The research team, led by Jingying Chen and Tengfei Gao of Central China Normal University together with colleagues at Wuhan University, recruited 45 children for the study. Twenty-two of them had been diagnosed with autism spectrum disorder, with a mean age of 5.29 years and an age range spanning two to eight years. The remaining 23 children formed a neurotypical control group, with a mean age of 4.37 years and an age range of two to six years. While the children watched carefully selected emotional video clips, the researchers synchronously recorded their brain activity using electroencephalography, a technique that captures the tiny electrical signals generated by populations of neurons firing near the scalp.

Electroencephalography, or EEG, is prized in developmental neuroscience precisely because it is non-invasive, relatively tolerant of movement, and capable of resolving brain activity on a millisecond timescale. For young children, including those as young as two years old, it is often the only practical method for probing live brain function. In this study, the raw EEG signals were decomposed into their constituent frequency bands, each of which is associated with different aspects of neural processing. Delta waves, the slowest oscillations, are typically linked to deep processing and attentional engagement. Beta waves, which are faster, are associated with active cognitive processing, motor planning, and the regulation of emotional responses.

The first major analysis focused on power spectral density, a measure of how much energy the brain devotes to each frequency band. Here the researchers found a clear spatial signature of difference. Children with autism showed reduced beta-band activity in the frontal regions of the brain, the areas responsible for executive control, emotional regulation, and the interpretation of social cues. At the same time, they exhibited enhanced delta-band activity in the temporo-occipital areas, regions involved in visual processing and the integration of sensory information. In other words, while the neurotypical children’s frontal circuits appeared to be actively regulating and evaluating the emotional content, the children with autism showed comparatively less frontal engagement alongside heightened slow-wave activity in visual and temporal processing areas.

Beyond simple power measurements, the team turned to entropy analyses, which quantify the complexity of brain signals. Sample entropy and differential entropy are mathematical tools that capture how unpredictable or richly structured a signal is over time. A brain signal with high entropy reflects flexible, dynamic neural processing; lower entropy suggests more rigid or less varied activity. The results were unambiguous: children with autism displayed lower brain complexity during emotional processing than their neurotypical peers. This reduction in neural complexity aligns with a growing body of literature suggesting that autism involves differences in how flexibly brain networks adapt to changing emotional and social stimuli.

The third analytical pillar was functional connectivity, which examines how synchronously different brain regions oscillate together. Coordinated activity between distant regions is thought to reflect communication within brain networks, and disruptions to this coordination have repeatedly been implicated in autism. In this study, the pattern was frequency-dependent. The children with autism showed increased high-frequency synchronization across brain regions, suggesting that their fast oscillatory networks were unusually tightly coupled during emotional stimulation. The control group, by contrast, displayed more coordinated low-frequency connectivity patterns, indicating that their slower networks carried the burden of integration. This dissociation hints that the two groups may achieve emotional processing through fundamentally different oscillatory architectures.

To guard against the possibility that these findings were statistical artifacts, the researchers applied a machine learning validation strategy. They trained an XGBoost classifier, a powerful gradient-boosted decision tree algorithm, to distinguish between the two groups based on their EEG features, and then used SHapley Additive exPlanations, or SHAP, to interpret which features drove the model’s predictions. SHAP values, borrowed from cooperative game theory, assign each feature a precise contribution to every prediction, making the model’s reasoning transparent rather than opaque. The SHAP-based analysis confirmed the significance and predictive value of the beta- and delta-band features in the frontal and occipital regions, lending independent computational support to the classical statistical results from the t-tests.

The identification of these features as potential biomarkers is arguably the study’s most consequential contribution. Biomarkers, objective biological measurements that correlate with a condition, are desperately needed in autism research, where diagnosis currently relies on behavioral observation and clinical judgment, often arriving years after parents first notice differences. If EEG signatures such as reduced frontal beta power and elevated temporo-occipital delta power can be reliably measured in children as young as two, they could eventually complement behavioral assessments, enabling earlier identification and earlier access to support. The authors suggest that these markers may also inform the development of targeted neurotherapeutic interventions, therapies designed to modulate specific neural circuits rather than addressing symptoms alone.

The study’s findings also speak to a broader scientific conversation about the nature of emotional processing differences in autism. Previous research has produced a complicated picture, with some studies reporting deficits in facial emotion recognition and others finding that apparent impairments depend heavily on task demands and measurement methods. By combining spectral analysis, entropy measures, connectivity analysis, and explainable machine learning within a single paradigm, the new work adds converging, multi-level evidence that emotional processing in autism involves measurable differences in both the location and the rhythm of brain activity, while the shared experimental setting underscores that children in both groups were engaged with the same emotional material.

As with any study, the findings come with scope for refinement. The sample was modest, and the age ranges of the two groups overlapped but were not identical, considerations that future work with larger and more closely matched cohorts will need to address. Nevertheless, the study demonstrates a methodological template for the field: pairing classical EEG analyses with explainable machine learning to extract robust, interpretable neural signatures from young children. As EEG technology becomes more portable and machine learning pipelines more refined, the prospect of using a child’s brain rhythms to understand, and ultimately support, their unique way of experiencing emotion moves steadily closer to the clinic.

Subject of Research: EEG-based comparison of emotional processing in children with and without autism spectrum disorder

Article Title: Similarities and Differences in Emotional Processing Between Children With and Without Autism Spectrum Disorder: Evidence from an Electroencephalogram Case Study

Article References: Chen, J., Mao, N., Yang, Z., Hu, X., Chen, D., Zuo, Y., & Gao, T. (2026). Similarities and Differences in Emotional Processing Between Children With and Without Autism Spectrum Disorder: Evidence from an Electroencephalogram Case Study. Frontiers of Digital Education, 3(1), Article 1. https://doi.org/10.1007/s44366-026-0075-1

Image Credits: AI Generated

DOI: 10.1007/s44366-026-0075-1

Keywords: autism spectrum disorder, electroencephalography, emotional processing, beta-band activity, delta-band activity, entropy analysis, functional connectivity, XGBoost, SHAP, biomarkers, neurodevelopment, children