, neural dynamics, and behavior in various gaming genres.

Conclusions
This study explored the distinctive dynamic responses elicited by various video gaming genres, including life simulation games (LSG), first-person shooter games (FPS), and action real-time strategy games (ARSG), using a seed-free Co-Activation Pattern (CAP) method based on Regions of Interest (ROIs).The findings showed extensive, significant dynamic variations across both primary and higher-order networks between action-like video games (FPS and ARSG) and LSG.Notably, FPS and ARSG viewing induced dynamic disparities in sensorimotor-related CAPs, attentional control-related CAPs, and Transition Probability from sensorimotor-related CAPs to attentional control-related CAPs.
Fig. 2. The spatial distribution of eight Co-Activation Patterns (CAPs).The average time series of each Region of Interest (ROI) represents its baseline activation (Z = 0), as the time series is Z-normalized.Absolute Z-values show the degree to which activation (depicted in red) or deactivation (depicted in blue) deviates from baseline.

Fig. 3 .
Fig. 3.The correlations between the spatial distribution of Co-Activation Patterns (CAPs) and the association maps of 56 Neurosynth topics.The horizontal coordinates of the histograms indicate the correlation coefficients.Note: *** p < 0.001, ** p < 0.01, * p < 0.05 using a spin-based spatial permutation test.The word cloud illustrates correlation coefficients, where larger font sizes indicate stronger positive or negative correlations.Red words indicate positive correlations, while green words indicate negative correlations.

Fig. 4 .
Fig. 4. The differences in the Fraction of Time, Persistence, and Counts among three gaming viewing.Positive values indicate increases in dynamic features compared to resting, while negative values indicate decreases.The short green lines highlight the position of the zeros.LSG, life simulation game; FPS, first-person shooter game; ARSG, action real-time strategy game.Note: *** p < 0.001, ** p < 0.01, * p < 0.05 after a correction with false discovery rate (FDR).

Fig. 5 .
Fig. 5.The differences in the Transition Probability among three gaming viewing.LSG, life simulation game; FPS, first-person shooter game; ARSG, action real-time strategy game.The upper section of Fig. 5 presents the differences between the Transition Probability of each gaming viewing and that during resting.Positive values indicate increases in Transition Probability compared to resting, while negative values represent decreases.The lower section of Fig. 5 presents the Transition Probability that existed significant differences between different gaming viewing.The red arrow indicates the former had a larger Transition Probability than the latter, while the blue arrow indicates the opposite.Note: *** p < 0.001, ** p < 0.01, * p < 0.05 after a correction with false discovery rate (FDR).

Table 1
The corresponding functional networks and cognitions of the Co-Activation Pattern (CAPs).
Note: Red words indicate positive correlations between the spatial distribution of Co-Activation Patterns (CAPs) and the association maps of Neurosynth topics, while green words indicate the opposite.Key network abbreviations include VN (Visual Network), SMN (Somatomotor Network), DAN (Dorsal Attention Network), VAN (Ventral Attention Network), FPN (Fronto-parietal Network), DMN (Default Mode Network).