Modulation of a Rubber Hand Illusion by Different Levels of Mental Workload: An EEG Study

Abstract

The current study aimed to investigate the impact of externally evoked mental workload on the level of an artificial hand ownership sensation, a component of the embodiment phenomenon (feeling an external object, in this case a fake upper limb, as part of one’s body). The process of embodiment is extensively investigated in the literature also to find solutions for promoting the acceptance of prosthetic limbs. Before a traditional procedure for summoning in healthy subjects a Rubber Hand Illusion (RHI), the participants performed memory-related tasks in easy or demanding conditions to generate, respectively, low and high mental workloads. Alongside the behavioral correlates of the body ownership in the form of a proprioceptive drift (the measure of the correspondence between the perceived position of the actual limb and the fake one), EEG data was also collected. The results, both behavioral and neural, suggest that a high mental workload before the RHI experience leads to a low level of body ownership, whereas a low one enhances it. This can be interpreted as a consequence of distracting mental resources (possibly a specific type of them) from the embodiment stimulation session.

1. Introduction

The sense of body ownership refers to the feeling or perception that one’s body belongs to them. It is a fundamental aspect of self-awareness and is closely related to bodily self-consciousness and, in particular, is one of the main components of the embodiment (feeling an external object as part of the individual body scheme) [1,2,3,4]. This sense of ownership is typically experienced as a seamless integration between sensory input, motor control, and body awareness [5,6], and it is studied to find strategies to promote the embodiment of artificial limbs to improve prosthetic acceptance (several users tend to abandon their bionic limbs if they do not show enough reliability immediately [7]).

Research in neuroscience [8,9] and psychology [10,11,12] has revealed that the brain integrates information from various sensory modalities to explore the surrounding environment and the self in relation to it. This multisensory integration underlies a wide range of perceptual, cognitive, and motor functions, enabling adaptive interaction with the world. For example, studies using paradigms like the Rubber Hand Illusion (RHI) [13] have shown how the brain can be tricked into feeling ownership over an artificial limb when sensory input (such as touch) is synchronized between the artificial limb and the participant’s real hand. By simultaneously applying tactile stimulation on both the rubber hand (observed by the subject) and the real (hidden) hand, the brain integrates visual and tactile information, leading to the sensation that the rubber hand is part of their body. Multisensory congruency is crucial in inducing the Rubber Hand Illusion [14]. This refers to the alignment or synchronization of sensory information across different modalities, such as vision and touch, in this case. When the visual input of seeing the rubber hand being touched is congruent with the tactile feeling, the brain is more likely to merge these signals and incorporate the rubber hand into one’s body representation by recruiting attentional resources throughout the sensory modalities. Nonetheless, to effectively maintain the attentional resources and keep the feeling of the embodiment on a stable level, the mental workload needs to be monitored, since it is overloading may reduce the ability to process the rubber hand as an external stimulus [15,16,17,18]. Such a premise seems supported by recent studies [19,20,21] on virtual and augmented settings for prosthetic embodiment training demonstrating how secondary tasks can alter the latter by engaging mental resources probably necessary for this type of process.

In the present EEG, study two levels of mental workload [22] were imposed on the participants during a task performed before the induction of the Rubber Hand Illusion (RHI) to see its impact on the strength of the resulting body ownership. The latter was assessed behaviorally by means of proprioceptive (perceptual) drift towards the fake hand and neurally by EEG measures. The results suggest that the higher mental workload in a previous task lowers the ownership level obtained through the execution of a subsequent RHI protocol whereas the lower workload enhances it. Besides the behavioral evidence, the EEG findings provide neural mechanisms underlying this modulation. The EEG correlates of the RHI, computed by using Sample Entropy (SE) [23,24], differ from each other, depending on the level of priorly imposed mental workload. Sample Entropy, which quantifies the irregularity and complexity of brain dynamics, was higher during RHIs preceded by low workload, indicating more flexible and adaptive neural activity. In contrast, after high workload conditions, entropy values were reduced, reflecting more regular and predictable neural patterns typically associated with cognitive fatigue and diminished processing capacity.

Moreover, the SE differences were most prominent in frontal and central scalp regions, which are known for attentional control [25]. This suggests that high cognitive demand before RHI induction constrains the dynamical repertoire of frontal networks, leaving fewer neural resources available for integrating visuotactile signals and constructing a sense of ownership over the artificial hand. Thus, the EEG results provide converging evidence that embodiment is not solely a perceptual phenomenon but also critically dependent on the availability of cognitive resources and the adaptive capacity of cortical dynamics.

2. Materials and Methods

2.1. Subjects

A total of 8 healthy, right-handed female participants (mean age = 26.5, SD = 1.6) were recruited for the study. The subjects are students and technicians of the Rehab Technologies lab of Istituto Italiana di Tecnologia (IIT), Via Morego, 30-16163, Genoa, Italy. Since each subject was tested two times with an interval of one day, the levels of the mental workload tasks were randomized.

A simulation-based post hoc power analysis was conducted using the current sample size (8 subjects). The analysis revealed that the study had approximately 75.2% power to detect a large effect size (Cohen’s d = 1.0) at a significance level of 0.05, which slightly falls below the commonly accepted threshold of 80% for adequate power, but the current study has an explorative objective, which will give a start for a broader study both by means of sample size and experimental upgrades, such as implementing a closed-loop paradigm to detect an embodiment online.

2.2. Apparatus and Stimuli

Two types of stimuli were used during the experiment. The first one was used to provide a controlled mental workload called the dual N-back task, and the second one was the Rubber Hand Illusion (RHI).

2.2.1. Dual N-Back Task

The stimulus used to manipulate the mental workload is the well-known N-back task [26]. Particularly, the picture version of Dual N-back task [27] PC application was used. In this setup, participants are presented with a series of stimuli, particularly fruit pictures accompanied with letter sounds, and are asked to indicate whether the current stimulus (either audio or visual) matches the one presented n steps back in the sequence.

The duration of the task was 20 min.

After completing the task, the participants were asked to fill out the NASA Task Load Index (NASA TLX) questionnaire [28], which is a widely used subjective workload assessment tool developed by the Human Performance Group at NASA Ames Research Center. It is designed to evaluate perceived workload in a variety of domains, including aviation, healthcare, and other complex task environments. The tool was originally developed for the assessment of workload in aviation, but it has been adapted for use in different fields.

NASA TLX assesses workload based on the following six dimensions:

• Mental Demand: How much mental and perceptual activity was required?
• Physical Demand: How much physical activity was required?
• Temporal Demand: How much time pressure did you feel due to the pace of the task?
• Performance: How successful were you in accomplishing the goals of the task?
• Effort: How hard did you have to work to accomplish your level of performance?
• Frustration Level: How annoyed, irritated, or frustrated were you during the task?

2.2.2. Rubber Hand Illusion (RHI)

The life-sized rubber hand was placed in front of the participant, making sure that the participant’s real hand was hidden from view under a cloth. Moreover, the rubber and real hand were placed in a similar posture.

Participants were instructed to visually attend to the rubber hand throughout the procedure. A vibrating sensor was affixed to the participant’s real hand and was wirelessly triggered in synchrony with tactile stimulation of the rubber hand using a custom-built actuator resembling a pen-like clicking device. Each time the rubber hand was touched, the sensor on the real hand delivered a corresponding vibration. This synchronous visuotactile stimulation was designed to induce the Rubber Hand Illusion, whereby participants may experience the tactile sensations perceived on the rubber hand as originating from their own hidden hand. As a marker for successful embodiment during RHI, the so-called proprioceptive drift was used [13]. Proprioception refers to the sense of the relative position of one’s own parts of the body and the effort being employed in movement. In the context of the RHI, proprioceptive drift refers to the perceived movement or shift in the position of the participant’s real hand toward the location of the rubber hand.

As the participant’s brain integrates visual and tactile information, they may feel as though their real hand is moving towards the rubber hand, even though it remains in the same position.

2.3. Experiment

The experiment consists of the following blocks:

• Resting (baseline) state 5 min;
• Rubber Hand Illusion 5 min;
• One-back and Two-back tasks for each subject on different days; each task lasts 20 min followed by the NASA TLX questionnaire;
• Rubber Hand Illusion 5 min.

2.4. EEG Recording and Preprocessing

The EEG signals were acquired by a BrainAmp system (Brain Products, Zeppelinstrasse 7, 82205, Gilching, Germany). Its output was digitized with a resolution of 16 bits and sampled at a rate of 1000 Hz via the Lab Streaming Layer (LSL) protocol [29], a framework used for real-time collection and transmission of time-sensitive data such as EEG and EMG. For data acquisition, we mounted a set of 64 Ag/AgCl electrodes according to the 10/20 system and referenced it to FCz. We kept the impedances below 20 kΩ with the ground and the reference electrode below 5 kΩ throughout the recording. The offline analysis of the acquired data was performed by using the EEGLAB toolbox [30] operating under the MATLAB R2022b software. The EEG data were re-referenced to the average reference, which involves subtracting the average voltage across all electrodes from each electrode’s voltage and bandpass filtered between 0.5 Hz and 45 Hz to remove slow drifts and high-frequency noise. Artifacts, including eye blinks and muscle activity, were removed using Independent Component Analysis (ICA) [31]. ICA components were rejected based on characteristic spatial topographies, time-course patterns, and frequency profiles. In addition, residual artifacts such as electromyographic or electrooculographic activity were minimized by visually inspecting the cleaned data and excluding contaminated epochs if necessary. The pre-processed data were visually inspected to ensure the removal of artifacts and were saved for further analysis.

2.5. Sample Entropy of EEG Data

Sample Entropy is a metric used to quantify the irregularity, complexity, or predictability of a time series. It is commonly applied in the analysis of physiological signals, such as EEG (electroencephalogram), ECG (electrocardiogram), and other biomedical data. Applying Sample Entropy (SE) analysis to EEG signals in the context of the Rubber Hand Illusion provides a quantitative measure of the complexity and regularity of brain activity during the illusion. It can offer valuable insights into the neural mechanisms underlying body perception and multisensory integration. Therefore, we used the Sample Entropy of the acquired EEG data as a discriminative tool to detect the modulation of the RHI.

SE is calculated in the following way:

Given a time series of length N, Sample Entropy is calculated by counting the number of similar sequences of length m (embedding dimension) that are close to each other within a specified tolerance level r. The formula for Sample Entropy ($\mathrm{S}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{E}\mathrm{n}$) is as follows:

$$\mathrm{S}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{E}\mathrm{n}\left(\mathrm{m},\mathrm{r},\mathrm{N}\right)=-\ln {\displaystyle \frac{{\mathrm{A}}^{\mathrm{m}+1}(\mathrm{r})}{{\mathrm{A}}^{\mathrm{m}}(\mathrm{r})}}$$

where ${\mathrm{A}}^{\mathrm{m}}(\mathrm{r})$ is the average number of sequences of length m that are similar within a tolerance r, and ${\mathrm{A}}^{\mathrm{m}+1}(\mathrm{r})$ is the average number of sequences of length m + 1 that are similar within a tolerance r.

Parameters are as follows:

• m: embedding dimension, representing the length of sequences to be compared.
• r: tolerance level, indicating the maximum acceptable difference between two sequences.
• N: length of the time series.

The Sample Entropy of the preprocessed EEG series corresponding to the Rubber Hand Illusions after both levels of mental workload was computed for each 1 s period. This way the time development of the neural correlates of the Rubber Hand Illusion could be observed.

3. Results

3.1. Psychophysiological Results

As a psychophysical (quantitative) measure of the exposed Rubber Hand Illusion, we used the proprioceptive drift (PD) of the subjects, particularly the difference between the pre- and post-Rubber Hand Illusion exposure. The results show (Figure 1) that the perceptual drift is observed after low mental workload, but not after the high one, meaning that high mental workload weakened the feeling of body ownership and thus of embodiment. To assess the difference between the PDs corresponding to low and high mental workloads, a two-sample t-test was performed. To justify the choice of t-test, we evaluated the normality of data distribution by using the Lilliefors test (lillietest) and the Anderson–Darling test (adtest) in MATLAB. The data appeared to be normally distributed, since both tests returned non-significant results of p > 0.05. The t-test result showed a significant difference (p = 0.0064, t (7) = −3.9676) between the samples.

3.2. Self-Report (NASA TLX) Results

The results of self-assessment using the NASA TLX form are shown in Figure 2.

3.3. EEG Results

The entropies of the EEG data corresponding to the 5-min Rubber Hand Illusions after one-back and two-back, accordingly, were computed for pre- and post-exposures, and the difference between them was considered as a final entropic measure. Sample Entropy (SE) was computed for each chunk of 1 s throughout the whole 5-min EEG stream. The scalp distribution of the mean entropy values through the whole exposure period for both tasks shows activations of primary visual and frontal-central areas (Figure 3). To assess exactly which areas show significant difference by means of activation, we performed two-sample t-test analysis for each channel separately. The scalp distribution of the p-values (Figure 4) shows that the underlying neural dynamics (entropies) of the Rubber Hand Illusions after low and high mental workload tasks are significantly different in frontal central areas.

Figure 5 illustrates the variation of Sample Entropy over time for the Rubber Hand Illusions after one-back and two-back conditions. The blue curve represents the one-back condition, while the red curve corresponds to the two-back condition. Shaded regions indicate the variability (standard errors) for each condition.

Initially, Sample Entropy for both Rubber Hand Illusions starts at comparable levels, but over time, a distinct divergence emerges. The one-back condition exhibits a progressive increase in Sample Entropy, particularly beyond the 150-s mark, indicating greater complexity and variability in neural dynamics. In contrast, the two-back condition shows a relatively stable trend initially, followed by a gradual decline in Sample Entropy after 150 s. This may suggest that increased cognitive load in the two-back task is associated with reduced dynamical complexity, possibly reflecting higher task demands and cognitive fatigue.

3.4. Statistical Analysis and Results

The normality of the data distributions was evaluated using both statistical tests and visual inspection in MATLAB. Specifically, we applied the Lilliefors test (lillietest) and the Anderson–Darling test (adtest) to each continuous variable. These tests assess whether a sample deviates significantly from a normal distribution. A variable was considered approximately normal if both tests returned non-significant results (p > 0.05).

For SE: Lilliefors test: h = 0, p = 0.5000; Anderson–Darling test: h = 0, p = 0.7415;

For proprioceptive drift: Lilliefors test: h = 0, p = 0.3938, Anderson–Darling test: h = 0, p = 0.1020;

For NASA-TLX questionnaire components: 0.1 < p < 0.5, h = 0.

Linear mixed-effects model (LME) [32] was performed to assess the influence of Entropy_EEG, Mental Demand, and task on Drift, including all interactions, with subjects as a random effect in the form of the following formula: Drift∼Entropy_EEG × MentalDemand × task + (1∣subjects), where Drift is the dependent variable, and Entropy_EEG, Mental Demand, and task are fixed effects, including all their interactions. A random intercept for subjects is included to account for inter-individual variability.

The analysis revealed (

Table 1. ANOVA results for the effects of Entropy_EEG, Mental Demand, and Task on Drift.

Term	F-Statistic	DF1	DF2	p-Value
(Intercept)	9.5661	1	8	0.014824
MentalDemand	94.576	1	8	1.0448 × 10−5
Task	0.0079291	1	8	0.93123
Entropy_EEG	23.157	1	8	0.0013346
MentalDemand:task	284.87	1	8	1.5392 × 10−7
MentalDemand:Entropy_EEG	54.729	1	8	7.6348 × 10−5
task:Entropy_EEG	10.328	1	8	0.012354
MentalDemand:task:Entropy_EEG	29.314	1	8	0.00063511

) a significant main effect of Mental Demand (F(1,8) = 94.58, p < 0.001) and Entropy_EEG (F(1,8) = 23.16, p = 0.001), suggesting that both mental workload and EEG entropy are key predictors of drift. However, the main effect of the task was not significant (F (1,8) = 0.008, p = 0.931), indicating that the task type alone did not influence drift.

Several significant interaction effects emerged. A strong two-way interaction was observed between Mental Demand and task (F(1,8) = 284.87, p < 0.001), suggesting that the effect of cognitive workload on drift depends on the task context. Additionally, significant interactions were found between Mental Demand and Entropy_EEG (F(1,8) = 54.73, p < 0.001) and between task and Entropy_EEG (F (1,8) = 10.33, p = 0.012), highlighting the role of EEG entropy in modulating drift across different levels of cognitive demand and task conditions.

Crucially, the three-way interaction between Mental Demand, task, and Entropy_EEG was also significant (F(1,8) = 29.31, p < 0.001). Specifically, we found that EEG entropy (as measured by SampEn) decreased with increasing mental workload, but this effect was modulated by task demands: the reduction in entropy was more pronounced under [two-back-task or high-complexity] conditions than under [e.g., low-complexity or one-back-task] conditions. This suggests that the relationship between neural complexity and workload is not linear or uniform across contexts.

Overall, these results emphasize the critical role of EEG entropy and cognitive workload in modulating drift, with their effects being contingent on task demands.

4. Discussion

The Rubber Hand Illusion (RHI) is a proven experimental paradigm that investigates the sense of body ownership by creating the illusion that a rubber hand is part of one’s own body. This is typically achieved through synchronous visuotactile stimulation, where both the real hidden hand and the visible rubber hand are stroked simultaneously, leading to the sensation that the rubber hand belongs to the participant.

Electroencephalography (EEG) studies have provided insights into the neural mechanisms underlying the RHI [33,34,35,36,37,38,39]. Notably, reductions in theta and beta-band oscillatory activity have been observed during the illusion, suggesting increased sensory-motor system excitability associated with altered body ownership.

Mental workload, defined as the cognitive demand placed on an individual during task performance, has been shown to influence neural activity patterns. EEG studies [40,41,42] have demonstrated that increases in mental workload are associated with elevated theta activity, particularly in frontal regions, reflecting heightened cognitive engagement.

The interplay between mental workload and RHI is an emerging area of research [43]. It is plausible that varying levels of mental workload could modulate the strength or occurrence of the RHI by influencing attentional resources and multisensory integration processes. For instance, a high mental workload might either enhance or diminish the illusion depending on whether cognitive resources are directed towards or away from processing the multisensory stimuli essential for the RHI.

An EEG study investigating this interaction could involve manipulating mental workload levels while inducing the RHI and measuring corresponding neural responses. Such a study would provide valuable insights into how cognitive demands affect body ownership experiences and the underlying neural dynamics.

In this study, Sample Entropy (SE) [23,44,45,46,47,48] was employed to analyze EEG signals associated with the Rubber Hand Illusion (RHI) under varying mental workload conditions. The choice of SE over conventional time-frequency methods was motivated by its unique ability to capture the complexity and irregularity of neural activity, which is particularly relevant when investigating dynamic cognitive states such as body ownership and attentional modulation.

Unlike traditional spectral analyses (e.g., Fourier or wavelet transforms), which focus on the distribution of power across frequency bands, SE quantifies the unpredictability of signal patterns over time. This allows for the detection of subtle changes in neural dynamics that may not manifest as shifts in oscillatory power but rather as alterations in signal regularity. In the context of the RHI, where perceptual and cognitive states fluctuate rapidly, SE provides a more sensitive measure of the brain’s adaptive responses.

Moreover, SE is robust to non-stationarity, a common characteristic of EEG data during naturalistic tasks. Time-frequency methods often rely on assumptions of local stationarity or require windowing strategies that can obscure transient neural events. SE, by contrast, can be applied directly to short and variable-length EEG segments, making it well-suited for capturing the moment-to-moment changes in brain activity associated with illusion onset and cognitive load.

The results of this study demonstrated that SE effectively distinguished between illusion and non-illusion states, with lower entropy values observed under a high mental workload conditions. This reduction in complexity is consistent with previous findings on cognitive fatigue and attentional depletion [49,50,51,52], suggesting that SE can serve as a neural marker of cognitive resource allocation.

Understanding the modulation of the RHI by mental workload has practical implications. For example, in neurorehabilitation, where virtual reality and multisensory feedback are used to restore motor functions, tailoring the cognitive load could optimize therapeutic outcomes by enhancing the sense of embodiment.

In conclusion, exploring how different levels of mental workload influence the RHI through EEG measures offers a promising avenue for understanding the cognitive and neural mechanisms of body ownership and could inform the development of interventions that leverage these mechanisms for therapeutic purposes.

The study presented had an objective to examine the effect of mental workload on the RHI. Our results indicated that increased mental workload significantly changes the experience of RHI. Particularly, subjects imposed to higher mental workload conditions reported lower (or not at all) proprioceptive drift towards the rubber hand compared to those after lower mental workload conditions. These results stipulate that attentional resources play a crucial role in the integration of multisensory information necessary for RHI. The reduced susceptibility to the illusion after high mental workload conditions may be due to the competition for attentional resources. When the brain is still preoccupied with a demanding cognitive task, it may allocate fewer resources to the processes underlying the RHI, such as the integration of visual and tactile inputs and the recalibration of body ownership.

The neurophysiological results indicate that the time development of sample entropies of EEG signals can distinguish between two perceptions of the rubber hand. The results suggest that there are measurable differences in the regularity of neural activity associated with each state. Higher Sample Entropy values typically indicate more complex and less predictable signals, whereas lower values suggest more regular and predictable patterns. In the presented case the lower entropy values correspond to RHI after the exposure to the higher mental demand task, which is consistent with study results stipulating that the EEG complexity (entropy) tends to decrease during cognitive fatigue [48,49,52]. The latter may suggest that the difference in EEG correlates of RHIs in the form of Sample Entropy features may be due to cognitive fatigue.

One notable limitation of this study is the relatively small sample size, which may affect the generalizability of the findings. While the observed effects of mental workload on the Rubber Hand Illusion (RHI) and the associated EEG entropy measures were statistically significant, a larger and more diverse participant pool would strengthen the robustness of these conclusions. Small sample sizes can increase the risk of Type I and Type II errors and may limit the ability to detect subtle individual differences in susceptibility to the illusion or cognitive fatigue. Future studies should aim to replicate these findings with larger cohorts to validate the observed trends and explore potential moderating factors such as age, cognitive capacity, or sensory sensitivity.

5. Conclusions

This study explored how varying levels of mental workload modulate the Rubber Hand Illusion (RHI), combining behavioral assessments with EEG-based neural complexity analysis. The findings provide compelling evidence that increased cognitive demand significantly weakens the strength of the illusion, as reflected by reduced proprioceptive drift and changes in EEG entropy. These results highlight the critical role of attentional resources in multisensory integration and body ownership.

Importantly, the use of Sample Entropy as a neural marker revealed that cortical activity becomes more regular and less complex under high workload conditions, suggesting that cognitive fatigue or resource competition may impair the brain’s ability to sustain the illusion. This dynamic interaction between mental workload and embodiment has practical implications for improving prosthetic technology acceptance. By understanding how cognitive demands influence the sense of ownership over artificial limbs, therapeutic interventions and device interfaces can be better tailored to support embodiment, especially in cognitively demanding environments.

Future research should expand on these findings by incorporating a broader range of cognitive tasks that engage distinct components of working memory, such as phonological versus visuospatial processing. This would allow for a more nuanced understanding of how different types of mental resources affect prosthetic embodiment and multisensory integration.

However, the study’s conclusions are tempered by the limitation of a small sample size, which may restrict the generalizability of the results. Larger-scale studies are needed to validate these findings and explore individual differences in susceptibility to the illusion under varying cognitive loads.

In summary, this work contributes to the growing body of research on the cognitive modulation of body ownership and offers valuable insights for the design of adaptive prosthetic systems and rehabilitation protocols that account for mental workload and attentional capacity.