Document Type : original article
Authors
1 Department of Cognitive Neuroscience, Faculty of Educational Sciences and Psychology, University of Tabriz, Tabriz, Iran.
2 Department of Mathematics, Institute for Advanced Studies in Basic Sciences, Zanjan, Iran.
3 Department of Biomedical Engineering, Faculty of Engineering, Shahed University, Tehran, Iran.
Abstract
Background:
Autism Spectrum Disorders (ASD) are developmental conditions characterized by impairments in social communication and repetitive behaviors. Verbal impairments are among the core features of ASD. Verbal fluency tasks are commonly used neuropsychological tools to assess language abilities in children with ASD. Recently, speech graph analysis—derived from graph theory—has emerged as a novel method to provide deeper insights into verbal fluency performance.
Methods:
This study compared speech graph features obtained from phonemic and semantic verbal fluency tasks between children with ASD and typically developing (TD) peers. Participants included 25 children with ASD (ages 7–12 years; IQ 70–85 based on the Goodenough Draw-a-Person Test) from a specialized autism school in Tabriz, and 30 age-matched TD children from mainstream schools. Verbal fluency was assessed using the Kormi-Nouri task, which included phonemic cues (A, N, M) and semantic categories (boy names, girl names, body parts, fruits, colors, kitchen utensils). Spoken words were represented as nodes, and the sequence of word production as edges, forming speech graphs. Standard verbal fluency scores and speech graph metrics were analyzed using independent t-tests and Mann–Whitney U tests.
Results:
Children with ASD produced fewer words in both phonemic and semantic tasks. Their speech graphs showed fewer nodes and edges, smaller largest connected components, lower average shortest paths and diameters, higher graph density, and reduced average total degrees compared to TD peers.
Conclusion:
Speech graph analysis provides a novel computational framework for identifying verbal fluency deficits in children with ASD. These findings highlight its potential for designing computer-based diagnostic and rehabilitation tools. Future research should include larger, more diverse samples and consider variables such as age and sex.
Keywords
)