Individual Variability in Cognitive Engagement and Performance Adaptation During Virtual Reality Interaction: A Comparative EEG Study of Autistic and Neurotypical Individuals

Abstract

Many studies have recognized that individual variability shapes user experience in virtual reality (VR), yet little is known about how these differences influence objective cognitive engagement and performance outcomes. This study investigates how cognitive factors (IQ, age) and technological familiarity (tech enthusiasm, tech fluency, first-time VR experience) influence EEG-derived cognitive responses (alpha and theta activity) and task performance (trial duration) during VR interactions. Sixteen autistic and sixteen neurotypical participants engaged with various VR interactions while their neural activity was recorded using a Muse S EEG. Correlational analyses showed distinct group-specific patterns: higher IQ correlated with elevated average alpha and theta power in autistic participants, while tech fluency significantly influenced performance outcomes only in neurotypical group. Prior VR experience correlated with better performance in the neurotypical group but slower adaptation in the autistic group. These results highlight the role of individual variability in shaping VR engagement and underscore the importance of personalized design approaches. This work provides foundational insights toward advancing inclusive, user-centered VR systems.

1. Introduction

Virtual reality (VR) has emerged as a promising tool to support individuals with autism spectrum disorder (ASD). VR’s unique strength lies in its capacity to deliver structured yet adaptable experiences, which provide controlled environments that can be tailored to specific user needs and preferences. Numerous studies have reported positive outcomes of using VR to improve social, cognitive, and daily living skills among autistic users [1,2]. However, an essential yet often underemphasized aspect of VR design and application is the considerable individual variability among autistic users.

Individual differences such as sensory preferences, cognitive abilities, and technological familiarity can significantly affect how comfortable, engaging, and ultimately effective a given VR interaction technique is for a user. Recognizing and accommodating this variability is crucial because autistic individuals often exhibit diverse sensory and cognitive profiles [3], benefiting more from personalization than from one-size-fits-all solutions. Many recent studies have acknowledged the importance of these individual differences in shaping VR experiences [4,5,6].

Although existing studies have explored the relationship between individual traits and cognitive or performance responses in VR, this work has largely been conducted in neurotypical populations, leaving a gap in our understanding of these dynamics in autistic individuals. According to a systematic review by Li et al., most existing VR interventions for autism primarily focus on social and affective skill training, with little attention paid to how individual differences might influence VR interaction effectiveness [7]. Karami et al. similarly reported varying levels of effectiveness of VR training across different skill domains for individuals with autism, suggesting that user-specific traits are likely to play a role in determining the outcomes of VR experience [8]. For instance, the strongest effect was observed for daily living skills, while more moderate effects were found for cognitive, emotional regulation, and social skills.

A particularly relevant study by Bozgeyikli et al. explored various VR interaction techniques among autistic users and identified signs that individual variability influences VR interaction preferences [9,10,11]. Their study highlights that techniques such as joystick-based navigation and “point and teleport” were typically favored, whereas methods involving flying or complex gestures were considered less comfortable [11], suggesting a likely connection between sensory–motor profiles and VR interaction preferences. Additional studies have also emphasized the role of cognitive factors. For instance, recent research on VR-based social skills training found that autistic adults with stronger planning skills, a key aspect of executive function, demonstrated greater improvements following VR interventions [8].

Taken together, we identify a research gap regarding the explicit investigation of how individual characteristics influence cognitive engagement and performance adaptation during VR interactions, particularly in autistic populations. To address this gap, this study explicitly examines the impact of individual variability including cognitive ability (IQ), age and technological experience on EEG-derived cognitive responses (alpha and theta activity) and performance (trial duration) during repeated VR interaction tasks. Guided by the research question “How do individual differences influence cognitive engagement and performance adaptation during repeated VR interactions among autistic and neurotypical individuals?”, this research aims to inform the development of more inclusive VR environments. In this paper, we address this question by first outlining the methods, then presenting the findings, followed by a discussion of their implications for VR design. By accounting for individual profiles, we hope to enhance the efficacy and accessibility of VR experiences for both autistic and neurotypical users.

2. Methods

The dataset analyzed in this study is part of a larger research project exploring EEG-based evaluations of VR interaction techniques for autistic and neurotypical individuals. To ensure that this article is self-contained, this section includes relevant methodological details, which are also described in a separate manuscript currently under review.

Ethical approval was granted by KEP UNPAD, with official permission granted from Bakesbangpol Kota Bandung. Sixteen autistic participants were recruited through local institutions and autism-focused organizations in Bandung, Indonesia. Due to limited formal diagnoses within these institutions, autism status and intellectual functioning were verified using the Wechsler–Bellevue Intelligence Scale (WBIS), administered individually by certified Indonesian psychologists. Sixteen neurotypical participants were recruited from the Indonesian community in Berlin, Germany. All participants or caregivers provided informed consent.

Participants engaged in four VR interaction techniques: two locomotion methods (teleportation and step) and two object selection/manipulation methods (hand-tracking and controller). Each technique involved at last five trials, preceded by a brief guided practice. The locomotion task required participants to follow directional cues, presented as a red arrow, along a virtual road (see Figure 1a), while the object interaction task involved placing 3D objects from a desk to a basket (see Figure 1b). The VR environment was intentionally minimalistic and visually neutral to reduce sensory distraction. The overall session lasted approximately 30 min, with short breaks provided between techniques to minimize fatigue.

The task was performed using Oculus Quest 2 VR headset (Meta Platforms, Inc., Menlo Park, CA, USA) and Muse S Gen 2 EEG headset (InteraXon Inc., Toronto, ON, Canada) with placement setup as seen in Figure 2. A baseline EEG recording (±2 min, eyes-open) was collected prior to the experimental tasks using the MALOKA VR application (version 1.4.3). The EEG data underwent notch filtering at 50 Hz, band-pass filtering (0.5–80 Hz), and z-score normalization. EEG segments were synchronized with VR task events via manual video annotation.

2.1. Variables of Individual Differences

Participants’ demographic details, including distribution of gender, age, and IQ scores, are summarized in

Table 1. Participants demography.

Variable	Autism Group	Neurotypical Group
Number of participants	16	16
Age	Range: 14–38	Range: 25–32
	Mean: 22, Median: 21	Mean: 28.25, Median: 28.5
IQ	Range: 36–107	Range: 98–139
	Mean: 69.5, Median: 69.5	Mean: 120.31, Median: 122
Gender distribution	Male: 14 (87.5%)	Male: 8 (50%)
	Female: 2 (12.5%)	Female: 8 (50%)

. In this study, we use the term ‘predictor’ to refer to variables hypothesized to correlate with or explain variability in cognitive engagement and performance responses during VR interactions. Cognitive profiles of autistic participants were assessed using the Wechsler–Bellevue Intelligence Scale (WBIS) [12]. The WBIS was utilized due to substantial barriers to access fully normed or licensed versions of more recent Wechsler editions (e.g., WAIS-IV or WAIS-5) in Indonesia. These barriers include licensing costs, the absence of nationally representative normative data, and the limited availability of certified psychologists trained in administering the newer editions. Thus, the WBIS remains the most practical and commonly used instrument for assessing cognitive functioning in the Indonesian context.

Due to the limited awareness of neurodevelopmental disabilities and early intervention services in Indonesia, many autistic participants received delayed diagnosis and therefore exhibited intellectual limitation, which resulted in a lower mean IQ compared to neurotypical group. Recognizing this inherent difference, we specifically included IQ as a predictor variable in our analyses. This explicit inclusion allowed us to transparently account for IQ differences when interpreting group differences in cognitive and performance outcomes.

Participant characteristics related to technology use (summarized in

Table 2. Overview of individual difference variables.

Metrics	Scale
Tech enthusiasm	7-point Likert scale (1 = not enthusiastic, 7 = extremely enthusiastic).
First VR experience	(0 = no, 1 = yes)
Tech fluency	Autism group Frequency of smartphone use (1–7) Familiarity with complex software applications (0/1) Number of electronic devices mastered (numeric count) Initiatives in learning new technologies (1–7)	Neurotypical group Tech background (job, major) (0/1) Gaming experience (0/1)

) were assessed through structured questionnaires and semi-structured interviews. For autistic participants, the information was provided by their caregiver, whereas neurotypical participants responded directly. Tech enthusiasm was measured to capture participants’ subjective interest and motivation toward technology, and the first VR experience was documented based on participants’ prior exposure to VR technology as reported in the interviews.

Tech fluency was operationalized differently for each group to reflect their distinct demographic and technological contexts. For both groups, individual indicators were first standardized using z-score normalization and then averaged to compute a composite technology fluency score. Group-specific measures were necessary due to differences in context. Using identical items across groups would likely have resulted in ceiling effects among neurotypical participants, thereby limiting variability and interpretability.

2.2. EEG and Performance Measurements

Alpha (8–13 Hz) and theta (4–8 Hz) frequency bands were explicitly selected due to their established association to cognitive processes. Alpha activity typically reflects attentional engagement, with decreased alpha power (alpha suppression) indicating sustained or increased cognitive focus [13,14,15]. Theta activity, meanwhile, is closely linked to cognitive control and working memory demands, and typically increases under conditions requiring greater mental effort [16,17,18]. These EEG measures have also shown different patterns in autistic populations [19,20,21], making them particularly relevant. Given this targeted relevance, other EEG bands, such as beta or gamma, were not analyzed to maintain a clearer focus on engagement and cognitive processes central to this study.

While performance in VR scenarios can be assessed through various metrics such as error rates, success rates, or task accuracy, this study focuses on trial duration as the primary indicator. Trial duration refers to the time spent completing a single trial, measured from the start of the task to its completion. In the context of this study, it serves as a practical and objective proxy for assessing task efficiency and user adaptation during the VR interaction.

To achieve a comprehensive understanding of cognitive profiles and performance responses, we analyzed both average and adaptation measures of EEG and performance. The average measure was determined by calculating the mean EEG power and trial durations across all trials for each participant. On the other hand, the adaptation measure captures variations in EEG activity and performance over repeated trials, measured as slopes derived from simple linear regression. These slopes indicate the direction and magnitude of adaptation, reflecting either increasing or decreasing cognitive engagement and performance improvement or decline with repeated task exposures.

2.3. Statistical Analysis

We specifically examined relationships between individual differences and cognitive performance adaptation during VR interactions. Spearman’s rank-order correlations were conducted to identify associations between demographic and technology-related variables (IQ, age, tech enthusiasm, tech fluency, first VR experience) and the outcome variables (average EEG alpha PSD, average EEG theta PSD, average trial durations, and adaptation slopes for EEG alpha PSD, EEG theta PSD, and trial durations). All statistical analyses were performed using Jamovi software (version 2.6.24.0). Due to the exploratory nature and modest sample size of the study, correlations with p-values between 0.05 and 0.10 are explicitly identified as exploratory trends rather than conclusive evidence.

3. Results

This section presents the results of the correlational analyses, examining the relationships between individual difference variables, cognitive engagement, and performance responses during repeated VR interaction tasks. The findings are organized into two main parts: (1) associations with average performance and neural activity, and (2) patterns of adaptation across repeated trials.

3.1. Average Performance and Neural Activity

To investigate participants’ cognitive state and task performance during VR interaction, we analyzed the relationships between participant characteristics and average EEG power (alpha and theta), as well as average trial duration. Significant correlations are summarized in

Table 3. Significant correlations between participant characteristics and average performance and EEG activity.

Measure	Predictor	Autism (ρ, p)	Autism—Interpretation	Neurotypical (ρ, p)	Neurotypical—Interpretation
Trial Duration	IQ	ρ = −0.484 (p = 0.057 †)	Higher IQ moderately predicts faster overall performance (marginal significance).	ρ = −0.473 (p = 0.064 †)	Higher IQ moderately predicts faster overall performance (marginal significance).
	Tech Fluency	ρ = −0.273 (p = 0.306)	n.s.	ρ = − 0.420 (p = 0.083 †)	Higher tech fluency moderately predicts faster overall performance (marginal significance)
Alpha	IQ	ρ = 0.721 (p = 0.002 *)	Higher IQ is significantly associated with higher alpha baseline.	ρ = −0.021 (p = 0.939)	n.s.
	Age	ρ = −0.015 (p = 0.956)	n.s.	ρ = 0.460 (p = 0.055 †)	Older participants had marginally higher alpha baseline.
Theta	IQ	ρ = 0.536 (p = 0.032 *)	Higher IQ is significantly associated with higher theta baseline.	ρ = 0.137 (p = 0.612)	n.s.
	Age	ρ = −0.263 (p = 0.325)	n.s.	ρ = 0.456 (p = 0.058 †)	Older participants had marginally higher theta baseline.

Note: ^† marginally significant (p < 0.1), * significant (p < 0.05), n.s. not statistically significant (p > 0.05).

, with the full correlation matrix provided in Appendix A.1.

For autistic participants, higher IQ correlated with elevated alpha and theta power (alpha: ρ = 0.721, p = 0.002; theta: ρ = 0.536, p = 0.032). In contrast, IQ showed no significant association with either EEG measure in the neurotypical group. Regarding task performance, IQ was marginally associated with faster average trial durations in both groups (autism: ρ = −0.484, p = 0.057; neurotypical: ρ = −0.473, p = 0.064).

Tech fluency further distinguished the groups; it showed weak-to-moderate evidence of faster task performance in the neurotypical group (ρ = −0.420, p = 0.083), but no significant association was found in the autism group. Lastly, age was found to marginally influence the neurotypical group, with older participants exhibiting higher alpha and theta power (alpha: ρ = 0.460, p = 0.055; theta: ρ = 0.456, p = 0.058). No age-related effects were observed in the autism group.

3.2. Adaptation Across Repeated Trials

We then examined how participant characteristics influenced adaptation over repeated VR trials using adaptation slopes of EEG power (alpha and theta) and trial duration. Significant correlations are summarized in

Table 4. Significant correlations between participant characteristics and adaptation slopes across trials.

Measure	Predictor	Autism (ρ, p)	Autism—Interpretation	Neurotypical (ρ, p)	Neurotypical—Interpretation
Trial Duration Slope	IQ	ρ = 0.120 (p = 0.671)	n.s.	ρ = 0.560 (p = 0.024 *)	Higher IQ significantly predicts less steep improvement across trials.
	Age	ρ = 0.009 (p = 0.975)	n.s.	ρ = 0.496 (p = 0.036 *)	Older participants had less steep improvement across trials.
	First VR Experience	ρ = 0.545 (p = 0.036 *)	Previous VR experience significantly associated with slower improvement across trials	ρ = − 0.591 (p = 0.010 **)	Previous VR experience significantly predicts faster adaptation across trials.
Alpha Slope	Tech Fluency	ρ = 0.079 (p = 0.780)	n.s.	ρ = −0.472 (p = 0.048 *)	Higher tech fluency significantly predicts steeper alpha suppression.
	Tech Enthusiasm	ρ = −0.535 (p = 0.040 *)	Higher enthusiasm significantly predicts steeper alpha suppression.	ρ = 0.296 (p = 0.232)	n.s.
Theta Slope	Tech Enthusiasm	ρ = −0.327 (p = 0.233)	n.s.	ρ = 0.627 (p = 0.005 **)	Higher enthusiasm significantly predicts positive theta.

Note: * significant (p < 0.05), ** highly significant (p < 0.01), n.s. not statistically significant (p > 0.05).

, with the full correlation matrix provided in Appendix A.2.

In the autism group, prior VR experience was significantly associated with trial duration slope (ρ = 0.545, p = 0.036), while in the neurotypical group, prior VR experience correlated with faster performance (ρ = −0.591, p = 0.010). IQ (ρ = 0.560, p = 0.024), and age (ρ = 0.496, p = 0.036) was positively correlated with trial duration slope in the neurotypical group, indicating less improvement across trials. Additionally, tech enthusiasm showed divergent neural patterns across groups: in autistic participants, higher tech enthusiasm correlated with increasing alpha suppression across trials (ρ = −0.535, p = 0.040), whereas in neurotypical participants, it significantly correlated with increasing theta activity over trials (ρ = 0.627, p = 0.005).

4. Discussion

This study explored how individual characteristics such as cognitive ability and technological familiarity impact both cognitive (EEG) and performance (trial duration) adaptation during repeated VR interaction tasks. It is important to note that all findings are based on correlational analyses. Thus, observed relationships should be interpreted as associations rather than causal effects. Additionally, in light of the study’s exploratory scope and limited sample size, results with p-values between 0.05 and 0.10 are interpreted as indicative trends rather than definitive findings. These findings offer valuable insights that can guide future confirmatory studies but should be interpreted cautiously.

4.1. Cognitive Factors: IQ and Age

Our findings reveal distinct patterns related to IQ and age in shaping task performance and neural activity during VR interactions. Higher IQ marginally predicted faster average trial durations in both groups. This aligns closely with established links between intelligence and cognitive efficiency, where more intelligent individuals typically utilize neural resources more effectively, facilitating quicker information processing and responses [22]. This confirms that IQ broadly facilitates performance efficiency in cognitively demanding VR environments. However, in the neurotypical group, IQ also significantly predicted a less steep adaptation slope. This seemingly counterintuitive result may reflect a ceiling effect, where higher-IQ neurotypical individuals performed near optimally at baseline, leaving limited room for observable improvement across repeated VR interactions.

EEG findings offered further nuance. Higher IQ correlated significantly with higher average alpha and theta power, but only in autistic participants. The relationship between IQ and alpha power in individuals with autism is consistent with foundational findings [23] and more recent work by Stankova et al. [24], both reporting positive correlations between higher IQ levels and increased alpha power, potentially indicating optimized cognitive efficiency or neural maturation. Further supporting this interpretation, studies by Finn et al. recently reported that peak alpha frequency (PAF) correlates with non-verbal IQ specifically in autism, but only with age in neurotypical individuals [25]. Although alpha PSD and PAF measure slightly different alpha characteristics (total alpha power versus dominant frequency), both metrics indicate underlying cognitive efficiency and neural maturation.

The theta findings provided additional autism-specific insights, as they correlated positively with IQ. Elevated theta power has consistently been linked to improved working memory, attentional regulation, and executive function during tasks that require significant cognitive effort [26,27,28]. Thus, the observed higher theta in autistic individuals with greater IQ may reflect more effective engagement of cognitive control mechanisms or working memory processes. The absence of a similar relationship in the neurotypical group is consistent with previous findings suggesting that theta’s functional role may differ based on population characteristics and task demands [29]. Consequently, the distinct alpha and theta EEG patterns regarding IQ underscore the unique neural resource allocation strategies employed by autistic individuals during VR interactions.

Interestingly, age exhibited a marginally significant positive correlation with average alpha power in neurotypical participants. Although alpha power typically declines with advancing age [30], our younger adult sample may reflect cognitive maturity or optimized attentional control rather than age-related decline. Within this narrower age range, increased alpha might represent enhanced attentional regulation or cognitive efficiency associated with neural maturation in our sample.

Collectively, these findings reveal the distinct cognitive–neural strategies utilized by autistic and neurotypical participants during VR interactions, highlighting how cognitive factors such as IQ and age uniquely modulate performance and neural efficiency in each group.

4.2. Technology-Related Factors

This section specifically discusses how the predictors of tech enthusiasm (subjective motivation and interest), tech fluency (objective familiarity with technology), and first VR experience influence both behavioral and neural responses in each group during VR interactions. Correlational analyses revealed that both tech fluency and tech enthusiasm were associated with EEG adaptation slopes (alpha and theta), and tech fluency was also found to be correlated with performance average (trial duration).

The tech fluency predictor showed that neurotypical participants with greater tech fluency had a faster average performance (shorter trial duration average). This finding supports the idea that prior experience in a domain can influence cognitive load, as experts typically have more efficient cognitive schemas for processing domain-specific information [31]. In contrast, autistic participants did not exhibit significant relationships with tech fluency, possibly indicating different cognitive strategies which were less influenced by prior general tech familiarity.

Tech enthusiasm showed different patterns. Neurotypical participants with higher tech enthusiasm show increasing theta activity across trials. This sustained level or increase in theta might reflect heightened motivation or intensified cognitive control strategies employed progressively over repeated VR interactions. On the other hand, autistic participants with greater tech enthusiasm exhibited more pronounced alpha suppression, which similarly might reflect heightened engagement with the VR task.

The distinct patterns between tech fluency and tech enthusiasm likely arise because tech fluency assesses objective familiarity and proficiency, while tech enthusiasm captures subjective motivation and interest. This distinction also aligns with the recent study of Fares et al., which emphasizes perceived enjoyment, ease of use, attitudes, and prior experiences as critical determinants for VR adoption [32]. Wong et al. similarly highlights openness and subjective enthusiasm toward technology as critical elements in enhancing VR engagement [33]. Furthermore, psychological predictors, such as immersive tendencies and individual innovativeness identified by Cummings et al., underscore the distinct influences of subjective motivation and objective technological familiarity on VR adoption and ongoing usage [34].

Additionally, the first VR experience predictor showed differences in performance adaptation trajectories; it correlated positively in autism group, but negatively in the neurotypical group. For autistic individuals, first-time VR use may pose greater sensory–motor and cognitive processing challenges, resulting in slower initial adaptation. As a result, prior VR experience likely provides an advantage by familiarizing them with the sensory demands of the environment, resulting in greater performance improvements over trials (steeper trial duration slope). Conversely, neurotypical first-time users tend to show rapid improvement. For this group, prior experience may reduce the observable adaptation slope, as their performance is already relatively more optimized from the start.

4.3. Future Directions

This study opens several promising directions for future research. First, while our study specifically focused on stable user characteristics, such as IQ, age, and technological familiarity, these findings can serve as foundational guidelines for designing personalized VR systems. Integrating these baseline user profiles with real-time EEG and performance feedback could further enhance personalization. Since individual learning curves and cognitive states may dynamically change throughout VR interactions, an adaptive system capable of real-time adjustments could better accommodate moment-to-moment variations. While technical implementation might present challenges, advancements in wearable EEG technology and real-time signal processing increasingly make such adaptive systems achievable.

Second, future research might address key methodological limitations from this study. The pronounced IQ gap between autistic and neurotypical participants, largely attributable to systemic gaps in early diagnosis and intervention access in Indonesia, means that findings from this study are primarily generalized to autistic individuals requiring higher levels of support. Additionally, the group-specific measures of technology fluency employed in this study, while contextually justified, limit direct numerical comparability across groups. Employing standardized measures would facilitate clearer cross-group comparisons and more robust conclusions in future work.

Third, expanding the range of user characteristics studied presents additional opportunities. Investigating other user-specific factors, such as sensory sensitivity, executive functions (e.g., working memory, planning), attention profiles, or motivational states, could greatly deepen our understanding of VR interaction dynamics. Broader user profiling can yield more targeted strategies to improve VR accessibility and engagement. In addition, replicating and extending this research with larger samples in future studies would help strengthen the reliability and generalizability of these exploratory findings.

Finally, while this research focused specifically on alpha and theta EEG bands, future studies could incorporate additional physiological measures (e.g., beta or gamma bands, heart rate variability) to provide deeper insights into cognitive and affective processes during VR interactions. Integrating these additional measures may help clarify or extend interpretations of how individual differences shape cognitive and performance adaptation in VR.

Taken together, these future directions provide a pathway for making VR systems more responsive, personalized, and accessible to a wider range of users. As an exploratory study, our findings lay a valuable foundation for further empirical investigations, such as experimental studies to investigate causal relationships, and inform practical applications in VR development. Future research could also explore direct comparisons of VR interaction approaches, or even entirely different VR systems, in terms of their therapeutic efficacy or specific cognitive outcomes for autistic individuals. Such comparisons would further advance the field by identifying optimal VR interventions tailored to distinct user profiles and therapeutic goals. Continued research along these lines can significantly advance VR as an effective support tool for both autistic and neurotypical populations.

5. Conclusions

This study examined how individual variability influences cognitive engagement (EEG alpha and theta activity) and performance adaptation (trial duration performance) during repeated VR interactions among autistic and neurotypical individuals. By investigating participant characteristics such as IQ, age, technological enthusiasm, tech fluency and prior VR experience, the study aimed to address a critical gap in understanding how personal traits modulate VR experiences.

The findings revealed distinct cognitive and performance patterns across groups. Higher IQ predicted elevated alpha and theta power exclusively among autistic participants, suggesting unique neural efficiency strategies. Age marginally influenced alpha power in neurotypical participants, potentially reflecting cognitive maturation. Additionally, technology-related factors further differentiated performance and neural engagement across groups. Collectively, these results underscore the importance of accounting for individual cognitive and technological familiarity when designing VR environments for diverse user populations.

These insights have practical implications for the development of more adaptive and personalized VR systems. Tailoring VR experiences based on user profiles could enhance engagement, learning outcomes, and accessibility, particularly for neurodiverse users. While the current findings are correlational and based on a modest sample size, they offer a valuable foundation for advancing VR design and broadening the understanding of individual variability in immersive environments.