Effects of tRNS on perception of static and dynamic displays of social-emotional communication

Abstract

Accurately decoding emotions from facial and vocal cues is essential to successful social interaction. The human mirror neuron system (hMNS) is thought to support this through sensorimotor simulation of observed emotional expressions. While prior studies linked hMNS activity—indexed by mu rhythm desynchronization (mu-ERD)—to emotional action perception, causal evidence with dynamic, multimodal social stimuli remains limited. We investigated whether transcranial random noise stimulation (tRNS) to the inferior frontal cortex (IFC), a key node of the hMNS, enhances perception of static and dynamic emotional displays. Fifty-two participants received active or sham tRNS over the IFC. Consistent with pre-registered predictions, active tRNS led to better performance on static emotion perception tasks compared to sham, with no group difference on a static identity-matching control task—validating stimulation specificity. Extending prior work, active tRNS led to faster response times and greater mu-ERD measured by EEG during dynamic audio-visual emotion perception tasks, consistent with predictions relating to enhanced sensorimotor simulation. These findings suggest that tRNS to the IFC can augment rapid, embodied emotion perception—particularly when stimuli more closely approximate real-world social communication—and contribute to the causal mapping of the hMNS, opening new avenues for studying social-emotional function in neurotypical and clinical populations.

Emotion is at the heart of human communication. Whether in casual conversation or high-stakes interaction, we rely on a dynamic blend of facial expressions, vocal tone, and speech content to understand one another. A smile paired with a warm tone can signal friendliness, while a flat expression and monotone voice may hint at disinterest or distress—even when the words say otherwise. Facial and vocal cues often operate in tandem, each enriching and clarifying the emotional message conveyed. This interplay is so seamless in everyday life that we rarely reflect on the complex perceptual and neural processes that support it. Yet understanding how we decode emotion from faces and voices—especially when cues are dynamic and multimodal—has become a central focus in affective and social neuroscience.

A proposed mechanism involved in emotional communication is mirror neurons. While early animal research on mirror neurons was received with much excitement, there have been some important criticisms regarding its extension to humans. Mirror neurons were first discovered while recording individual neurons in the premotor cortex of macaque monkeys (Di Pellegrino et al. 1992). Critics highlight gross anatomical and connectivity differences limiting the generalization of these animal findings to humans (Hickok 2009). Additionally, because of the invasive methods required to validate whether a neuron truly has mirroring properties, there exists limited direct evidence for mirror neurons in humans (but see Mukamel et al. 2010). Nonetheless, there is some consensus that a network exists in humans that has functional properties resembling mirror neurons which can support the observation of intentional actions, known as the human mirror neuron system (hMNS). A broader network of brain regions activated during the observation of actions, regardless of whether those neurons are also active during execution, has been referred to as the action observation network. For the purposes of the current investigation, we are interested in brain regions associated with the hMNS specifically, and their involvement in the execution and observation of vocal-facial emotion.

 

The human mirror neuron system

The hMNS is one of the brain networks proposed to be involved in action perception. It is active when one executes or observes intentional biological movement (Oberman et al. 2005). The hMNS is thought to map observed actions onto the observer’s own sensory and motor representations to aid in action understanding (Rizzolatti and Craighero 2004, Oberman et al. 2007). While earlier work focused on hand movements and action understanding, biological movement also includes facial movements. More recently, researchers proposed that the hMNS is engaged when processing facial expressions, supporting understanding of facial expressions (Prochazkova and Kret 2017). Consequently, it has been linked to imitation, theory of mind, empathy, and social-emotional communication (Rajmohan and Mohandas 2007).

Functional magnetic resonance imaging (fMRI) identifies the inferior frontal cortex (IFC), inferior parietal lobule (IPL) and the superior temporal sulcus (STS) as core areas of the hMNS. The STS processes multisensory information from observed actions which are then projected to the IPL and the IFC. The IPL codes for somatosensory information (Iacoboni et al. 1999, Rajmohan and Mohandas 2007) while the IFC creates motor plans that together form a sensorimotor simulation of the observed action (Iacoboni et al. 1999).

Mu-ERD

Measurement of the mu rhythm using electroencephalogram (EEG) allows for non-invasive tracking of the hMNS. Event-related desynchronization of activity in the mu-band (i.e. mu-ERD) is the reduction in power (i.e. amplitude) of the mu rhythm (an alpha wave which oscillates in the 8–13 Hz frequency band) after stimulus onset, resulting from independent sources of activity within and across the nodes of the network. In contrast, an event-related synchronization (ERS) can be observed when an individual is at rest, leading to synchronization within and across the nodes of the network.

Mu-ERD is typically measured over central electrodes (e.g. C3, Cz, and C4), situated between the IFC and IPL; two major nodes of the hMNS. Accordingly, EEG-fMRI studies show that mu-ERD is correlated with blood oxygenation in nodes of the hMNS (i.e. the IFC and IPL; e.g. Arnstein et al. 2011). Similarly, EEG source localization has identified frontal and parietal areas as sources of mu rhythm (Fox et al. 2016). Thus, mu-ERD over central electrodes is suggested to reflect the downstream activity of these sensorimotor areas.

Mu-ERD and action perception

Mu-ERD is associated with biological action perception and execution (Muthukumaraswamy and Johnson 2004). For instance, Gutsell and Inzlicht (2010) found that watching and performing a drinking action elicited significant mu-ERD from baseline, with stronger mu-ERD during action execution than observation. The relation between mu-ERD, action execution, and action observation changes as a function of social–emotional communication ability.

Autistic people are broadly characterized by challenges in social-emotional communication (Clark et al. 2008), imitation, and theory of mind (Oberman et al. 2005). Compared to controls, autistic people also exhibit decreased activation (Bookheimer et al. 2008) and anatomical differences in hMNS regions like decreased grey matter and cortical thinning (Hadjikhani et al. 2006). Correspondingly, they exhibit differences in mu-ERD response to biological movement compared to controls.

In Oberman et al. (2005), the control group exhibited mu-ERD when executing and observing hand movement, but the autistic group only exhibited mu-ERD during execution. Neither group showed mu-ERD when viewing bouncing balls (i.e. non-biological motion). Similarly, in Bernier et al. (2007), autistic adults and controls showed mu-ERD when executing or imitating a grip, but the control group had greater mu-ERD than the autistic group when observing a grip. These results suggest that autistic people show typical mu-ERD when executing action but reduced or absent mu-ERD during passive observation of others’ actions, indicating impaired sensorimotor simulation of others’ movements.

People living with Parkinson’s Disease (PD) also experience social communication deficits and altered mu-ERD compared to controls. While the controls exhibited mu-ERD when watching hand movements, mu-ERD was absent in persons with PD (Heida et al. 2014). Abnormal mu-ERD to biological movement may reflect impaired sensorimotor simulation of observed actions; an ability that supports understanding others’ actions and social interaction. Recent research extends earlier work on non-emotional hand movements to emotional facial expressions, supporting the relevance of mu-ERD in studying social-emotional communication.

Mu-ERD and emotional facial perception

Mu-ERD is also associated with execution and observation of facial expressions. Children exhibited significant mu-ERD from baseline when executing and observing emotional and neutral facial movements from the Amsterdam Dynamic Facial Expression Set (ADFES). This effect was unique to biological facial movement as children showed greater mu-ERD for human facial movements than for moving pixels of a pixelated human face (Rayson et al. 2016). Similarly, adults elicited greater mu-ERD to images of emotional faces than buildings (Moore et al. 2012).

Mu-ERD is particularly sensitive to socially relevant stimuli. Ensenberg et al. (2017) found that adults showed greater mu-ERD when ADFES faces expressed emotions when facing toward participants versus facing away, implicating mu-ERD in higher perceptual processes like social interactions. Rovetti et al. (2022) found greater activation in the left pre-supplementary motor area—a frontal region associated with the extended hMNS—when participants viewed emotional stimuli compared to neutral ones. This effect was observed consistently across facial expressions, vocal cues, and combined audio-visual presentations. A neural-behavior link has also been found with mu-ERD and emotion perception. Across all modalities (i.e. audio, video, audio-visual), greater mu-ERD was significantly related to greater emotion perception accuracy (Genzer et al. 2022).

Transcranial electrical brain stimulation

Brain stimulation can modulate neural activity and assess the role of particular brain regions in emotion perception, allowing researchers to examine the directionality of the relationship between brain activity and emotion perception. Transcranial electrical brain stimulation (tES) non-invasively alters cortical excitability, with 10–30 minutes of stimulation producing aftereffects for up to 90 minutes, allowing researchers to study behavioral and neural changes (Fertonani and Miniussi 2017, Thair et al. 2017).

Wu et al. (2018) used a within-subjects design to apply anodal, cathodal, and sham transcranial direct current stimulation (tDCS) to the IFC. After each stimulation, participants completed the Multifaceted Empathy Test, identifying emotion from static facial images. Cathodal tDCS reduced task accuracy compared to sham while anodal tDCS improved task accuracy, illustrating the role of the IFC in empathy. Gundlach et al. (2017) modulated the mu-rhythm by applying blocks of active and sham transcranial alternating current stimulation (tACS) to electrode sites CP3 and CP4 while recording EEG. There were no differences in the mu-rhythm amplitude prior to stimulation, but after stimulation, active tACS led to a decrease in the mu-rhythm amplitude compared to sham. Penton et al. (2017) used transcranial random noise stimulation (tRNS) to stimulate the IFC, an area implicated in emotion perception. Adults received either sham or active tRNS and then completed the three versions of the Cambridge Face Perception Test (CFPT): the CFPT-Identity, CFPT-Angry and CFPT-Happy. The groups did not differ on the facial identity (i.e., control) task, but the active tRNS group outperformed the sham group on the emotion perception tasks, demonstrating that tRNS to the IFC can increase emotion perception in young adults. Similarly, in older adults, active tRNS to the IFC improved perception of angry faces, but not the perception of happy faces or facial identity compared to sham (Yang and Banissy 2017).

The study by Penton et al. (2017) is of particular interest since the IFC is a major node of the hMNS, but no neural data was collected to assess the neural effects of stimulation. Neural data like mu-ERD could clarify whether tRNS to the IFC engages the hMNS which may explain the improved emotion perception performance. If tRNS to the IFC can engage the hMNS, it would allow future researchers to modulate the hMNS and study its role in emotion perception outside of correlational studies.

Stimuli used within the literature

Emotion perception studies tend to use static faces (e.g. CFPT), but responses to static versus dynamic stimuli can differ. Dynamic stimuli (e.g. a neutral face morphing into an emotional face) can elicit greater arousal (Sato and Yoshikawa 2007a), emotion perception accuracy (Ambadar et al. 2005) and facial mimicry than static stimuli (Sato and Yoshikawa 2007b; Rymarczyk et al. 2011). fMRI studies also find neural differences like greater STS activation for dynamic versus static stimuli (Schultz and Pilz 2009).

While some studies use dynamic stimulus sets (e.g. ADFES), these sets are still not representative of real-world social communication as they tend to involve a face changing from a neutral to emotional expression. In contrast, real-world social interactions generally do not follow such a prescribed trajectory, and audio cues like vocal intonation tend to be available. Mu-ERD is strongest for biological stimuli, especially when presented audio-visually (McGarry et al. 2012, Copelli et al. 2022). Likewise, facial mimicry is greater for dynamic audio-visual stimuli than for audio- and visual-only stimuli (Isomura and Nakano 2016). Therefore, dynamic audio-visual stimuli may better approximate real-world emotion perception than a dynamic visual-only stimulus set.

The current study

The current study aimed to replicate and extend the findings of Penton et al. (2017) to explore whether the hMNS is engaged through high frequency tRNS to the IFC by assessing whether mu-ERD is modulated post-stimulation. It also explores the effects of tRNS on the accuracy and response times during an emotion perception task using dynamic audio-visual stimuli to better approximate real-world social communication.

We hypothesized that our results would replicate Penton et al. (2017) such that those that receive active tRNS will outperform the sham group on the emotion perception tasks using static stimuli, but not the control task. For emotion perception tasks using dynamic audio-visual stimuli, we expected that active tRNS to the IFC would elicit more mu-ERD, greater task accuracy and faster response times than sham. All hypotheses were pre-registered on Open Science Framework: https://doi.org/10.17605/OSF.IO/QS8FP.

Methods

Participants

Fifty-four participants were recruited from the Toronto Metropolitan University community and undergraduate psychology research pool. The target sample size was determined by G*Power 3.1 (power = .9, d = .82, α = .05, one-tailed; Faul et al. 2007) based on the between-group effect of tRNS on emotion perception task accuracy in Penton et al. (2017). Two participants were removed due to a procedural error wherein the sound booth door was not fully closed, resulting in a final sample of 52 participants (71%, n = 37 female) ranging from 17 to 50 years old (M_age = 22.33, SD_age = 6.93). Due to a technical error in which the RAVDESS task computer crashed, neural and behavioral data during the RAVDESS tasks are missing from four participants. All participants were proficient in English and had no neurological disorders or brain injury. The Research Ethics Board at Toronto Metropolitan University (protocol number 2019-300) approved this study.

Study design

A single-blind design was used with random assignment to one of two stimulation groups: active tRNS (n = 26; M_age = 23.7, SD_age = 8.19) or sham tRNS (n = 26; M_age = 20.9, SD_age = 5.18). Participants completed all static (i.e., CFPT-Identity, CFPT-Angry, CFPT-Happy) and dynamic audio-visual tasks (i.e. RAVDESS-Happy, RAVDESS-Angry). Tasks were blocked by stimulus type, all CFPT tasks were presented together and all RAVDESS tasks were presented together with task order randomized within each block. CFPT tasks were always completed prior to RAVDESS tasks because interweaving the tasks may confound the replication. The independent variable was stimulation group: active tRNS or sham tRNS. The dependent variables were accuracy on CFPT and RAVDESS tasks, response time on RAVDESS tasks and mu-ERD during RAVDESS tasks.

Transcranial random noise stimulation

The Starstim 20 (Neuroelectrics, Barcelona, Spain) was used with two NG Pistim electrodes (Ag/AgCl electrode, 12 mm diameter, π cm² circular contact area) to administer tRNS through the NIC software (version 2.0.11.1). Following Penton et al. (2017), stimulation was applied to the IFC at F7 and F8 (International 10-10 system) for 20 min (plus 15 s fade-in and 15 s fade-out). The active tRNS group received 1 mA for the entire 20 min, while the sham group received 1 mA for only 15 s to mimic the sensation of stimulation without neurophysiological effects (Ambrus et al. 2010), effectively blinding participants as stimulation is most noticeable in the first 30 s (Gandiga et al. 2006). All stimulation electrodes had an impedance below 15 kΩ before stimulation began. Due to device limitations, our tRNS frequency ranged from 101 to 500 Hz, a slightly constricted range compared to 101–640 Hz used by Penton et al. (2017).

EEG recording

Neuroelectric data were recorded using Starstim 20 and NIC software on a MacBook Air (2019, Apple, macOS Mojave version 10.14.6) at 500 Hz. Eighteen NG Geltrodes (Ag/AgCl sintered pellets, 4 mm diameter, about 1 cm² contact area) were placed across the scalp to record brain activity at F7, F3, Fz, F4, F8, T7, C3, Cz, C4, T8, P7, P3, Pz, P4, P8 O1, Oz, O2 (International 10-10 system). All scalp electrodes had an impedance below 20 kΩ before recording commenced. Two electrodes recorded horizontal and vertical right eye movements. A dual reference electrode clipped onto the right earlobe connected the common mode sensor (CMS) and driven right leg (DRL).

Behavioral tasks

Visual stimuli were presented on a 1080p monitor 59 cm away from the participant and auditory stimuli played at 60.30 dB SPL from a loudspeaker positioned 99 cm in front of the participant and 127 cm above the ground.

Cambridge face perception tests (CFPTs)

Three versions of the CFPTs were used to assess facial identity perception (CFPT-Identity) and facial emotion perception for angry (CFPT-Angry) and happy faces (CFPT-Happy). Participants sorted six static images from most to least similar to a target face (CFPT-Identity with 16 trials) or from most to least intense expression of a given emotion (CFPT-Happy and CFPT-Angry with 10 trials each). Each task also included one practice trial. See Fig. 1 and Penton et al. (2017) for details and scoring procedure.

Figure 1.

A practice trial from the CFPT-Angry task (top) and the CFPT-Identity task (bottom). In each trial, the upper left corner displayed the trial number and the upper right corner displayed the trial timer.

RAVDESS stimuli and stimulus selection process

The full RAVDESS database contains audio, visual, and audio-visual clips of 24 professional actors (12 female) singing and speaking statements in various emotions at two intensity levels (normal and strong). The spoken statements are “Dogs are sitting by the door” and “Kids are talking by the door.” In a validation study, 247 participants rated the clips on emotional validity, intensity, and genuineness. The results reflect emotional validity, moderate-to-substantial interrater reliability and high test-retest reliability (Livingstone and Russo 2018). The present study used a subset of happy and angry speech audio-visual clips based on intensity and genuineness ratings from the validation study to create clip pairs for each emotion separately. See supplementary material for stimulus selection details.

RAVDESS tasks

RAVDESS-Angry and RAVDESS-Happy were computer tasks containing only angry or happy stimuli, respectively. Both tasks contain the same 15 actors (8 female), one practice, and 14 experimental trials. Trials began with a 1000 ms fixation cross, followed by two sequential dynamic audio-visual clips (i.e. a pair from the final stimulus set) of an actor expressing an emotion. Participants had four seconds to select the clip that displayed the emotion more intensely using the keyboard (‘z’ for the first video, ‘c’ for the second video; see Fig. 2).

Figure 2.

Schematic of a trial from the RAVDESS-Angry task. In frames 2 and 3, a RAVDESS clip is played with an approximate duration of 3000 ms per clip. The final frame depicts a four-second timer in the center of the screen, indicating the time left to provide a key response.

Details of both tasks can be downloaded from Open Science Framework: https://doi.org/10.17605/OSF.IO/5C6JK.

Accuracy was defined as the percentage of trials correctly selected as the more emotionally intense clip based on average ratings provided by Livingstone and Russo (2018). Response time was defined as the time from the end of the second clip to the keypress. Trials with no response were scored as “incorrect” with response time of four seconds.

Experimental procedure

Participants provided informed consent and demographic information (i.e., sex and age), were fitted with an EEG cap (International 10-10 system) and randomly assigned to either active or sham tRNS for 20 min and 30 sec. During stimulation, participants sat quietly with the experimenter nearby. Afterward, in a double-walled sound-attenuated chamber (Industrial Acoustics Corp., Bronx, NY), participants completed the three CFPT tasks (randomized order), had a three-minute break and finished with the RAVDESS tasks (randomized order). The CFPT tasks preceded the RAVDESS tasks to prioritize the replication and validation portion, but even 10 minutes of tRNS can increase cortical excitability for up to 60 minutes following stimulation, covering the duration of all tasks (Terney et al. 2008). Before each task, the experimenter reviewed the task instructions and remained present during practice trials to address questions. EEG was recorded throughout the experiment following stimulation. Upon completion, participants were debriefed, informed of their assigned condition and monitored for five minutes for any adverse effects.

EEG data pre-processing

The EEG data were processed with MATLAB version R2023a using the EEGLAB toolbox (Delorme and Makeig 2004). A high-pass filter at 1 Hz and a low-pass filter at 50 Hz were applied to the data to remove linear trends and line noise, respectively. Scalp and external channels were identified as noisy using pop_rejchan() with the threshold at 5 standard deviations, compute kurtosis for each channel and normalization on. Noisy scalp channels were interpolated while noisy external channels were rejected. The data were then re-referenced to the average excluding the external channels.

The data were epoched with windows of 1000 ms pre-stimulus onset and 6000 ms post-stimulus onset. Baseline correction (removal) was done using the pre-stimulus onset epoch as the baseline to compare the change in band power post-stimulus onset. Independent components analysis (ICA) with the extended option and pca option was applied to identify eye components. Based on the ICA, noisy epochs were automatically detected and rejected using pop_autoreject(). Multiple rounds of ICA help refine the algorithm, so a second round of ICA was applied to identify eye components. Eye components result from eye movements and blinks which add noise to the data; eye components identified by ICA with 90% confidence were removed.

The data were then used to calculate event-related spectral perturbation (ERSP) values, representing the average event-related changes in spectral power at a given time and frequency. The ERSP values of interest were those occurring from 0 to 6000 ms post-stimulus onset in the 8–13 Hz frequency range for electrodes C3, Cz, and C4.

Statistical analyses

Analyses were conducted in R version 4.3.1 with packages: “tidyverse” version 2.0.0 and “rstatix” version 0.7.2. Following Penton et al. (2017), we collapsed across emotion and conducted t-tests to assess the between-groups effect of stimulation on CFPT task accuracy, RAVDESS task accuracy, RAVDESS response time and mu-ERD during the RAVDESS tasks. Welch’s t-test was used as it is equivalent to an independent samples t-test for groups with equal variances but more reliable in cases of unequal group variances.

T-tests were one-tailed (α = .05), aligned with the pre-registered hypotheses except the t-test on CFPT-Identity performance was two-tailed since we did not specify directionality a priori. Univariate outliers were removed on all dependent variables for each group independently; a data point was considered an outlier if it fell outside the whiskers of a boxplot (Wilcox and Rousselet 2018).

Results

Stimulation validity check

There was a significant and small-to-medium main effect of group on the CFPT-Emotion task with the active tRNS group (M = 82.40, SD = 4.73) outperforming the sham tRNS group (M = 79.60, SD = 6.51), t(45.63) = −1.74, P = .04, d = −0.49, 95% CI [−Inf, −0.10]. The active tRNS (M = 66.30, SD = 4.94) and sham tRNS group (M = 64.30, SD = 9.17) did not significantly differ on their accuracy for the control task (CFPT-Identity), t(38.41) = −0.99, P = .33, d = −0.28, 95% CI [−6.16, 2.10]; see Fig. 3.

Figure 3.

Effect of tRNS on performance in the CFPT-Emotion task and CFPT-Identity task. Dotted line indicates chance performance. Error bars represent the standard error.

RAVDESS task performance

There was a small marginally significant effect of group on accuracy for the RAVDESS task between the active tRNS (M = 71.60, SD = 8.09) and sham tRNS group (M = 67.90, SD = 8.50), t(45.94) = −1.56, P = .06, d = −0.45, 95% CI [−Inf, 0.29]; see Fig. 4. A significant medium effect of group on response time was found with the active tRNS group (M = 763.00, SD = 181.00) responding faster than the sham tRNS group (M = 909.00, SD = 342.00), t(37.70) = 1.84, P = .04, d = 0.53, 95% CI [12.27, Inf]; see Fig. 5. See supplementary material for results with a combined speed-accuracy measure.

Figure 4.

Effect of tRNS on performance across RAVDESS tasks. Dotted line indicates chance performance. Error bars represent the standard error.

Figure 5.

Effect of tRNS on response time across RAVDESS tasks. Error bars represent the standard error.

Mu-ERD during RAVDESS task

There was a significant medium effect of stimulation group on power in the mu-rhythm relative to baseline such that the active tRNS group (M = −0.90, SD = 1.71) showed greater mu-ERD than the sham tRNS group (M = 0.14, SD = 1.65), t(41.36) = 2.05, P = .02, d = 0.62, 95% CI [0.19, Inf]; see Fig. 6.

Figure 6.

Effect of tRNS on the event-related desynchronization of the mu-rhythm across the RAVDESS tasks. Lower values indicate greater mu-ERD. Error bars represent the standard error.

Discussion

The results of the current investigation replicated Penton et al. (2017) such that active tRNS to the IFC improved performance on the emotion perception task compared to sham tRNS. In addition, there was no group difference on the control task, which validates the stimulation protocol. We also extended Penton et al. (2017) from static stimuli to dynamic audio-visual portrayals of emotion. When perceiving emotions from dynamic audio-visual stimuli, active tRNS to the IFC led to significantly more mu-ERD, faster response times, and marginally better accuracy than sham tRNS, suggesting an enhanced embodied response to dynamic audio-visual stimuli.

Active tRNS to the IFC may have led to faster response times because embodied cognition is a fast-acting system. Stel and Van Knippenberg (2008) proposed dual routes to emotion perception: a slow route involving cognitive matching of visual stimuli to stored knowledge, and a fast route relying on proprioceptive cues and embodied cognition. Facial mimicry, another hMNS index supporting the fast route, influences emotion recognition speed more than accuracy. Lydon and Nixon (2014) found that disrupting mimicry by biting a stick slowed response times without affecting accuracy. Similarly, Livingstone et al. (2016) recorded facial mimicry via zygomatic activity (indicative of smiling) during a RAVDESS emotion recognition task. In persons with PD—a condition known to also impair mu-ERD (Heida et al. 2014)—recognizing positive emotions more quickly was moderately correlated with more zygomatic activity, demonstrating a motor-behavior link. Importantly, persons with PD and controls did not differ on recognition accuracy of positive stimuli. Likewise, among 1239 adults aged 18–76 years, autistic adults had emotion recognition accuracy comparable to controls, but slower response times (Jertberg et al. 2025). These findings suggest that hMNS impairment slows response times without significantly affecting accuracy which may explain why enhancing the hMNS via active tRNS led to significantly faster response times, but only marginally improved accuracy.

Nonetheless, active tRNS to the IFC significantly improved emotion perception accuracy for static stimuli (i.e. CFPT tasks) but only marginally for dynamic audio-visual stimuli (i.e. RAVDESS tasks). This may reflect differences in stimuli complexity and naturalism. Dynamic audio-visual stimuli provide richer emotional cues (e.g. facial expression, tone of voice, movement, and colour) than static stimuli (e.g. facial expression), making the RAVDESS stimuli less ambiguous for all participants, possibly reducing the effect of stimulating the hMNS. The stimuli could be made more ambiguous in future work by adding visual or auditory noise.

Another explanation for the significantly improved task accuracy for static stimuli, but only marginal improvement for dynamic audio-visual stimuli is that response time may be a more sensitive measure than accuracy in the RAVDESS tasks. The RAVDESS tasks use a binary response style where trials are scored as correct or incorrect whereas the CFPT allows for subtle differences in performance because trials are scored based on the degree of deviation from the correct order. While trial accuracy in RAVDESS tasks is all-or-nothing, CFPT tasks allow participants to earn partial scores.

Limitations and future directions

To align with Penton et al. (2017), this study used only angry and happy stimuli and averaged across them. However, the full RAVDESS database includes six emotions, from basic (e.g. happy, sad, angry) to more complex emotions (e.g. fear, surprise, disgust). Prior work shows that autistic adults exhibit slower response times than controls across all emotions (happy, sad, angry, fear, surprise, disgust), but autistic adults only had worse accuracy with fear (Humphreys et al. 2007). There may be emotion-specific effects of engaging the hMNS on emotion recognition. Future research should test a broader range of emotions to explore interactions between stimulation and emotion (e.g. happy, sad, angry, fear, surprise, disgust) on emotion recognition. This type of study would also provide an opportunity to explore whether the effect of tRNS on emotion recognition is modulated by the complexity of emotion. A future study may also explore an interaction between stimulation, emotion, and age given that Yang et al. (2017) found that in older adults, active tRNS to the IFC improved anger perception, but not happiness perception compared to sham.

The stimuli in the RAVDESS and CFPT tasks differ such that the RAVDESS stimuli are more ecologically valid, being dynamic instead of static and including audio-visual cues rather than visual-only cues. The observed effects of active tRNS on emotion perception in the RAVDESS task may be driven by the dynamic component, multimodal component or their combination. The full RAVDESS database includes the audio-only and visual-only recordings of the audio-visual stimuli used in this study. Future research can use the full database to isolate whether the observed effects are attributable to the dynamic or multimodal features.

A control task using dynamic audio-visual stimuli was not included as tRNS aftereffects would not last the entire experiment. Future work could include one to assess whether the tRNS effect on response time is unique to emotion perception or reflect a general decrease in response time. Although the IFC is a major node of the hMNS, it also supports other brain networks like the central executive network (CEN), implicated in cognitively demanding tasks. Since the RAVDESS tasks involve fast, simple binary responses, it should not be cognitively demanding and likely engage the hMNS more than the CEN. Future research could use EEG to index the hMNS via mu-ERD and the CEN via theta-band phase synchronization (Mizuhara and Yamaguchi 2007) to clarify whether active tRNS to the IFC uniquely engages the hMNS when perceiving dynamic audio-visual emotion.

Although we applied tRNS over the IFC and observed anticipated effects on behavior and mu-ERD, it is possible that the stimulation did not exert its effects through direct modulation of the hMNS as hypothesized. While tRNS is generally understood to produce its strongest effects locally—at or near the site of electrode placement—there is evidence that its influence can extend beyond the targeted cortical region (Datta et al. 2011, Fertonani and Miniussi 2017). Stimulation may engage anatomically or functionally connected brain regions through network-level dynamics, so it is conceivable that the observed stimulation effects were mediated, at least in part, by indirect activation of adjacent or downstream neural systems, rather than a focal modulation of the hMNS alone.

Nevertheless, we consider this alternative explanation to be unlikely for two reasons. First, our stimulation protocol was guided by a strong a priori theoretical rationale rooted in the role of the IFC in action perception and mirroring. Second, behavioral and neural results align well with existing predictions derived from literature on mirror system function. Still, given the diffuse nature of current spread in transcranial stimulation and the limitations in our ability to directly measure neural targets with high spatial resolution, we cannot fully exclude the possibility that non-hMNS regions, such as limbic regions (Rymarczyk et al. 2018, Del Vecchio et al. 2024), contributed to the observed effects.

Conclusion

The present study builds meaningfully on prior work by demonstrating that tRNS applied over the IFC enhances emotion perception not only for static facial expressions but also for more ecologically valid dynamic audio-visual portrayals of emotion. Our findings support the notion that stimulation of the IFC can amplify sensorimotor simulation during emotion recognition, reflected in both behavioral (faster response times) and neural (increased mu-ERD) markers. By extending previous work to include dynamic stimuli, incorporating electrophysiological measures, and considering both speed and accuracy, this study advances our understanding of how non-invasive brain stimulation may selectively modulate components of the human mirror neuron system. These results not only validate the efficacy of tRNS in enhancing social-emotional communication but also offer new methodological directions for probing the neural mechanisms underlying emotion perception in both typical and atypical populations.

Author contributions

Carmen Dang (Conceptualization [Equal], Data curation [Lead], Formal analysis [Lead], Investigation [Equal], Methodology [Equal], Project administration [Lead], Visualization [Equal], Writing—original draft [Lead], Writing—review & editing [Equal]), Frank A. Russo (Conceptualization [Equal], Funding acquisition [Lead], Investigation [Equal], Methodology [Equal], Resources [Lead], Supervision [Lead], Visualization [Equal], Writing—original draft [Supporting], Writing—review & editing [Equal])

Supplementary material

Supplementary material is available at SCAN online.

Conflicts of interest

The authors declare no conflicts of interest.

Funding

This work was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant awarded to F.A. Russo (Reference Number: 2017-06969).

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.