Back

Neural processing of natural speech by children with developmental language disorder (DLD): EEG speech decoding, power and classifier investigations

Keshavarzi, M.; Richards, S.; Feltham, G.; Parvez, L.; Goswami, U.

2025-07-18 neuroscience
10.1101/2025.07.14.664754 bioRxiv
Show abstract

The sensory/neural Temporal Sampling (TS) theory of developmental language disorder (DLD) is based on the sensory and linguistic impairments in rhythm processing that are found in children with both developmental dyslexia (DD) and DLD. These sensory/linguistic impairments include decreased sensitivity to amplitude rise times (ARTs), syllable stress patterns and speech rhythm. They are explained via a neural oscillatory speech processing framework drawn from the adult literature. Notably, at the neural level TS theory predicts impairments in the cortical tracking of different rates of amplitude modulation (AM) in the speech signal <10Hz. To date, the accuracy of low-frequency cortical tracking in natural continuous speech has not been measured in children with DLD. Here, EEG was recorded during story listening from children with and without DLD aged around 9 years, and decoding analyses in the delta, theta and alpha (control) bands were carried out. EEG power was computed in the delta, theta and gamma bands, as was phase-amplitude versus phase-phase coupling (PAC, PPC) between bands. The expectation that the accuracy of low-frequency decoding (delta, theta) of the speech signal would be impaired in DLD was not supported. Further, theta-delta and theta-gamma PAC were not atypical, contrary to prediction. EEG power in all bands was elevated for the DLD group, however. Receiving-operator-characteristic (ROC) curves estimated using support vector machine and logistic regression showed that these simple band power measures provided classifier models with the highest areas under the curves (AUC). The data are discussed using TS theory. HighlightsO_LICortical decoding accuracy for speech envelope information < 10 Hz is intact in DLD. C_LIO_LIChildren with DLD show greater EEG power in delta, theta, and low gamma frequencies. C_LIO_LILow-gamma EEG power yields the best DLD classification performance (AUC = 0.81). C_LI

Published in Brain and Language (predicted rank #20) · training set

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.