Neurophysiological differentiation of Boder's dyslexia subtypes using harmonised quantitative EEG cross-spectra

Developmental dyslexia is a heterogeneous disorder classically divided into distinct subtypes. Elena Boder's, 1973 model conceptualised reading and spelling as functions of distributed, interactive neural networks that require a dynamic interplay between visual-gestalt and analytic-auditory processes. Disruption of this interplay produces three main dyslexic subtypes-dysphonetic (DD), dyseidetic (DYD), and mixed (MD). These subtypes have also been described in Italian speakers, and recent research suggests they correspond to specific neurophysiological patterns. In this study, we analysed quantitative EEG (qEEG) from a cohort of 227 children. The dyslexic group comprised DD (n = 169; 110M/59F, age 7-15), DYD (n = 18; 17M/1F, age 7-15), and MD (n = 40; 25M/15F, age 7-18), compared to 100 age-matched typically developing controls (58M/42F, age 6-16). Methodologically, we moved beyond conventional power analysis by employing a stable and sparse regression classifier (SSRC) based on harmonised cross-spectral complex z-scores. This approach is neurophysiologically grounded in the view that reading emerges from the dynamic interplay within distributed cerebral networks. We therefore hypothesised that dyslexia subtypes would be best discriminated by distinct patterns of functional connectivity, which are directly captured by cross-spectral metrics. Consistent with this, while conventional log-spectral measures showed poor subtype discrimination, our cross-spectral analysis revealed distinct connectivity patterns that differentiated the DD, DYD, and MD groups. Classification accuracy was high (areas under the ROC: DD vs DYD = 0.85; DD vs MD = 0.94; DYD vs MD = 0.78), and dyslexic subtypes were clearly separated from controls using fewer variables. These findings provide the first neurophysiological validation of Boder's subtypes, demonstrating that DD and DYD reflect distinct network dysfunctions. The partial overlap between dysphonetic and mixed groups aligns with clinical evidence that MD represents a more severe dysphonetic form. Overall, our results highlight the potential of qEEG connectivity biomarkers to refine dyslexia diagnosis and support personalised interventions.