Social Modulation of Imitation in Children with Autism Spectrum Disorder: Evidence from EEG and Reciprocal Imitation Training

Abstract

Imitation is crucial for social learning, yet children with autism spectrum disorder (ASD) often show atypical imitation abilities. To probe the neural dynamics that precede overt imitation, electroencephalography (EEG)—with a focus on α (8–12 Hz) and β (13–30 Hz) activity commonly linked to action observation and sensorimotor processing—was used to index pre-imitation processing in preschool-aged children with ASD. Grounded in the social motivation framework, this study combined an EEG experiment and a naturalistic behavioral intervention. In Study 1, 11 preschool children with ASD completed an action-observation (pre-imitation) task under low- versus high-sociality video conditions. Time–frequency and spectral analyses were conducted to compare α- and β-band responses across conditions. In Study 2, four children received a six-week Reciprocal Imitation Training (RIT) program, and imitation and social-communication outcomes were assessed pre-, mid-, and post-intervention. The results showed that low-sociality stimuli elicited stronger frontal and prefrontal power increases in both α and β bands, whereas high-sociality stimuli elicited more temporally dynamic β-band responses but with lower overall power engagement. Although inferential support was limited by sample size, behavioral trends suggested improvements following RIT in imitation and related social functioning, with larger gains in children with mild-to-moderate ASD. Together, these findings suggest that social context modulates pre-imitation neural activity in ASD and that socially grounded imitation training may support broader social development.

1. Introduction

Imitation is a core capacity in human cognition and social development, permeating multiple stages such as language acquisition, skill learning, and interpersonal communication. It involves complex psychological processes including information processing, intention understanding, and action reproduction [1]. Beyond its cognitive function in learning, imitation also plays a crucial role in fostering interpersonal relationships and social interaction. Previous studies have shown that children with Autism Spectrum Disorder (ASD) generally exhibit atypicalities in imitation [2]. Autism Spectrum Disorder (ASD) is a pervasive developmental disorder characterized by deficits in social interaction, stereotyped behaviors, language abnormalities, and atypical visual abilities [3,4,5]. Social communication impairments may affect the cognitive processes involved in imitation (DSM-5). In addition to imitation difficulties, children with ASD often present atypical behaviors such as impairments in social communication, reduced joint attention, repetitive and restricted behaviors, and difficulties in emotional reciprocity, all of which may further constrain their opportunities for social learning through imitation. These deficits may impair their social and emotional development from an early age [6,7]. Although a substantial body of empirical research has demonstrated that children with ASD perform poorly in areas such as spontaneous imitation and fine-motor imitation [8,9], research findings are not entirely consistent. Such discrepancies may stem from the lack of a unified definition of imitation as well as variations in experimental paradigms. Traditional explanatory frameworks, such as the mirror neuron system theory and theory of mind, remain insufficient to fully account for these phenomena. It should be noted that while these theories provide important insights, their explanatory power remains debated in the current literature, and further discussion of these theoretical limitations is provided in the subsequent section. Consequently, some scholars argue that imitation deficits in ASD should be understood from the perspective of the social functions of imitation [10].

At present, there is still a lack of specific therapeutic methods for ASD, with rehabilitation education and training remaining the main forms of intervention. Importantly, this does not imply that ASD interventions are ineffective; rather, there is no single curative treatment, and evidence-based behavioral and developmental approaches—such as Applied Behavior Analysis (ABA) and Naturalistic Developmental Behavioral Interventions (NDBIs)—are widely used to support social, communicative, and adaptive functioning in children with ASD.

Among them, Reciprocal Imitation Training (RIT), a Naturalistic Developmental Behavioral Intervention (NDBI), has been shown to effectively improve imitation skills in children with ASD [11]. Moreover, parent-mediated CI-RIT has demonstrated certain feasibility within early intervention systems [12]. However, existing research has predominantly focused on action imitation, while relatively little attention has been given to social imitation, which is closely tied to social interaction. Furthermore, there is a scarcity of systematic investigations into the neural basis of imitation from the perspective of cognitive neuroscience.

In the field of electroencephalography (EEG), studies have indicated that children with ASD often exhibit abnormal background activity and epileptiform discharges [13]. In addition, EEG abnormalities are correlated with certain clinical manifestations of ASD [14]. Beyond these clinical markers, EEG has been widely applied to investigate the temporal dynamics of action observation and social information processing in ASD, including atypical modulation of sensorimotor rhythms (e.g., mu/alpha suppression) and differences in beta-band synchronization during socially relevant tasks. These neurophysiological findings suggest that EEG can provide a sensitive window into the neural mechanisms underlying imitation and its social modulation. These findings provide methodological support for examining the initiation of social imitation in children with ASD using EEG indicators.

Drawing upon both behavioral and cognitive neuroscience approaches, the present study focuses on the pre-imitation (action observation) stage, which occurs before the actual execution of imitation, by analyzing EEG features associated with social cues. Specifically, this study addresses the following research questions:

(1) Do social factors influencing imitation deficits in children with ASD have a neural basis? Are there significant differences in neural activation between imitation under high-social and low-social conditions?

(2) Can RIT, grounded in the principles of NDBI, enhance social imitation in children with ASD and improve their sensitivity to social information?

2. Literature Review

2.1. Imitation, Autism, and Imitation Deficits

Imitation is a complex process, and its definition has been approached from different perspectives, which can be broadly categorized into two views. One emphasizes imitation as the replication of model behaviors, while the other stresses that imitation also involves understanding the model’s intentions. The first perspective suggests that imitation allows individuals to acquire new skills by observing others’ behaviors. Thorpe proposed three defining features of imitation: the target behavior is novel, the imitator could not previously perform it, and the behavior is not innate [15]. The second perspective highlights the functions of imitation, which include demonstration, inhibition–disinhibition, response facilitation, and ripple effects [16]. More recently, Galli reported that in a motor imagery task [17], children with ASD and typically developing peers did not differ significantly in their ability to understand intentions or imagine goal-directed actions, providing new empirical support for the view emphasizing intentional understanding. Specifically, in this paradigm, children were asked to mentally simulate object-directed hand actions without executing overt movements, allowing researchers to assess intention-related motor representation independent of actual motor performance.

A meta-analysis by Edwards revealed that children with ASD demonstrated only 27% of the imitation abilities of control groups [18], with no moderating effects of age, gender, or task modality. Among the earliest and most extensively studied forms of imitation is action imitation, in which children with ASD show deficits in action representation [9,19,20]. Such deficits appear to be general, specific (e.g., mirror imitation), and persistent. Some studies have suggested that object-related imitation is less impaired [21]. Moreover, children with ASD make more errors when imitating meaningless gestures [22], whereas their performance is relatively better when imitation involves clear intentions or functional significance [23]. It is important to acknowledge that ASD is highly heterogeneous, and imitation difficulties may vary substantially across individuals, overlapping with impairments observed in other neurodevelopmental conditions. Therefore, imitation deficits should be interpreted as probabilistic patterns rather than uniform diagnostic features.

Two mainstream theoretical frameworks have been proposed to account for imitation deficits: the mirror neuron system (MNS) theory and the social motivation theory. In addition, theory of mind (ToM) accounts have also been discussed in the literature, particularly in relation to intention understanding and social-cognitive inference, and these perspectives are briefly considered alongside MNS explanations.

The mirror neuron system theory attempts to explain imitation deficits from a neurobiological perspective. The MNS is considered the neural basis of imitation [24]. Both animal and human studies have shown that when observing an action, relevant brain regions exhibit neural activity similar to that generated during action execution [25]. Broken-Mirror Theory (BMT) posits that imitation deficits in children with ASD result from MNS dysfunction, thereby impairing language, theory of mind, empathy, intention understanding, and imitation abilities [26]. Evidence from EEG and fMRI studies indicates that during action observation tasks, children with ASD show reduced activation in the motor cortex and other MNS-related regions compared to controls [27,28].

However, the MNS theory remains controversial. Some studies have found that children with ASD perform normally in meaningful or object-related action observation tasks, suggesting that standard imitation tests may not directly capture true imitation ability [29,30]. Furthermore, imitation, empathy, and theory of mind reasoning may rely on a broader “social brain” network rather than solely the MNS [31,32].

The social motivation theory argues that reduced early social motivation in children with ASD leads to diminished sensitivity to the rewarding value of social stimuli, which subsequently hampers social attention and the accumulation of social experiences [33,34]. Such motivational deficits not only weaken social competence but may also contribute to imitation impairments [35]. Experimental studies have shown that under external sensory or incentive conditions, children with ASD are still capable of completing action observation tasks, suggesting that their potential for imitation exists but lacks spontaneous social drive. For example, Ingersoll [36] reported that when action observation tasks were paired with explicit reinforcement or salient sensory cues, children with ASD were able to reproduce modeled actions at levels comparable to controls, indicating that performance may depend on motivational context rather than absolute incapacity. Nejati, through a systematic review, further noted that the transfer effects of social training (e.g., imitation training, social skills interventions) are significantly influenced by age, severity, duration of intervention, and environmental factors [37].

Neurophysiological investigations of imitation deficits have primarily relied on methods such as fMRI, EEG, and ERP, each with distinct strengths and limitations. fMRI provides high spatial resolution for identifying cortical networks involved in action observation, whereas EEG/ERP offer superior temporal resolution for tracking rapid neural dynamics during social perception and imitation processing, although they are more limited in precise anatomical localization.

fMRI studies have demonstrated that children with ASD exhibit atypical activation in brain regions such as the inferior frontal gyrus (IFG), superior temporal sulcus (STS), inferior parietal lobule (IPL), and premotor cortex (PMC) during imitation and action observation tasks [38]. EEG studies have further shown that children with ASD display reduced suppression of the mu rhythm while observing actions, indicating MNS dysfunction [39]. Event-related potential (ERP) studies have revealed differences in components such as P1, N1, N190, and P3 in children with ASD during detailed processing and social action observation tasks [40], suggesting that external stimuli and social factors may modulate imitation-related brain activity. In addition, microstate analyses have uncovered significant differences in overall brain activity patterns between children with ASD and controls [41]. Recent work further suggests that EEG-based pattern analysis may help characterize subtypes within ASD and potentially reduce heterogeneity by identifying distinct neurophysiological profiles [42], an emerging direction that strengthens the contribution of neural markers to ASD research.

Regarding the association between EEG and clinical features, Ruffini found that EEG abnormalities in children with ASD—such as altered background spectral power and atypical fluctuations—were significantly correlated with phenotypes including language impairments, social deficits, and repetitive behaviors. In this large-scale study of children spanning early childhood to adolescence, Ruffini et al. reported moderate correlations between spectral power alterations and symptom severity across language and social domains, highlighting that EEG markers may track broader emotional and behavioral difficulties in ASD [14]. These findings suggest that EEG markers may serve as potential biological indicators of differences in imitation and social competence.

In sum, imitation deficits in children with ASD may result from the interaction between atypical neural mechanisms and reduced social motivation.

2.2. The Relationship Between Imitative Behavior and Sociality

Imitation deficits in children with ASD are not uniform across different types of imitation. Certain tasks, such as object manipulation and gesture imitation, are not significantly impaired [19]. Since imitation involves visual information processing and attentional mechanisms, deficits in joint attention—a hallmark of ASD—negatively affect socially oriented imitation abilities. Researchers often distinguish between cognitively oriented imitation (instrumental, goal-directed) and socially oriented imitation (over-imitation, immediate imitation), with children with ASD performing more poorly on the latter [1].

Low-social imitation refers to imitation aimed primarily at reproducing actions or goals without involving explicit social functions, such as action imitation, object-related imitation, and meaningful imitation. Action imitation ranges from reproducing localized movements to replicating entire action sequences. In children with ASD, this is typically manifested as uncoordinated movements, slower execution, reliance on visual control, and increased tension and fatigue. Depending on the type of content, action imitation can be categorized into object manipulation, hand movements, oral–facial actions, and vocal imitation, which together provide a framework for investigating imitation deficits in ASD. Importantly, meaningless imitation tasks—designed to exclude the influence of semantic understanding or object involvement—remain a key paradigm for testing action reproduction ability. Recent advances in the development of assessment tools and automated evaluation methods have provided methodological improvements for more precise measurement of these tasks [43].

High-social imitation refers to socially oriented imitation, including synchronous imitation, over-imitation, and imitation of sounds or facial expressions. Synchronous imitation, which requires the imitator to perform actions simultaneously with the model, is closely associated with social motivation [44]. Over-imitation refers to copying extraneous actions irrelevant to the task itself and has been interpreted as reflecting social affiliation motives [45]. Children with ASD exhibit significantly fewer instances of over-imitation compared to typically developing peers [46], suggesting that this atypicality is related to diminished social motivation.

2.3. Research on RIT Based on NDBI

Imitation is a fundamental basis of learning, and imitation interventions for children with ASD are designed to foster spontaneous imitation abilities. Such interventions often stimulate motivation through children’s interests and follow a gradual progression: moving from gross motor actions to fine-motor movements, facial expressions, oral-facial gestures, and speech sounds, with reinforcement strategies to facilitate generalization. Imitation training is typically embedded within broader intervention models, such as SCERTS (imitation of expressions, gestures, and language) and PRT (imitation of play, action sequences, and verbal expression). A systematic review of SCERTS-based interventions by Yi reported moderate to large improvements in language expression and social interaction, particularly in components involving imitation of expressions and gestures [47]. Although traditional ABA-based interventions are widely supported as effective, concerns have been raised regarding challenges, in general, ization and spontaneity when training is delivered in highly structured formats, which motivates the development of more naturalistic approaches [48].

Naturalistic Developmental Behavioral Interventions (NDBIs) emerged from the integration of applied behavioral principles (e.g., ABA) with developmental science, aiming to address early social-communication difficulties in young children with ASD within natural everyday contexts. Unlike highly structured therapist-led formats, NDBIs emphasize child-initiated interaction, shared control, and the use of naturally occurring reinforcement, thereby promoting both skill acquisition and generalization. This approach has become a widely adopted framework in early ASD intervention due to its strong developmental rationale and ecological validity.

NDBI integrate principles from behavioral and developmental sciences, implemented in naturalistic settings. They emphasize shared control between the child and therapist, using natural events and behavioral strategies to teach developmentally appropriate skills. Representative approaches include Incidental Teaching (IT), Pivotal Response Training (PRT), the Early Start Denver Model (ESDM), Joint Attention, Symbolic Play, Engagement, and Regulation (JASPER), and Reciprocal Imitation Training (RIT).

RIT was specifically developed to address the limitations of traditional imitation training—namely, the lack of generalization in highly structured settings, reliance on reinforcement, and training of isolated skills. RIT emphasizes naturalistic interactions, reciprocal imitation, and skill generalization, and its five-phase implementation includes: adaptation, familiar action modeling, novel action modeling, modeling with different toys, supplemented by conditional imitation and language mapping, all designed to promote spontaneous imitation and enhance social interaction.

Research on the effectiveness of NDBIs has demonstrated significant benefits in improving core ASD symptoms, social engagement, expressive language, and play skills [49,50,51]. These outcomes have typically been evaluated using standardized developmental and diagnostic instruments (e.g., ADOS, Vineland Adaptive Behavior Scales, and language measures), with meta-analytic evidence indicating small-to-moderate effect sizes in social-communication domains, particularly for preschool children receiving early intervention.

Studies on RIT in particular have shown that it significantly enhances both spontaneous imitation and social interaction skills [52], and it can be delivered effectively by therapists, teachers, and parents [53]. However, empirical research on RIT in China remains scarce, and further studies are needed to support its localization and broader implementation.

3. Methodology

3.1. Ethics Statement

All procedures performed in this study complied with the ethical standards of the 1964 Declaration of Helsinki and its later amendments, or with comparable ethical standards. The study protocol and instruments were reviewed and approved by the appropriate ethics committee. Written informed consent was obtained from the parents of all participants prior to their inclusion in the study. In addition to parental consent, verbal assent was obtained from each child whenever developmentally appropriate. Before the EEG setup and experimental tasks began, the researchers explained the procedures using child-friendly language and simple visual demonstrations. Participation was initiated only when the child appeared comfortable and willing to proceed. Given that preschool children with ASD may experience sensory sensitivities, the research team continuously monitored signs of discomfort (e.g., agitation, avoidance behaviors, distress reactions) during the sessions. Short breaks were provided whenever necessary, and the experiment was discontinued immediately if a child could not tolerate the procedure or showed sustained distress. No adverse events were reported. To improve clarity and reproducibility, the overall workflow of Study 1a (EEG experiment) and Study 1b (RIT intervention) is summarized in Figure 1.

3.2. Study 1a: EEG Experiment (RQ1)

3.2.1. Participants

To address Research Question 1, participants were preschool children diagnosed with Autism Spectrum Disorder (ASD). A total of 15 children were recruited through a local rehabilitation hospital.

Participants were enrolled through clinician referral and posted recruitment invitations at the hospital. Eligible children were screened by licensed clinicians based on medical records and standardized diagnostic assessments. Due to the clinical nature of the population, participant enrollment followed a convenience sampling approach rather than true random sampling. This recruitment procedure is now reported to improve transparency and reproducibility.

All children met DSM-5 diagnostic criteria and exceeded the diagnostic threshold on the Childhood Autism Rating Scale (CARS). ASD symptom severity was characterized using CARS total scores, with higher scores indicating more severe symptom presentation. In line with commonly used clinical cutoffs, scores between 30 and 36.5 were considered indicative of mild-to-moderate autism, whereas scores above 36.5 reflected more severe symptomatology. Given the small sample size, severity categories were reported descriptively rather than used for subgroup-level statistical inference. During the experimental process, two participants were unable to complete the task calmly, and EEG data from another two participants were of insufficient quality. Consequently, valid EEG data were collected from 11 participants (2 females and 9 males). The observed male–female imbalance is consistent with the higher prevalence of ASD in males reported in epidemiological studies. Nevertheless, we acknowledge that the limited representation of females restricts the generalizability of gender-related interpretations, which is noted as a limitation. All participants were screened to exclude organic diseases and other psychiatric histories, and none had taken psychotropic medication recently.

Information regarding cognitive functioning and language ability was obtained from available clinical records and therapist reports. However, standardized developmental quotient or IQ assessments were not consistently available for all participants, which represents an additional limitation of the current pilot sample. Future studies will incorporate comprehensive developmental and cognitive profiling to better characterize participant heterogeneity.

Given clinical recruitment constraints, a sensitivity power analysis (G*Power 3.1.9.7. framework; matched-pairs design, two-tailed $\alpha$ = 0.05) was conducted to clarify the detectable effect sizes with the available sample. With n = 11, the design has 80% power to detect effects of approximately $d_z\approx 0.94$, indicating that the present EEG sample is primarily sensitive to large within-subject effects.

3.2.2. EEG Instruments and Recording

This study employed the Emotiv EPOC+ wireless EEG system, which follows the international 10–20 electrode placement system and includes 14 channels (AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4) along with two mastoid reference electrodes (CMS/DRL). The device and electrode distribution are shown in Figure 2. Continuous sampling was performed at a rate of 128 Hz, with a bandwidth of 0.2–43 Hz. The portability and comfort of this system make it suitable for preschool ASD populations.

We acknowledge that the Emotiv EPOC+ system provides lower spatial resolution and signal quality compared with high-density clinical EEG systems. However, its wireless and child-friendly design has been widely adopted in developmental and special-needs populations, where participant tolerance and movement constraints are critical. In the present study, the system was selected to maximize feasibility and compliance in preschool children with ASD. Nevertheless, the limited number of channels and reduced spatial precision are considered important methodological limitations and are discussed accordingly.

3.2.3. Experimental Design and Procedure

The EEG experiment was conducted in a quiet, distraction-free room. Participants were seated comfortably in front of a computer used for stimulus presentation, and the experimenter assisted the child in wearing the EEG device. The task was initiated only after electrode contact quality and signal quality met the system’s requirements. A representative recording setup and the experimental environment are shown in Figure 3.

The experimental paradigm consisted of three instruction phases interleaved with baseline and stimulus presentation. First, participants received an initial warm-up instruction (Instruction 1) to help them relax and become familiar with the experimental setting. This was follows: by a one-minute baseline recording during which a short “dolphin breathing” video was presented to capture resting-state activity while maintaining the child’s attention on the screen. After baseline, Instruction 2 was delivered, asking the child to observe the upcoming actions in the videos and mentally simulate them without performing any overt movements.

Stimulus operationalization and validation. The core manipulation in Study 1a was the sociality level of the action-observation stimuli. Specifically, Videos 1–2 constituted the low-sociality condition and were derived from a child mindfulness exercise video, in which the actor primarily displayed gross motor movements (e.g., raising arms, swaying, rotating arms) with minimal social cues. In this condition, the stimuli were defined a priori as “low sociality” because they contained (i) no spoken language or vocal interaction, (ii) minimal or neutral facial expression, (iii) no explicit gaze-to-viewer engagement or communicative gestures, and (iv) a single agent with a relatively simple action context. Videos 3–4 constituted the high-sociality condition and were derived from a children’s gesture-dance video, in which the actor’s movements were embedded in richer social signals, including (i) explicit facial expressions, (ii) speech or rhythmic vocalization, (iii) gaze engagement and socially directed gestures, and (iv) a more socially communicative presentation context. Representative screenshots of the two conditions are provided to support transparency (low sociality vs. high sociality; see the examples shown in the figures in Section 4).

To provide an objective basis for this manipulation, we additionally conducted a stimulus validation procedure prior to formal analyses. Two independent raters (not involved in data collection) viewed each video segment and scored perceived sociality on a 5-point Likert scale (1 = very low social content; 5 = very high social content), while also coding the presence/absence of key social cues (face visibility, direct gaze, speech/vocalization, and communicative gestures). Inter-rater agreement was evaluated, and the high-sociality videos were consistently rated higher than the low-sociality videos across raters, supporting the intended manipulation.

To reduce potential confounds from low-level stimulus properties, we attempted to keep the two conditions comparable, in general, format (child-friendly videos presented on the same screen and with the same duration per clip). Moreover, both conditions involved whole-body movement and similar viewing distance and resolution. We acknowledge that perfect matching of low-level properties (e.g., motion energy, visual complexity, or novelty) is challenging when manipulating sociality in naturalistic stimuli; therefore, we report these operational criteria and validation scores to allow readers to evaluate the contrast transparently.

Participants then watched Videos 1 and 2 under this pre-imitation (action observation) stage. Subsequently, Instruction 3 introduced the next stimulus sequence, again emphasizing mental imagery of the observed actions without execution. Each video was separated by a 1.5 s inter-stimulus interval.

Presentation order and counterbalancing. To control for potential order effects, the presentation order of the two sociality conditions (low-sociality first vs. high-sociality first) was counterbalanced across participants using a two-sequence design. Within each condition, the two videos were presented in a fixed order to maintain narrative continuity, and each video was separated by a 1.5 s inter-stimulus interval.

Throughout the experiment, children were instructed to remain as still as possible to minimize movement-related artifacts.EEG recording ended simultaneously with the conclusion of the final video stimulus. After completion, the child’s scalp was gently cleaned with a damp cloth to remove any residual saline solution from the electrodes, ensuring that participants experienced no physical discomfort as a result of the procedure.

3.2.4. EEG Preprocessing and Artifact Control

EEG preprocessing was conducted using standard procedures to ensure data quality, particularly given the high likelihood of motion artifacts in preschool children with ASD. Raw EEG signals were first band-pass filtered between 0.5 and 40 Hz to remove slow drifts and high-frequency noise. A notch filter was applied when necessary to reduce power-line interference. Artifact-contaminated segments were identified through a combination of automatic thresholding and visual inspection. Epochs containing excessive amplitudes (e.g., >±100 μV), abrupt movement-related fluctuations, or muscle artifacts were excluded from further analysis.

Bad channels were detected based on abnormal variance or poor contact quality and were interpolated using spherical interpolation when appropriate. EEG data were referenced using the CMS/DRL mastoid reference scheme provided by the Emotiv system.

To specifically address motion artifacts, children were instructed to remain as still as possible, and the experimenter monitored movement continuously throughout the recording. Short breaks were provided when restlessness occurred, and trials with visible head or body movements were removed during preprocessing.

Quantitatively, across the final sample (n = 11), an average of approximately 15–20% of epochs per participant were rejected due to motion or other artifacts, leaving sufficient artifact-free trials for spectral and time–frequency analyses. These quality-control procedures were implemented to enhance the reliability of the EEG findings.

3.3. Study 1b: RIT Intervention Study (RQ2)

3.3.1. Participants

To address Research Question 2, a subset of four children was selected from the original EEG sample to participate in the Reciprocal Imitation Training (RIT) intervention study. This intervention component was designed as an exploratory pilot study aimed at examining the feasibility and potential behavioral trends associated with socially grounded imitation training in preschool children with ASD. Participants were chosen based on therapist and parent reports indicating marked difficulties in spontaneous imitation during daily interactions, as well as observable challenges in social engagement and reciprocal play behaviors. Only children who were able to attend the full six-week intervention schedule and complete repeated behavioral assessments were included. All selected participants met DSM-5 diagnostic criteria for ASD and exceeded the diagnostic threshold on the Childhood Autism Rating Scale (CARS), ensuring diagnostic consistency with the EEG cohort. Participant characteristics for Study 1b are reported in

Table 1. Participant Characteristics in Study 1b (RIT Intervention).

Participant	Age (Years)	CARS	MIS (Motor Imitation Scale)	SRS (Social Responsiveness Scale)
WY	5	44	14	137
WSY	3.5	37	25	130
DSK	4	46	16	145
XX	5	38	20	126

.

For the pilot RIT component (n = 4), sensitivity analysis under a matched-pairs framework indicates that only extremely large effects ($d_z\approx 2.13$) would be detectable with 80% power at $\alpha$ = 0.05 (two-tailed). Therefore, this component is intended to provide feasibility information and effect size estimates rather than definitive hypothesis testing.

3.3.2. Intervention Design and Procedure

Participants underwent Reciprocal Imitation Training (RIT) over a six-week period. Each child received a total of three hours of intervention per week, delivered in three 20 min sessions. The intervention was implemented by a licensed therapist with formal training and clinical experience in early ASD behavioral interventions. To evaluate changes in imitation and related social-communication abilities, behavioral assessments were conducted at three time points: prior to the intervention (pre-intervention), midway through the program (week 4), and immediately after completion (week 6). Outcome measures included imitation ability, social sensitivity, spontaneous language, stereotyped behaviors, joint attention, and play skills. To reduce potential observer bias, behavioral coding and outcome assessments were completed by an independent assessor who was blinded to the intervention phase (pre/mid/post).

The RIT intervention was implemented in three successive phases. During the baseline phase (week 1), sessions primarily consisted of free-play interactions between the therapist and the child, with the goal of establishing rapport and identifying the child’s spontaneous imitation level, and no structured teaching strategies were introduced at this stage. The intervention phase was carried out from weeks 2 to 6 and focused on reciprocal imitation exchanges designed to promote generalization and spontaneous imitation, with training progressing gradually from familiar toys and actions to novel actions and varied play contexts. Finally, a maintenance phase was conducted in week 7, during which the therapist interacted with the child under the same schedule but without applying specific intervention strategies, and post-intervention assessments were administered to examine whether gains in imitation and social behaviors were sustained beyond the active training period.

3.3.3. Outcome Measures and Behavioral Coding

To evaluate behavioral changes associated with the RIT intervention, outcome assessments were conducted at three time points (pre-intervention, mid-intervention, and post-intervention). Measures included standardized rating scales as well as structured behavioral observations during therapist–child interaction sessions.

Imitation ability was assessed using the Motor Imitation Scale (MIS), which evaluates children’s capacity to imitate actions across multiple domains, including object-related imitation and meaningful versus non-meaningful movements. Social functioning was assessed using the Social Responsiveness Scale (SRS; Constantino & Gruber), which captures autism-related social impairment across domains such as social awareness, social cognition, social communication, social motivation, and autistic mannerisms.

In addition to standardized scales, spontaneous language and stereotyped behaviors were coded during 15 min semi-structured play interactions. Spontaneous language was defined as contextually appropriate utterances that were not directly prompted or imitated from adult speech. Stereotyped behaviors were defined as repetitive, restricted, or ritualized actions and persistent non-functional interests, and were quantified based on observed frequency.

Joint attention behaviors were scored following established criteria [54], ranging from no response (0) to gaze alternation (1), pointing (2), showing (3), and gaze following (4). The therapist initiated joint attention opportunities approximately once per minute, and children’s responses were recorded accordingly.

Play skills were evaluated using a hierarchical developmental framework consisting of four levels: no play (1), manipulation (2), functional play (3), and pretend play (4). Across sessions, the therapist modeled appropriate toy use and children’s engagement levels were systematically documented.All behavioral assessments and coding were completed by an independent assessor blinded to intervention phase to reduce observer bias.

3.4. Data Analysis

For Study 1a, EEG power and time–frequency analyses were conducted to compare neural responses under high- versus low-sociality conditions, focusing on $\alpha$ (8–12 Hz) and $\beta$ (13–30 Hz) bands. For Study 1b, behavioral outcomes were analyzed using paired nonparametric Wilcoxon signed-rank tests due to the small sample size. Effect sizes were calculated to quantify the magnitude of change. Statistical assumptions (e.g., distributional constraints and suitability of nonparametric tests) were evaluated prior to analysis, and all tables include complete titles and corresponding test information to ensure interpretability and reproducibility.

4. Data Analysis and Results

4.1. EEG Experiment on Psychological Imitation in Children with Autism Under High/Low Sociality Styles (RQ1)

This experiment primarily analyzed the differences in EEG responses of children with ASD under high and low sociality conditions, focusing on the $\alpha$ (8–12 Hz) and $\beta$ (13–30 Hz) frequency bands, which are closely related to imitation. Through frequency-domain and time–frequency analyses, we describe preliminary neural response patterns associated with action observation under different sociality conditions.

4.1.1. Frequency-Domain Analysis

Before conducting the frequency-domain comparisons, EEG power in the $\alpha$ and $\beta$ bands was computed based on power spectral density (PSD) estimation. Specifically, artifact-free epochs were extracted for each stimulus condition, and spectral power was calculated using a Fast Fourier Transform (FFT) approach with Hanning windowing. Band-limited power was obtained by averaging PSD values within the predefined frequency ranges: $\alpha$ (8–13 Hz) and $\beta$ (13–30 Hz). Mean band power values were then extracted from relevant frontal and prefrontal electrodes (AF3, AF4, F3, F4) for subsequent statistical comparisons.

It should be noted that the classical $\mu$ rhythm is also typically defined within the 8–13 Hz range and is closely related to action observation and imitation processes. In the present study, however, $\mu$ activity was not analyzed as a separate component because the wearable Emotiv EPOC+ system does not include central sensorimotor electrodes (e.g., C3/C4), which are necessary for reliably isolating $\mu$ suppression. Therefore, the reported $\alpha$-band findings should be interpreted as general alpha-range oscillatory activity rather than a specific marker of $\mu$ rhythm modulation.

In the $\alpha$ band (Figure 4), both Condition A (low sociality) and Condition B (high sociality) showed frontal and prefrontal power distributions, with minor involvement of the occipital lobe, while activation in the parietal and central regions was relatively low. In comparison, Condition A displayed a broader distribution and higher spectral power, particularly in frontal regions, whereas activation under Condition B was relatively more restricted. Spectral analysis of relevant channels (F3, F4, AF3, AF4) further supported these findings, indicating that the energy levels in Condition A were higher than in Condition B.

In the $\beta$ band (Figure 5), the activation regions under both conditions were concentrated in the frontal, prefrontal, and temporal lobes, with a small distribution in the occipital lobe. Condition A demonstrated a wider activation range, with higher energy levels in the prefrontal and bilateral temporal regions compared to Condition B. Overall, low-sociality stimuli elicited more extensive frontal spectral activation across both frequency bands.

The results indicate that the neural responses of children with ASD during the action observation task were primarily concentrated in the frontal and prefrontal regions. Under the low-sociality condition, the activation was more extensive and stronger, whereas under the high-sociality condition, the activation was weaker, suggesting relatively reduced frontal engagement when children observed highly social stimuli, rather than providing direct evidence of behavioral “avoidance”.

4.1.2. Time-Frequency Analysis

Time–frequency representations were computed using ERSP measures, and statistical contrasts between conditions were evaluated using nonparametric approaches with correction for multiple comparisons across time–frequency points ($p\le 0.05$). All reported effects reflect within-participant condition differences rather than between-group comparisons.

At the F3 channel (Figure 6 ), Condition A exhibited event-related synchronization (ERS) at 1700 ms, whereas Condition B showed multiple ERS occurrences at 500 ms, 900 ms, and 1700 ms, indicating a more complex activation pattern. Significance testing revealed significant between-condition differences in the $\alpha$ band after 1500 ms ($p\le 0.05$), and in the $\beta$ band from 700 to 2000 ms.

At the F4 channel (Figure 7), Condition A showed ERS at 100–200 ms and around 1000 ms, while Condition B exhibited sustained ERS between 500 and 1000 ms, with larger fluctuations in the $\beta$ band. Differential testing indicated a significant difference in the $\alpha$ band at 1500 ms, while the $\beta$ band showed a more scattered pattern of significance.

At the FC5 channel (Figure 8), Condition A exhibited low-intensity ERS at 250 ms and 750 ms, and high-intensity ERS at 1500 ms. Condition B demonstrated weaker activation, with responses only at 750 ms and 1500 ms. Significant differences in the $\beta$ band were observed at 500 ms, 1000 ms, and 1500 ms ($p\le 0.05$).

At the FC6 channel (Figure 9), Condition A showed ERS post-baseline and at 1500 ms, with event-related desynchronization (ERD) around 1000 ms. Condition B exhibited sustained ERS post-baseline and at 500 ms. Differential testing revealed significant differences in the $\alpha$ band between 500 and 1500 ms.

Overall, the results indicate that children with ASD exhibited stronger and more stable neural activation under the low-sociality condition. Although the high-sociality condition elicited a more complex activation pattern, these findings should be interpreted as descriptive neural response differences rather than evidence of processing efficiency or avoidance mechanisms.

4.2. Intervention Study on Social Imitation Skills in Autism Through Reciprocal Imitation Training (RQ2)

Study 1b examined preliminary behavioral changes following a six-week Reciprocal Imitation Training (RIT) program in four preschool children with ASD. Pre–post comparisons were conducted using paired Wilcoxon signed-rank tests due to the small pilot sample size. Given the exploratory nature of this intervention component, results are reported primarily as descriptive trends supported by effect size estimates.

4.2.1. Imitation Skills (MIS)

At baseline, all four participants exhibited limited imitation abilities. Following the intervention, MIS scores increased across participants, with WSY and XX showing the most pronounced gains (Figure 10). Median MIS scores improved from 18 (pre-intervention) to 22 (post-intervention).Median MIS scores improved from 18 (pre-intervention) to 22 (post-intervention), corresponding to a median increase of 5 points (95% CI [2, 6], bootstrap). The Wilcoxon signed-rank test indicated a trend-level improvement that did not reach conventional statistical significance ($Z=-1.826$, $p=0.068$;

Table 2. Wilcoxon Signed-Rank Test Comparing Pre- and Post-intervention MIS Scores.

Outcome (MIS)	Median (Pre)	Median (Post)	Z	p	Effect Size (r)	Median Change (95% CI)
Total MIS score	18	22	−1.826	0.068	0.91	5 [2, 6]

Note. Wilcoxon signed-rank test (two-tailed). Effect size was calculated as $r=\mid Z\mid /\sqrt{N}$. Positive values indicate improvement in MIS scores.

). However, the associated effect size was large ($\mid r\mid =0.91$), suggesting potentially meaningful gains despite limited statistical power.

4.2.2. Social Sensitivity

SRS scores decreased from pre- to post-intervention in all four children, indicating a general reduction in autism-related social impairment. The largest decreases were observed in WSY and XX (Figure 11). Mean SRS scores declined from 134.5 (pre) to 123.5 (post), with median values decreasing from 133 to 122, corresponding to a median reduction of 10 points (95% CI [−21, −3], bootstrap).

This change did not reach statistical significance (Z = −1.826, $p=0.068$;

Table 3. Wilcoxon Signed-Rank Test Comparing Pre- and Post-intervention SRS Scores.

Outcome (SRS)	Mean (Pre)	Mean (Post)	Median (Pre)	Median (Post)	Z	p	Effect Size (r)	Median Change (95% CI)
Total SRS score	134.5	123.5	133	122	−1.826	0.068	0.91	−10 [−21, −3]

Note. Lower SRS scores indicate reduced social impairment. Wilcoxon signed-rank test (two-tailed). Effect size was calculated as $r=\mid Z\mid /\sqrt{N}$. Negative change values indicate reductions in SRS scores (improved social functioning).

), although the effect size estimate was again large ($\mid r\mid =0.91$). These findings should be interpreted as exploratory patterns within a small feasibility sample.

4.2.3. Spontaneous Language and Stereotyped Behaviors

Following the intervention, spontaneous language generally increased across participants, with WY demonstrating the most noticeable improvement (Figure 12). Reductions in stereotyped behaviors were observed in WY and WSY, while XX showed a slight decrease and DSK exhibited minimal change. Due to the limited sample size, these observations are presented descriptively.

4.2.4. Joint Attention

At baseline, WY and DSK displayed relatively low joint attention responses. Post-intervention, WY showed increased gaze alternation, and WSY demonstrated emerging pointing and showing behaviors (Figure 13). XX remained relatively stable, whereas DSK continued to exhibit limited joint attention engagement. These patterns suggest possible improvement trends following RIT.

4.2.5. Play Skills

At baseline, most children primarily engaged in manipulative play. After the intervention, WY and XX demonstrated increased functional play behaviors, and WSY occasionally exhibited pretend play (e.g., symbolic toy use). In contrast, DSK showed little change (Figure 14). These results may indicate developmental progress in play complexity, although no inferential conclusions can be drawn from the current pilot design.

Overall, descriptive improvements were observed across imitation ability, social sensitivity, spontaneous language, joint attention, and play skills following the RIT program. However, given the very small sample size and the absence of a control condition, these findings should be interpreted as preliminary feasibility evidence rather than confirmed treatment effects. Larger controlled studies are required to validate these initial behavioral trends.

5. Discussion

5.1. (RQ1) EEG Experiment on Psychological Imitation in Children with Autism Under High/Low Sociality Styles

Unlike previous studies, this experiment emphasized social imitation in promoting children’s imitative behaviors. Compared with highly structured imitation training, RIT advocates imitation within therapist–child interactions, allowing children to exert control over the environment, enhancing autonomy, and facilitating generalization of imitation. Furthermore, the interactive nature of RIT also contributes to interventions targeting social abilities. In addition to assessing basic imitation skills, this study evaluated related social factors, including social sensitivity, spontaneous language, stereotyped behaviors, joint attention, and play skills, all of which pertain to social communication. Comprehensive assessment of these domains allows investigation of RIT’s effects not only on imitation per se but also on socially relevant abilities.

The results indicated that RIT produced broad effects, with varying degrees of change observed across all six evaluation domains; however, given the small pilot sample and the absence of a control group, these findings should be interpreted as preliminary behavioral trends rather than definitive intervention outcomes. In addition, severity-related differences are reported descriptively and should not be taken as firm subgroup evidence.

5.1.1. Frequency-Domain Differences Under Different Sociality Styles

Frequency-domain analysis revealed that children with ASD showed EEG power differences across imitation-related stimuli, with activation patterns mainly observed in frontal and prefrontal regions. Importantly, because the EEG component included only children with ASD, the present findings should be interpreted as within-group contrasts between low- and high-sociality stimulus conditions, rather than ASD-specific atypicalities. Topographical maps showed activation concentrated in the frontal and prefrontal regions. Although these regions are often discussed in relation to higher-order cognitive processing, EEG scalp-level power patterns cannot directly specify underlying executive or social-cognitive mechanisms. Therefore, our interpretation is restricted to describing differential neural engagement across conditions.

Comparisons at four target channels (F3, F4, FC5, FC6) indicated that low-sociality stimuli elicited stronger activation than high-sociality stimuli. Previous research has reported $\mu$ suppression in ASD individuals when observing actions [55], without differentiating social factors. In the present study, we observed condition-related differences in alpha-range activity; however, due to the limited channel coverage of the wearable EEG system, these findings should be viewed as exploratory correlational patterns rather than localized neural mechanisms.

5.1.2. Time-Frequency Differences Under Different Sociality Styles

Time–frequency analysis demonstrated significant differences in neural response patterns under high- and low-sociality conditions. ERSP results revealed (1) Both conditions were dominated by ERS, but with distinct temporal and frequency distributions; (2) High-sociality conditions elicited more complex activation with greater fluctuations, particularly in the $\beta$ band; (3) Four-channel significance analysis highlighted prominent differences in the $\beta$ band.

ERS is generally associated with cortical activation and information processing. However, these neural signatures do not provide direct evidence for psychological constructs such as “avoidance”. Instead, reduced activation under high-sociality stimuli may reflect relatively lower neural engagement or altered processing demands within the ASD group.

This study demonstrated that children with ASD do not “ignore” imitation stimuli; rather, the observed differences suggest that neural responses vary depending on the social richness of the action-observation stimuli. Similar findings were reported by Paula, who observed high-frequency ERS activation in ASD individuals during facial expression tasks [56].

5.2. (RQ2) Intervention Study on Social Imitation Skills in Autism Through Reciprocal Imitation Training

The results of this study indicate that RIT may exert multifaceted influences on children with ASD, including imitation skills and interaction-related domains such as social sensitivity, spontaneous language, stereotyped behaviors, joint attention, and play skills. However, because the Wilcoxon tests did not reach statistical significance and the intervention sample was extremely small (n = 4), these changes should be interpreted as descriptive trends rather than statistically confirmed treatment effects.

Related studies support the feasibility of expanded RIT formats [8,57]. Nevertheless, the current study did not include longitudinal EEG measures, and thus no conclusions can be drawn regarding neural change following RIT.

5.2.1. Effects of RIT on Imitation Skills

To examine the effects of RIT on imitation, participants with relatively low baseline imitation abilities were selected. After intervention, MIS scores increased in three participants, consistent with previous studies [52]. Although improvements were observed descriptively, the statistical evidence remains preliminary and should not be overgeneralized.

However, RIT had limited effects on children with severe ASD, whereas improvements were more pronounced in children with mild to moderate ASD. Given the small sample size, this severity-related pattern should be viewed as exploratory.

5.2.2. Effects of RIT on Social Sensitivity

The effects of RIT on social sensitivity were heterogeneous. Two children with mild to moderate ASD showed decreases in total SRS scores and across subdomains after intervention, while children with severe ASD showed little change. These observations are reported as individual-level trends rather than confirmed group-level effects.

5.2.3. Effects of RIT on Spontaneous Language and Stereotyped Behaviors

After the intervention, all participants exhibited increases in spontaneous language. WSY, who initially showed slower language development, demonstrated the emergence of new utterances in the later stages of intervention. Prior research has indicated that RIT facilitates language development, although specific mechanisms are difficult to isolate. It is hypothesized that speech imitation, language mapping, and nonverbal imitation jointly contribute to language development. The present results support the roles of language mapping and nonverbal imitation.

Importantly, given the small sample size and the exploratory nature of the current intervention study, these language-related changes should be interpreted as preliminary developmental trends rather than statistically confirmed treatment effects. Nevertheless, the observed increase in spontaneous utterances is consistent with the view that reciprocal imitation contexts may provide richer opportunities for communicative engagement and social learning.

Concurrently, stereotyped behaviors decreased as interaction and imitation increased. Previous studies suggest that stereotyped behaviors in children with ASD function as a form of self-stimulation and can be reduced through positive interaction [58].

In the present study, reductions in stereotyped behaviors were noted descriptively during interactive sessions, which may reflect increased social engagement and attentional redirection during play-based imitation exchanges. However, these observations remain correlational and require validation in larger controlled trials.

5.2.4. Effects of RIT on Joint Attention

RIT facilitated the emergence of joint attention, with participants progressing from low levels at baseline to higher levels during the maintenance phase. Imitation involving objects, particularly toys, more readily captured the children’s attention and promoted subsequent social interaction. Moreover, when children acted as demonstrators and the therapist followed their attention, this may have further enhanced their interest in and understanding of joint behaviors.

Joint attention is widely regarded as a foundational component of early social communication development, and reciprocal imitation interventions such as RIT may support this domain by creating repeated opportunities for shared focus, turn-taking, and socially contingent responses.

Nevertheless, because the current study did not include a comparison condition, improvements in joint attention should be interpreted cautiously as individual-level trends that may also reflect natural developmental progression or contextual factors beyond the intervention itself. Future work with controlled designs will be necessary to clarify causal contributions.

5.2.5. Effects of RIT on Play Skills

RIT intervention not only promoted imitation skills but also enhanced children’s representational and social abilities through interactive imitation and interactive contexts, thereby fostering the development of play skills.

Play is an essential context through which children develop symbolic representation, flexibility, and peer-oriented social engagement. The interactive structure of RIT, which emphasizes reciprocal action exchanges and shared enjoyment, may provide a supportive framework for promoting more functional and socially meaningful play behaviors.

In the present pilot sample, some children demonstrated progress from simple manipulation toward more functional or pretend play activities. However, given the limited sample size and descriptive nature of the outcome evaluation, these findings should be viewed as preliminary evidence of feasibility rather than definitive conclusions about intervention efficacy. Larger-scale studies are needed to determine whether RIT produces reliable and sustained improvements in play development.

6. Conclusions, Implications and Prospects

6.1. Implications

The contributions of this study can be summarized in three main aspects: First, from the perspective of social motivation theory, it is proposed that imitation deficits in ASD are not solely attributable to cognitive representation or mirror neuron system impairments, but are also related to insufficient social motivation. The observed abnormal neural responses under high-sociality stimuli provide empirical support for the social motivation hypothesis [10]. Second, at the neural mechanism level, this study elucidates the social imitation characteristics of children with ASD. In the $\beta$ band, children with ASD exhibited more complex but inefficient synchronization responses to high-sociality tasks, indicating reduced processing efficiency for social information. This finding aligns with Paula [56] and offers neurophysiological evidence for the phenomenon of “social avoidance”. Third, the intervention results demonstrate that RIT can effectively enhance imitation skills and promote the development of related social abilities. The effects were particularly pronounced in children with mild to moderate ASD, confirming the theoretical foundation of Naturalistic Developmental Behavioral Interventions (NDBI) and highlighting the central role of social interaction in intervention. From a practical perspective, this study provides the following recommendations for assessment and intervention: In assessments, social action observation tasks should be included and combined with multimodal measures such as EEG to comprehensively capture children’s social motivation and imitation abilities. In interventions, interaction should be emphasized, incorporating social cues such as facial expressions and vocalizations, using objects as “scaffolds”, and enhancing social motivation through role-switching. Interventions should be individualized according to the severity of the child’s symptoms. Families and schools can promote children’s social engagement motivation through imitation-based play, while portable EEG and other tools can be used for real-time monitoring of intervention outcomes. At the policy level, the promotion and training of NDBI should be encouraged, particularly to implement social imitation intervention models in grassroots rehabilitation institutions.

6.2. Limitations and Future Research

The present study has several limitations that should be considered when interpreting the findings. First, the sample size was relatively small (11 participants in the EEG experiment and 4 in the intervention study), which constrains statistical power and limits the generalizability of the results.We acknowledge that no formal a priori power analysis was conducted, as the present work was designed as an exploratory pilot study in a clinically recruited preschool ASD population, where participant availability, tolerance, and artifact-free EEG acquisition pose substantial feasibility constraints. Future studies will incorporate power-informed sample planning and larger cohorts to confirm these initial observations. Second, the experimental paradigm could not precisely capture the exact moment when imitation was internally generated, and the stimulus materials were relatively unidimensional. This limits the specificity of the neural interpretations, although we attempted to reduce confounds by standardizing stimulus duration and focusing on the pre-imitation (action observation) stage. Third, motion artifacts remain an unavoidable challenge in EEG research with preschool children with ASD, which may affect signal quality. To address this, recordings were conducted under therapist supervision, and only data meeting quality criteria were included in the final analyses. Fourth, the RIT intervention period was relatively short, which restricts conclusions about long-term maintenance and transfer effects. Therefore, the intervention outcomes should be interpreted as short-term trends rather than sustained efficacy. Finally, because RIT was implemented in an individualized clinical context without a control condition, causal inferences regarding treatment effectiveness are limited, and the results should be viewed as exploratory evidence supporting feasibility. In Study 1b, intervention outcomes were illustrated mainly through individual trajectories due to the very small pilot sample (n = 4), which limited robust group-level inference. Future studies with larger cohorts will enable more comprehensive group-level analyses of treatment effects. Accordingly, the neural findings should be interpreted as correlational patterns rather than direct evidence of ASD-specific mechanisms or localized neural circuits. In addition, although the social motivation theory was used as a theoretical framework to interpret reduced engagement with highly social stimuli, the present study did not directly assess social motivation mechanisms through dedicated behavioral or reward-based measures (e.g., gaze preference, social reward sensitivity, or standardized motivation scales). Therefore, the link between our neural findings and social motivation theory remains indirect. Future studies should incorporate explicit social motivation assessments to test this theoretical account more rigorously.

Future research could strengthen the evidence base by: (1) expanding the sample size across different ages and severity levels; (2) optimizing EEG paradigms with richer and better-controlled social stimuli while further reducing movement-related artifacts; (3) extending intervention duration and incorporating follow-up assessments to evaluate long-term and generalized outcomes; (4) testing diversified intervention models involving parents, peers, and other social partners; (5) examining additional forms of social imitation, such as automatic imitation; and (6) integrating broader neurophysiological measures to construct a more comprehensive model of social imitation mechanisms in ASD.

Overall, despite these limitations, the present study provides preliminary neural and behavioral evidence highlighting the importance of social context and interaction-based interventions in understanding imitation difficulties in young children with ASD.Future studies could explore the potential of emerging digital technologies, including artificial intelligence and virtual or extended reality (VR/XR), as complementary tools for assessment and intervention. The integration of such technologies may offer new opportunities to enhance personalization, engagement, and scalability in the treatment and educational support of children with ASD [59,60].

6.3. Conclusions

This study examined how social context modulates imitation-related processing in preschool children with ASD using both an EEG paradigm and a Reciprocal Imitation Training (RIT) intervention. The EEG results showed that low-sociality stimuli elicited stronger frontal and prefrontal activation in the $\alpha$ and $\beta$ bands, whereas high-sociality stimuli were associated with more temporally dynamic but overall reduced neural engagement. These findings suggest that imitation difficulties in ASD may be linked not only to motor aspects but also to differences in processing socially relevant information. The intervention study further indicated that RIT was associated with positive trends in imitation ability as well as related social-communication behaviors, including social sensitivity, spontaneous language, joint attention, and play skills, with greater improvements observed in children with mild-to-moderate ASD than in those with more severe symptoms. Overall, the present findings provide preliminary support for the importance of social factors in imitation and highlight the potential value of socially grounded imitation-based interventions for young children with ASD.