Effects of video game immersion and task interference on cognitive performance: a study on immediate and delayed recall and recognition accuracy

Participants

We recruited 160 participants aged between 18 and 29 years, all of whom had at least 3 years of regular experience playing “shoot-em-up” video games. Participants were recruited from a university setting (67% male, 33% female) across various academic majors, resulting in a diverse sample with different cognitive and gaming backgrounds. Participants were randomly assigned to one of three experimental groups or a control group using a computer-generated randomization algorithm. This process was designed to ensure that each group was balanced in terms of age, educational background, and gaming experience. The randomization was stratified to maintain an even distribution of gender across all groups. This approach ensured that any differences observed in cognitive performance could be attributed to the experimental conditions rather than demographic imbalances. Recruitment was conducted on a voluntary basis, with participants fully informed and providing consent prior to their involvement in the study. Participants were primarily recruited through targeted announcements in psychology courses and by word of mouth among students attending various courses (Economics, Law, Educational Sciences, Engineering, Classical Studies, Sports Science, Social Work). We opted for presentations at the beginning or end of lectures to explain the study’s purpose, its relevance to understanding cognitive processes in gaming, and how students could volunteer. Additionally, students already enrolled were encouraged to inform their peers attending other majors about the study, leveraging peer networks to increase participant numbers. No incentives were provided for participation.

Design and procedure

The study employed a between-subjects design to examine the effects of video game immersion and task interference on cognitive performance. Each experimental condition was designed to vary the type of cognitive interference experienced by participants while engaging in a secondary task of mnestic recall and linguistic recognition.

Criteria for game selection

The game genre was restricted to first-person shooters (FPS), a genre well-known for its demand on attention, speed, and multitasking abilities. These games are highly popular among the age group of our participants (18–29 years), which ensures that the majority of participants are already familiar with the game mechanics, reducing the learning curve and focusing on the cognitive load aspect. The study utilized “Call of Duty: Modern Warfare”, a first-person shooter (FPS) game, chosen for its high cognitive demands, popularity among the target demographic, and ability to impose a significant cognitive load. This selection was based on pre-testing and literature review, ensuring the game’s suitability for examining the effects of task interference on cognitive performance. The game’s difficulty level was configured to be medium-high to ensure it was sufficiently challenging to affect cognitive performance without being overly frustrating for the participants.

Implementation

During the study, the game sessions were standardized across all participants to control for any variability in game difficulty and session duration. Participants played the game under monitored conditions to ensure consistent exposure and to accurately measure the game’s impact on subsequent cognitive task performance.

The Rey Auditory Verbal Learning Test (RAVLT) (Rey, 1958; Carlesimo et al., 1996) was chosen as the primary tool for assessing immediate and delayed recall as well as recognition memory. The test is made up of seven tasks. In the first task the examiner reads the words and asks the subject to repeat them. After the first task, the words are read another four times and the subject is asked to repeat them, including the previously recited words. After these five tasks, the subject under examination is “distracted” with visual-spatial tasks for about 15 min, and then he is asked to repeat the same words one more time. The subject under examination then undergoes another recognition task whereby he is presented verbally with 46 words, that is to say, the 15 words presented to him in the first task together with 31 distractors. They are asked to identify the words they remember from the original list. The test serves as a classic quantitative evaluation for immediate and deferred recall, but also for an interesting qualitative evaluation on how the subject has coped with the memorization test.

The subjects were randomly assigned to one of three experimental groups. The first group completed a 15-min gaming session in a familiar environment with a medium-high level of technical difficulty during the interval in the administration of the Rey test, i.e., between the first and the second session. The subjects were asked to focus on the video game task and simultaneously follow the verbal instructions of the experimenter before and after the gaming session. The second group was invited to play a video game while simultaneously responding to the verbal prompts of the experimenter who administered the first session of the Rey test (immediate recall of a list of fifteen words in five repeated phases). At the end of the fifth repetition, there was a 15-min interval, as prescribed by the test, and then we proceeded to complete the second part of the test that measured the deferred recognition of the preceding list from a new list of 46 words. The experimenters recorded the number of correct recalls and also noted the number and type of errors made during the recall and recognition sessions. The third group was first invited to play for 15 min and then to undertake the recall and recognition test sessions. Overall, an experimental structure was established in which the first group experienced an automated immersion in the video game activity between one test and the other; the second group operated under conditions of maximum interference and simultaneity between the automated video game activity and the mnestic recall; the third group performed the mnestic task immediately after the automated immersion. In addition to the experimental groups, we used another control group consisting of 40 subjects from the same age range (18–29) to whom we administered the Rey test without any experimental game conditions. All experimental sessions were conducted in soundproof rooms with standardized lighting conditions to minimize external distractions and ensure consistency across sessions. Experimenters were blinded to the group assignments to prevent bias in data collection and analysis.

Summary of experimental conditions

Group 1) Participants performed the RAVLT, followed by a 15-min video game session, and then completed the recall and recognition tasks.

Group 2) Participants engaged in video gaming simultaneously while trying to recall the words from the List, immediately followed by the recognition task.

Group 3) Participants played video games for 15 min before starting the RAVLT.

Control Group: Participants only completed the RAVLT without any interference from video gaming.

Confidentiality regarding the data collected was strictly maintained, with measures іn place tо protect the privacy and anonymity оf all participants. The study was conducted in accordance with the requirements of the Helsinki convention and approved by the local Institutional Review Board of the University of Cassino and Southern Lazio (IRB_DIPSUS 07:22/04/23). Informed written consent was obtained from all participants.

Data analysis

The performance on the RAVLT was quantified by the number of words correctly recalled during the immediate and delayed recall tasks, as well as the number of correctly recognized words in the recognition task. Differences between groups were analyzed using ANOVA to assess the impact of task interference on recall and recognition. Error types (intrusions and repetitions) were also recorded and analyzed to understand the patterns of errors under different multitasking conditions. In addition to standard ANOVA, interaction effects between task types and group assignments were analyzed to discern more nuanced impacts of task sequencing and multitasking on cognitive outcomes. These analyses help in understanding whether the sequence in which tasks are performed (i.e., before or during video game play) exacerbates or mitigates the cognitive load, potentially influencing the recall and recognition performance differently.

The level of statistical significance was set at p < 0.05. In order to ensure the robustness of our findings, a power analysis with G*Power (version 3.1.9.7) was conducted for the primary statistical tests (ANOVA) used in this study. Given the sample size of 160 participants distributed across four groups, the power analysis indicated that the study had a power of 0.95 or higher to detect medium to large effect sizes (Cohen’s f > 0.25) at an alpha level of 0.05. This high level of statistical power suggests that the study is well-equipped to identify significant differences between groups, minimizing the risk of Type II errors. To address potential overfitting, we employed k-fold cross-validation techniques, where the dataset was divided into k subsets, and the model was trained and validated k times, each time using a different subset as the validation set and the remaining subsets as the training set. This approach helps ensure that the model’s performance is more robust and generalizable, reducing the likelihood that the model is simply fitting noise in the training data. Missing data were handled using multiple imputation for cases where less than 5% of data was missing. For cases with more extensive missing data, those records were excluded from the analysis to prevent bias. Outliers were identified using z-scores and were handled by winsorizing the data at the 5th and 95th percentiles. This approach allowed us to reduce the influence of extreme values while retaining all data points for analysis. The impact of outliers on the overall results was assessed, and sensitivity analyses were conducted to ensure the robustness of the findings. Data analyses were performed using the IBM SPSS Statistics 26.

Immediate recall

Analysis of the fifth session of immediate recall demonstrated significant variation among the groups: F(3,156) = 64.65, p < 0.001 η² = 0.554 Cohen’s f = 1.115, see Fig. 1 below. The graph shows how the groups performed in the last session of immediate recall. Group 1 and Group 4 perform better compared to Group 2 and Group 3, which indicates less cognitive interference from multitasking or better task management. Group 1, which engaged in a video game session between the recall tasks, showed the highest recall rates, significantly outperforming the other groups, especially Group 2, which engaged in simultaneous video gaming and recall tasks and showed the lowest performance.

This box plot graph shows the average scores for immediate recall across the groups, illustrating the impact of different multitasking conditions.

The post-hoc analysis reveals significant differences in performance across groups during the fifth session of immediate recall. Group 2, engaging in simultaneous video gaming and recall tasks, exhibited the lowest performance, underscoring the adverse effects of high cognitive load and multitasking (see Table 2).

Delayed recall

The delayed recall results also indicated significant differences among the groups: F(3,156) = 11.74, p < 0.001 η² = 0.184 Cohen’s f = 0.475; see Fig. 2 below. Here again, Group 1 generally outperformed the other groups, showing the efficacy of structured task interference over simultaneous task execution. The trends are similar, with Group 1 showing slightly better retention compared to other groups.

This graph compares the groups in terms of their performance on the delayed recall task, highlighting differences in memory retention under varying task interferences.

Post-hoc analysis showed significant differences in the delayed recall performance. The results suggest that while the effect of task interference was apparent, it was most detrimental in Group 2 (See Table 3).

Recognition task

The recognition task highlighted significant disparities: F(3,156) = 35.20, p < 0.001 η² = 0.404 Cohen’s f = 0.823 (See Fig. 3 below). Group 2 exhibited notably more false recognitions compared to the other groups. This suggests a higher cognitive load and interference in managing task requirements under simultaneous multitasking conditions.

This graph displays the average number of false recognitions by each group, emphasizing the increased error rates under simultaneous multitasking conditions.

The Post-hoc analysis of false recognitions highlighted further differences. Findings indicate a significantly higher rate of false recognitions in Group 2, suggesting that simultaneous task execution led to the most pronounced errors (See Table 4).

Error analysis

Intrusion errors, where participants mistakenly recall words not presented in the list, were higher (but not significant) in Group 2, which engaged in simultaneous video gaming and recall tasks: F(3,156) = 0.62; p = 0.602 η² = 0.012 Cohen’s f = 0.109 (see Fig. 4 below). This suggests that the intervention involving immersion in video games and task interference did not have a significant impact on intrusion errors in this specific experimental condition.

In regard to repetition errors, the ANOVA analysis on the total repetitions across different groups reveals significant differences: F(3,156) = 7.42, p < 0.001 η² = 0.125 Cohen’s f = 0.378 (See Fig. 5 below). This indicates that the number of repetition errors made by participants varied significantly across the experimental groups.

Post-hoc results indicate that Group 4 consistently had fewer repetitions compared to other groups, suggesting better task management or lower cognitive interference in their experimental conditions. This insight might be useful for further investigating how task structure or timing influences cognitive load and error frequency in multitasking environments (See Table 5).

Interpretation

1) The results from the fifth session of immediate recall demonstrate pronounced variability among the groups. Group 2, which was engaged in simultaneous video gaming and recall tasks, showed significantly poorer performance compared to all other groups. This underscores the deleterious impact of high cognitive load when tasks are performed simultaneously.

2) Although there were significant differences in performance across the groups, the effect of task interference appears less pronounced over time. This may suggest some recovery or adaptation to the multitasking demands.

3) Similar to immediate recall, Group 2 also displayed a significantly higher rate of false recognitions. This further highlights their vulnerability to error under conditions of simultaneous multitasking. Interestingly, despite the increased number of intrusion errors noted in Group 2, these did not reach statistical significance, suggesting that while the cognitive load was high, it did not significantly alter the pattern of intrusion errors compared to other groups.

4) The significant mean difference in repetition errors between Group 4 and the elucidates the impact of task structure and timing on cognitive load. The results suggest that while task interference invariably affects cognitive performance, the structuring and sequencing of such tasks play pivotal roles in mediating these effects.

Model’s performance

The model’s performance was evaluated using AUC (Area Under the Curve) as the primary metric, with comparisons made between the training and test datasets. To address potential overfitting and ensure the model’s robustness, we employed a 10-fold cross-validation technique. This approach involved dividing the dataset into 10 subsets, training the model on nine subsets while validating it on the remaining subset, and repeating this process 10 times.

The cross-validation results showed consistent AUC values across the folds, with a mean AUC of 0.93 and a standard deviation of 0.03, indicating that the model’s performance was stable and less likely to be overfitting. These findings suggest that the model generalizes well across different subsets of the data, reducing the likelihood of performance being driven by noise in the training set.

Power analysis

In order to ensure the robustness of our findings, we conducted a Power Analysis for each of the ANOVA tests performed in this study. The Power Analysis was executed to estimate the likelihood that our study would detect a statistically significant effect, given the sample size and effect sizes observed. We assumed a common alpha level of 0.05 for all tests.

Immediate Recall: Effect Size (Cohen’s f): 1.115; Power: 100%

This indicates that our study has excellent power to detect the observed effects of video game immersion and task interference on immediate recall tasks.

Delayed Recall: Effect Size (Cohen’s f): 0.475; Power: 99.96%

The study also demonstrates high statistical power for detecting differences in delayed recall due to the experimental conditions. This nearly perfect power score ensures that any significant effects are unlikely to be missed.

Recognition Task: Effect Size (Cohen’s f): 0.823; Power: 100%

Similar to immediate recall, the study possesses perfect power for the recognition task, affirming the capability to detect significant cognitive impacts from multitasking and task immersion.

Repetition Errors: Effect Size (Cohen’s f): 0.378; Power: 98.6%

The analysis indicates a very high probability of detecting the effects of video game playing and task interference on the frequency of repetition errors.

These results underscore the adequacy of the study design and the reliability of the conclusions drawn from the data.

Our analysis explored also the influence of gender, years of gaming experience, and gaming preferences on the various cognitive performance metrics. Results indicate that gender significantly impacts the rate of false recognitions, with males and females showing different patterns in error processing (p < 0.01). Males exhibited an average false recognition rate of 0.94 ± 1.35, while females showed a higher rate of 1.42 ± 1.93. The statistical significance of this difference was confirmed with a p-value of less than 0.01 (p = 0.008). To quantify the difference between genders, we calculated the effect size, which was found to be d = 0.33, indicating a small effect according to Cohen’s standards. However, neither gender nor years of gaming experience significantly influenced the initial and delayed recall scores, suggesting a minimal impact of these variables on short-term memory tasks under the conditions tested.

Interestingly, gaming preferences, categorized as playing alone, with friends, or online, did not show a significant direct effect on recall abilities or error rates. This suggests that the social context of gaming, within the scope of our experimental design, does not alter cognitive performance in a statistically significant way.

Limitations and future directions

While this study contributes valuable insights into the impact of multitasking with digital media on memory tasks, it is not without limitations. One of the key limitations of our study is the potential for selection bias due to the recruitment of participants primarily from a university setting, which may limit the generalizability of the findings to broader populations. Also the reliance on self-reported gaming experience could introduce recall bias, as participants may have inaccurately reported their gaming habits or skill levels. Another limitation is the relatively homogeneous sample in terms of age and educational background, which may not fully capture the diversity of cognitive responses to video game immersion. As the sample was restricted to regular players of shoot-em-up video games, this may limit generalizability. Future research could explore a broader demographic to include non-gamers or occasional gamers to assess if the same patterns of cognitive interference occur. Exploring the long-term effects of such multitasking behaviors on cognitive functions could provide deeper insights into the chronic impacts of digital multitasking. The study’s outcomes may also be influenced by the technological limitations of the VR system used. Hardware constraints, such as the resolution of the VR headset and potential latency issues, could affect participants’ experiences and the fidelity of the tasks. Another consideration is the ecological validity of the VR tasks. While VR provides a controlled environment that can closely replicate real-world scenarios, there may still be differences between these simulated tasks and real-life activities. The extent to which the VR tasks used in this study mirror actual cognitive challenges faced in everyday life is a key factor in interpreting the results. Future studies should continue to refine VR scenarios to enhance their realism and applicability to real-world contexts.

Future research should focus on longitudinal studies to assess the long-term effects of video game immersion on cognitive functions. It would also be beneficial to replicate this study across diverse demographics, including different age groups and professional backgrounds, to evaluate the generalizability of these findings and to understand better the variability in cognitive responses to multitasking across different segments of the population.

While our study focused on the cognitive impacts of engaging with a high-demand non-VR game, future research could explore similar effects within a VR context using games specifically designed for VR platforms. Such studies would provide valuable insights into whether the immersive nature of VR gaming exacerbates or mitigates cognitive load and task interference compared to traditional gaming environments. Researchers should also consider conducting longitudinal studies to track changes in cognitive functions over time using VR assessments like VR-EAL. Such studies could provide valuable insights into how cognitive abilities evolve, particularly in response to interventions, and how these changes are reflected in immersive VR environments (Kourtesis & MacPherson, 2021). The VR-EAL platform also holds potential for broader clinical applications, particularly in populations with neurological disorders such as Alzheimer’s disease, traumatic brain injury, or stroke. Using VR-EAL in these contexts could enhance diagnostic accuracy and support tailored rehabilitation programs by providing detailed assessments of cognitive functions in environments that closely mimic real-life tasks. Integrating VR with other emerging technologies, such as artificial intelligence and neuroimaging, offers exciting possibilities for advancing neuropsychological assessments. For example, AI could be used to analyze complex data patterns within VR tasks, while neuroimaging techniques could provide complementary insights into the neural correlates of cognitive performance in these immersive environments. Such integrations could significantly enhance our understanding of cognitive functions and lead to more personalized and effective interventions.