Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Predicting children’s language ability from resting state EEG
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
Henry Railo1, Anna Kautto1, Joonas Karhula2, Heli Parviainen1, Antti Airola1; 1University of Turku, 2Aalto University
Children show considerable variation in language acquisition. Understanding the sources that contribute to this variation is important not only from the basic science point of view, but also to shed light on the mechanisms underlying language disorders. Here, we ask how well children’s language ability can be predicted based on resting state electroencephalography (EEG). By and large, previous research suggests that stronger slow wave activity is associated with poorer language ability whereas higher power in higher frequency bands predicts better language skills. However, these findings are based on studies with small sample sizes that have typically not differentiated between aperiodic and periodic (i.e., rhythmic, oscillatory activity) components of neural activity. We analyzed a large resting state EEG dataset (N = 2415 children aged 5–17 years), and decomposed the EEG spectra into aperiodic and periodic components using the SpecParam model. The children’s language ability was estimated using the Verbal Comprehension Index (VCI) of WISC-IV. Using linear regression, we tested for associations between the aperiodic and periodic components of the EEG spectra and the (age-normalized) VCI scores while controlling for age, sex, and nonverbal cognitive ability (Fluid Reasoning Index, WISC-IV). The results suggested that higher VCI was associated with steeper EEG spectrum (aperiodic exponent) and smaller alpha (8–13 Hz) periodic activity across the whole age range. VCI did not significantly predict the aperiodic offset, or periodic theta (4–8 Hz) or beta (13–30 Hz) activity. However, while statistically significant, the effect sizes of the associations between language ability and EEG were small. We are currently running machine-learning analyses to test how well the language ability of out-of-sample cases can be predicted based on the resting state EEG. We discuss the potential of resting state EEG in characterizing the sources that contribute to individual variation in children’s language ability.
Topic Areas: Language Development/Acquisition,