Interaction effects of Action Real-Time Strategy Game experience and trait anxiety on brain functions measured via EEG rhythm

Abstract

Purpose

Action Real-Time Strategy Games (ART SGs) is a prominent component of lifestyles and offers an ecological venue to investigate brain functions. Anxiety is an adaptive emotion and may exert its wide influence on brain functions. However, little is known about the potential interaction effects of trait anxiety and ART SG exper ence on brain functions.

Materials and methods

In this study, EEG and behavioral data of 215 League of Legends (LOL) video game players were collected. And then, two-way analysis of variance (ANOVA) were conducted on relative power indices, and relations between power indices and game performance scores were calculated using Spearman's correlation.

Results

Results showed that (a) gamma (resting and game play states) and beta (game play state) powers of high-level ART SG players were greater than those of low-level players in the high-Trait Anxiety Inventory (TAI) score group (p<0.05); (b) during the game play state, the delta power of highlevel players was lower than that of the low-level players in the high-TAI score group (p<0.05); and (c) game performance scores were positively correlated with gamma powers.

Conclusions

These findings indicated that brain functions may be influenced by interactions of the ART SG experience and trait anxiety levels, and provided new insights into the potential mechanisms of brainapparatus conversation.

Keywords:

• Action Real-Time Strategy Game
• trait anxiety
• interaction effects
• EEG rhythm
• game play state

1. Introduction

Video games, especially Action Real-Time Strategy Game (ARTSGs), have become one of the most prominent components of lifestyles and can provide both physical and spiritual experiences that are similar to true lifestyles. Furthermore, playing Action Real-Time Strategy Game requires players to focus on solving different problems at a far greater pace than what is typical in daily life [1,2]. Obviously, video gaming is interactively conducted with various information from apparatuses outside to the brain, and the brain vividly makes the decision for the next action. As a result, ARTSGs may offer an important and ecological venue to examine relationships between active learning and brain function [2–4], and further reveal the potential mechanisms of brain-apparatus conversation [5]. A number of behavioral findings in cognitive science have demonstrated that ARTSGs may enhance various cognitive abilities [2,3], including visual attention [6,7], sensory integration and working memory [3]. There is additional evidence of the cognitive benefits of ARTSG experience. For example, using resting-state functional magnetic resonance imaging (fMRI) and structural MRI, increased functional connectivity and gray matter volume in insular subregions were found [8]. Gong et al. also found that high-level players have enhanced functional integration between salience and central executive networks, which are related to visual attention and working memory, compared to low-level players [9]. These implied that ARTSG high-level players outperformed ARTSG low-level players in the speed of processing information during gameplay.

Anxiety is an adaptive emotion that exerts a wide influence on cognitive processing [10,11]. On the one hand, there is a processing efficiency theory; that is, anxiety may not impair quality of performance and may lead to the use of compensatory strategies and processing resources [12]. For example, using trait anxiety questionnaire scores, Edwards et al. found that working memory capacity may moderate the influence of trait anxiety on behavioral performance [11]. On the other hand, anxiety may exert a different influence on the cognitive function of different populations. For example, using fMRI, the interaction effect of trait anxiety and depressive symptoms on cognitive function in young and older adults was found, and the effects of trait anxiety on attention and working memory were greater in older adults [13]. Using behavior experiments, Alder et al. found that expert tennis players had better performance under high anxiety conditions [14]. However, considering it always contains motivation (eagerness to win) during playing ARTSGs, little is known about how trait anxiety affects brain functions such as attention and working memory for high-/low- level game players during ecological stimulation (e.g. ARTSGs, which offer rich and complex stimulation, and more closely approximate naturalistic lifestyles) as opposed to during traditional lab-based tasks.

Neural oscillations of cerebral cortex, which can be measured by EEG rhythm, underlie the dynamical interactions of multiple brain regions forming a functional network, and likely relate to a wide variety of cognitive processes necessary for perception and action [15,16]. For example, gamma-band EEG rhythm may relate to a broader range of processes, including feature integration, stimulus selection, attention, memory formation and even conscious awareness [17–19]. Beta-band EEG activity may be related to visual attention [20,21]. These cognitive processes may be critical to player attention during ARTSGs. However, little research has examined the potential interactions effects of ARTSG experience and trait anxiety on brain activity via EEG rhythms. This raises the question of how interactions of ARTSG experience and trait anxiety in brain functions affect the processing of more naturalistic stimuli (e.g. playing ARTSGs), and such research will help us to further understand the mechanism of brain activity measured via EEG rhythm during naturalistic stimuli.

In this study, given that the oscillatory frequencies during gameplay state may reflect different levels of cognitive processing, we investigated potential interactions between ARTSG experience and trait anxiety via EEG rhythms. Power spectrum indices during resting-state and game play state were calculated using time-frequency analysis, and then two-way analysis of variance (ANOVA) was used to examine the interactions of ARTSG level and trait anxiety. Behavioral data were also analyzed using two-way ANOVA, and relationships between EEG power indices and game scores were investigated using Spearman’s linear correlation.

2. Materials and methods

2.1. Participants

The League of Legends (LOL) video game used in this study is one of the most popular Action Real-Time Strategy Game among young people in China. A total of 215 participants with right-handed and a normal state of mind (males, mean age  =  21.42 ± 2.27 years, ranging from 18 to 28 years) were included in this study. All participants did not have a history of neurological disorders and relevant physical illnesses and were not currently taking medication known to affect their EEG data. This study was approved by the local ethical committee of the University of Science and Technology of China (UESTC) in accordance with the standards of the Declaration of Helsinki. Written informed consent were obtained from all participants.

Participants were recognized as ARTSG high-level players (∼top 25%) according to the expertise ranking provided by the LOL, which is a widely used method for calculating the relative skill levels of LOL players (The Elo rating of expertise ranking was based on an online inventory). In addition, ARTSG low-level players (lowest 29.92 ∼ 45.11%) were recognized as amateurs based on their rankings. There was no relationship between the Expertise Ranking distribution and age. The Trait Anxiety Inventory (TAI) questionnaire was used to assess trait anxiety, and the TAI score was used to generate subgroups with high (≥37) and low (<37) TAI scores [22]. A correlation analysis was conducted between TAI scores and Expertise Ranking, and there was no significant relationship between the two factors. For each subject, game scores (team performance) were first assessed independently by 6 experts (top 1%, have more than 3 years LOL experience) according to grade methods described in the Delphi method. Then, the game score of each subject was averaged across 6 experts (ranging from 0 to 5). Details of the demographic information of all the participants are shown in Table 1.

Table 1. Details of demographic information of all participants (means ± standard deviations).

Subgroup	Number	Game score	TAI score	Age (years)
Low-level, low-TAI score	55	3.19 ± 0.51	30.94 ± 4.23	21.65 ± 2.31
Low-level, high-TAI score	49	3.04 ± 0.53	43.16 ± 6.54	20.65 ± 2.15
High-level, low-TAI score	72	3.19 ± 0.55	30.14 ± 4.68	21.21 ± 2.18
High-level, high-TAI score	39	3.34 ± 0.55	44.56 ± 6.41	21.82 ± 2.25

2.2. Experiment

Each subject was asked to sit in a provided armchair for the duration of the experiment in a quiet and dimly lit room (temperature ∼25 °C). Each subject was asked to keep his eyes closed and stay fully relaxed for 5 min (resting state). After resting, participants were given an additional period of game readiness of approximately 5 min. Finally, the participants were asked to play a LOL game (game state) which lasted approximately 1 h. During the entirety of the experiment, EEG data with video was recorded.

2.3. EEG recording

The EEG data were recorded with an EEG system (BORUIEN, EEG32-BT, Chengdu, China) with 32 Ag-AgCI electrodes arranged in accordance with the international 10–20 system. The sampling rate was 1000 Hz. The impedance of all electrodes was kept below 5 kΩ during recording. All electrodes were recorded from the frontal vertex (i.e. FCz) as the reference, and AFz served as the ground electrode during recording. participants’ heads were fixed on the brackets to avoid head movement artifacts. To control eye movement artifacts, horizontal and vertical electrooculograms (EOGs) were recorded from electrodes above the right eye and at the outer canthus of the left eye, respectively. The electrode cap was placed on the participants’ heads during the entirety of the experiment to ensure that the recorded EEG data were optimal.

2.4. Data preprocessing and analysis

Data were analyzed using WeBrain pipelines (version 1.0, a cloud computing platform, https://webrain.uestc.edu.cn/) [23] and EEGLAB (version 14_1_0b, https://sccn.ucsd.edu/eeglab/index.php). In brief, EEG channels were visually inspected (Ruifang Cui and Xue Li) first, and “bad channels” were replaced with the extrapolated average values from the neighboring 3–4 channels. Then, the EEG data were divided into two segments according to the two states (resting state and gameplay state). The EEG data of each state were converted to average reference data and processed by a filter with a passband of 0.5–60 Hz and a notch filter (45–55 Hz). Next, the EEG data were re-referenced to the “zero” reference provided by the reference electrode standardization technique (REST) off-line [24,25]. Further, to reject EEG data with artifacts such as eye blinking, eye movement and head movement, the WeBrain pipeline of threshold method based on global field power (GFP) was used. GFP was calculated using the standard deviation of the signal at all channels: the threshold during the resting state was 25, and the threshold during the game play state was 30. Clean EEG data (the percentage of clean EEG data for each subgroup can be seen in Table 2) were used for further analysis.

Table 2. Percentages of clean EEG data for each subgroup (means ± standard deviations). There were no significant differences among the percentages of subgroups.

Subgroup	Resting state (%)	Gameplay state (%)
Low-level Low-TAI score	0.74 ± 0.25	0.52 ± 0.22
Low-level High-TAI score	0.74 ± 0.23	0.53 ± 0.25
High-level Low-TAI score	0.74 ± 0.23	0.53 ± 0.21
High-level High-TAI score	0.72 ± 0.27	0.58 ± 0.24

For each subject, EEG signals without artifacts were then divided into small epochs (5 s), and EEG data of each epoch, without overlap, were subjected to time-frequency analysis with Fast-Fourier Transform (FFT) to obtain the EEG power spectrum at each channel. Each epoch data unit was linearly detrended before time-frequency analysis. Finally, the mean indices across epochs (relative powers and power ratios) were obtained. The power indices were calculated in the delta (1-4 Hz), theta (4–8 Hz), alpha1 (8–10.5 Hz), alpha2 (10.5–12.5 Hz), beta1 (12.5–18.5 Hz), beta2 (18.5-21 Hz), beta3 (21.0–30 Hz), gamma1 (30-40 Hz) and gamma2 (40–60 Hz) frequency bands [26–29]. The relative power was calculated as the ratio of power in a given band/sum of power from 1 to 60 Hz (i.e. total power)×100%. The WeBrain pipeline of calculating power indices was used, and a brief description of these power spectrum indices is given in Table 3.

Table 3. Power indices extracted from EEG data.

Power indices	Description
Total power	Total power of the EEG power spectrum(1–60 Hz)
Relative delta	Absolute delta/Total power
Relative theta	Absolute theta /Total power
Relative alpha1	Absolute alpha1/Total power
Relative alpha2	Absolute alpha2/Total power
Relative beta1	Absolute beta1/Total power
Relative beta2	Absolute beta2/Total power
Relative beta3	Absolute beta3/Total power
Relative gamma1	Absolute gamma1/Total power
Relative gamma2	Absolute gamma2/Total power
R1	theta/(alpha1 + alpha2 + beta1)
R2	(theta + delta)/(alpha1 + alpha2 + beta1 + beta2)
R3	theta/(alpha1 + alpha2)
R4	theta/(beta1 + beta2 + beta3)
R5	delta/theta
PAF	The maximum power in the alpha(8-12.5) frequency band

2.5. Statistical analysis

The effects of game level (expertise ranking) and anxiety level (TAI score) on power indices were analyzed using a 2 (high-TAI score vs. low-TAI score) ×2 (high-level player vs. low-level player) two-way analysis of variance (ANOVA). A post-hoc t-test analysis was further conducted on the power indices. Noting that, a Jarque-Bera test was used to judge the distribution of the EEG power spectrum indices in each group, and a logarithmic transformation was used to translate the data distribution into a normal distribution before statistical analysis. Furthermore, the effects of game level and anxiety level on game scores were also analyzed using a two-way ANOVA and post-hoc t-test analysis. In addition, Spearman’s linear correlation analysis was conducted between game score and power spectrum indices in the 2 subgroups (high-TAI score and low-TAI score).

3. Results

The results of two-way ANOVA showed that there were significant (p < 0.05) interaction effects between game level and anxiety level on the EEG relative power spectrum. The main effects of game and anxiety levels were not significant. The EEG power spectrum indices with significant interaction effects are shown in Figure 2 and Figure 3. During the resting state, significant interaction effects in beta2 relative power (F_peak = 9.81, p = 0.0021, Fp1, F3, C3, C4, P3, P4, T5, T6, Fz, Fc4, Cp3, Ft8, Tp7, FCz, and CPz), beta3 relative power (F_peak = 4.75, p = 0.031, P3, T5, Cp3, and Tp7), gamma1 relative power (F_peak = 5.50, p = 0.020, Fp1, F3, C3, P3, P4, T5, Fz, Pz, Fc3, and Cp3), and gamma2 relative power (F_peak = 5.78, p = 0.017, Fp1, F3, C3, P3, P4, T5, Fz, Cz, Pz, Fc3, Cp3, Ft7, FCz, and CPz) were found. During game play state, significant interaction effects in delta relative power (F_peak =  14.91, p = 0.0001, Fp1, Fp2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5, T6, Fz, Cz, Pz, Fc3, Fc4, Cp3, Cp4, Ft8, Tp7, Tp8, FCz, CPz, and OZ), gamma1 relative power (F_peak = 10.54, p = 0.0014, Fp1, Fp2, F3, F4, C3, C4, F7, F8, T3, T4, T5, T6, Fz, Cz, Fc3, Fc4, Cp3, Cp4, Ft7, Ft8, Tp7, Tp8, FCz, OZ, and CPz), and gamma2 relative power (F_peak =  7.87, p = 0.0056, Fp1, Fp2, F3, F4, C3, C4, P3, P4, O2, F7, F8, T3, T4, T5, T6, Fz, Cz, Pz, Fc3, Fc4, Cp3, Cp4, Ft7, Ft8, Tp7, Tp8, FCz, CPz, and OZ) were found. Figure 1(C,D) demonstrated that, with high TAI scores, participants in the high-level group showed higher relative gamma1 (T_peak = 3.00, p = 0.0032), gamma2 (T_peak = 2.68, p = 0.0081), beta2 (T_peak = 2.28, p = 0.024) and beta3 (T_peak = 2.17, p = 0.031) power than did the participants in the low-level groups. Furthermore, with low TAI scores, participants in high-level group showed lower relative power than did those in the low-level group (gamma1 (T_peak = −2.66, p = 0.0085), gamma2 (T_peak = −2.55, p = 0.012), beta2 (T_peak = −2.76, p = 0.0065) and beta3 (T_peak = −2.26, p = 0.025)). Figure 2(C,D) demonstrated that, with high TAI scores, participants in high-level group showed higher relative gamma1 (T_peak = 3.26, p = 0.0014), gamma2 (T_peak = 2.80, p = 0.0057), and lower relative delta power (T_peak = −3.91, p = 0.00013) than did participants in the low-level groups, while wi th low TAI scores, gamma1 (T_peak = −2.71, p = 0.0073) and gamma2 (T_peak = −2.37, p = 0.0188) values were lower and delta (T_peak = 3.29, p = 0.0012) values were higher in high-level groups.

Figure 1. Experiment of EEG recording during ARTSG gameplay. After resting-state, each subject was asked to play League of Legends game (game state, about 1 h). The subject’s head was fixed on the brackets to avoid head movement artifacts. During the entirety of the experiment, EEG data with video was recorded.

Figure 2. Results of two-way ANOVA for relative power indices during resting state: (A) p values of interaction effects (p < 0.05); (B) F values of interaction effects; (C) interactions between ARTSG and anxiety levels were revealed by posthoc t-test analyses corresponding to the peak F value; and (D) differences (no log transformed) between high-level players and low-level players in high- and low-TAI subgroups (high level – low level).

There were significant (F = 4.04, p = 0.046) interaction effects between game level and anxiety level on game scores (Figure 4). And with high TAI scores, participants in high-level group showed higher game score (p = 0.017, T = 2.44) than did participants in the low-level groups. Significant correlations between game scores and power spectrum indices were found in subgroups. In the high-TAI score group, gamma1 relative power (p = 0.00079, R_peak = 0.37) and gamma2 relative power (p = 0.00080, R_peak = 0.37) were positively correlated with game score, while no statistically significant correlations were found in low-TAI score group.

4. Discussion

In this study, various EEG power spectrum indices (relative powers and power ratios) of ARTSG players were investigated using time-frequency analysis. The results revealed that the (a) gamma (resting state and gameplay state) and beta (gameplay state) relative powers of the high-level players were significantly higher than those of low-level players in the high-TAI score group; (b) during the gameplay state, the delta relative power of the high-level players was lower than that of the low-level players in the high-TAI score group; and (c) in the high-TAI score group, game scores of high-level players were higher than those of the low-level players, and game scores were positively correlated with gamma relative power. These findings suggested that the interactions between ARTSG experience and trait anxiety level might have potential effects on brain activity via EEG rhythms.

Figures 2 and 3 showed that high-level participants had a greater gamma relative power (located at the frontal and parietal channels) than did the low-level participants in the high-TAI score groups. In previous studies, it has been found that gamma rhythm may be closely, and positively related to many cognitive processes such as selective attention, working memory and decision making [17–19]. Using behavioral experiments, it has been demonstrated that Action Real-Time Strategy Game may improve visual attention abilities [2,6,7] and working memory [3,30]. In neuroimaging studies, it has been found that the brain functional network related to attention and working memory may be enhanced by ARTSG experience. For example, using fMRI, Gong et al. found that high-level players have increased functional connectivity between the attentional and sensorimotor networks [8] and enhanced functional integration between salience and central executive networks related to visual attention and working memory compared to low-level players [9]. Using event-related potential (ERP) and fMRI, Focker et al. found that enhanced perceptual and attentional control functions in AVGPs appear linked to parietal lobes [31,32]. These findings implied that changes in gamma rhythms in brain functions may be affected by ARTSG experience. On the other hand, previous studies have found that brain functions such as visual attention [33] and working memory [34] abilities may be affected by anxiety, which is related to EEG gamma rhythms [35,36]. Preliminary studies have suggested that working memory capacity [11] and cognitive load [37] may moderately influence performance in the presence of trait anxiety. For example, Owens et al. found that trait anxiety had a differential relationship with cognitive test performance depending on available working memory resources [38]. Using behavior experiments, Alder et al. found that expert tennis players had a better performance under high anxiety conditions [14]. In our work, our results were consistent with these studies, and perhaps implied that higher gamma rhythm in high-level ARTSG players with anxiety was likely to be explained by an increased motivation to do well during ARTSG playing, which may be driven by anxiety.

Figure 3. Results of two-way ANOVA for relative power indices during gameplay state: (A) p values of interaction effects (p < 0.05); (B) F values of interaction effects; (C) interactions between ARTSG level and anxiety level were revealed by posthoc t-test analyses corresponding to the peak F value (p < 0.05); and (D) differences (no log transformed) between high-level players and low-level players in high- and low-TAI subgroups (high level – low level).

Furthermore, during the resting state, the beta2 relative power and beta3 relative power (located at the temporal and parietal channels) of high-level players were significantly higher than those of the low-level players in the high-TAI score groups (Figure 3). It has been found that beta-band EEG activity may be related to visual attention [20,21]. In our work, increased beta power in high-level players with high TAI scores provided further evidence that trait anxiety may have positive effects on visual attention in high-level game players. On the one hand, visual attention may be improved by ARTSG experience [3,39]. On the other hand, it has been reported that anxiety may impair performance of ‘difficult’ tasks that require high attention [40], which may lead to decreased performance of low-level players in the high-trait anxiety group. During the gameplay state, high-level participants had a lower delta relative power (located at the parietal and frontal channels) than did the low-level participants in the high-TAI score groups. Delta is closely related to irrelevant information filtering [41] and working memory [42]. It was found that delta-beta coherence has been linked to emotional state, and increasing delta rhythm may be related to behavioral inhibition [42,43]. Our results were in line with these previous studies and implied that decreased delta power may be inversely related to enhanced brain function and better performance [42,44].

Figure 4 shows that in the high-TAI score group, the game scores of the high-level players were greater than those of the low-level players. The scale score, team performance of ARTSG, assessed by the Delphi method could reflect players’ comprehensive cognitive abilities [45], such as spatial cognition, visual attention, speed of processing inference, working memory and other cognitive abilities [3,46]. It showed that high-level ARTSG players with high anxiety have better performance of ARTSG, and it might be likely to be explained by an increased motivation to do well during ARTSG playing, which may be driven by anxiety. Meanwhile, game scores in the high-TAI score group were positively correlated with gamma relative power, which may be related to selective attention and working memory. This finding further indicates that ARTSG experience and trait anxiety level may have positive effects on brain functions in high-level players via EEG beta and gamma rhythm.

Figure 4. Results of behavioral data. A: interactions between ARTSG and anxiety levels on game scores were revealed by a post-hoc t-test analysis (p < 0.05); B and C: Spearman’s correlations between game scores and gamma relative powers in the high-TAI score group.

However, several limitations should be noted in this study. First, the current study is a preliminary and exploratory study to reveal the interaction effect of ARTSG experience and trait anxiety. The statistical results were not corrected for multiple comparisons. Meanwhile, under the current experimental design and conditions, it was difficult (or might be not appropriate now) to locate and analyze the specific events during the game process in more detail. ARTSGs offer rich and complex stimulation, which closely approximate naturalistic lifestyles, and may have comprehensive effects on brain functions. Therefore, a series of follow-up studies will be expected to replicate current findings and improve interpretations. Second, in the process of the experiment, we controlled for the interference of many factors, including age, gender and education experience. However, there may be other factors (e.g. depression) that can affect the ARTSG player’s brain function and game performance. Third, it should be noted that the Delphi method has limitations of subjectivity. Thus, an increased number of experts and objective indicators will be expected to make up for this limitation. Finally, dichotomizing was used in this work, due to its advantages of simplifying the statistical analysis and presentation of results. However, considering the potential risk of dichotomizing (e.g. losing useful information or underestimating the extent of variation between two groups) [47], Spearman’s correlations between power indices and TAI scores in high- or low-level players were further calculated to compensate for these issues. Results showed significant correlations between power indices (delta, gamma1 and beta2 bands) and TAI scores in high-level players (see supplementary materials Figure S1–S7), and it further supported our findings.

5. Conclusions

In summary, our results revealed that trait anxiety level and ARTSG experience may have a significant interaction effect on brain activity via EEG gamma and beta rhythms and provide new insights into the potential mechanism enabling high-level ARTSG players to outperform low-level players, and how trait anxiety affects brain functions during naturalistic lifestyle activities. And, it may be useful to guide the design of game training under reasonable anxiety conditions, or as candidate features in brain-computer interfaces [48,49] in the future.

Author contributions

Conceived and designed the work: LD. Wrote code: LD and XL. Acquired the data: LQ, RC, XL and DG. Analyzed the data: CL, RY and XL. Wrote and revised the manuscript: LD, XL, LQ and DG. Equally contributing authors: LQ and XL. All authors revised the work for important intellectual content. All of the authors have read and approved the manuscript.

Disclosure statement

The authors report no conflict of interest.

Data availability statement

The EEG data used to support the findings of this study are not publicly available due to the unfinished study of the whole project but are available from the corresponding author upon reasonable request.

Additional information

Funding

This work is supported by the MOST 2030 Brain Project [Grant No. 2022ZD0208500], and the Sichuan Science and Technology Program [2022NSFSC1398, 2021YJ0167].