Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Predictive Language Processing in Preschool Children: Naturalistic EEG Data Reveal Individual Differences in Surprisal Effects
Poster Session C, Friday, October 25, 4:30 - 6:00 pm, Great Hall 3 and 4
Ina Bornkessel-Schlesewsky1, Sophie Jano1, Chloe Dziego1, Imogen Weigall1, Jonathan Kum2, Jerome Mak2, Brian Caswell3, Matthias Schlesewsky1; 1University of South Australia, 2MindChamps PreSchools, Singapore, 3MindChamps Applied Integration Research (AIR)
Predictive coding is a prominent contemporary theory of brain function, positing that the human brain actively generates predictions about its sensory input. Through its unified account of human information processing across a range of cognitive domains, it provides a parsimonious explanation for language processing. In addition, predictive-coding-based approaches are appealing for the neurobiology of language, since there are detailed proposals regarding their neurobiological implementation. There has been considerable discussion about the possible relation between predictive coding and neurophysiological correlates of language processing like the N400. However, considerably less is known about how predictive processing matures during language acquisition, particularly under ecologically valid language processing conditions. Here, we examined predictive processing in preschool children using a naturalistic paradigm. We recorded EEG data from thirty-five children (mean age: 5:6; range: 4:8–6:4) as they performed music exercises in one-on-one sessions with a teacher at a Singaporean preschool (language of instruction: English). Sessions lasted approximately 30 minutes and consisted of a structured sequence of successively more complex tasks targeting rhythm, pitch and melody. Resting-state EEG data was collected in a two-minute eyes-closed session prior to the main task. We calculated regression-based event-related potentials (rERPs) for the instructions provided by the teacher during approximately the first 5–6 minutes of the test session, using the onset of each word as a time-locking point (mean number of epochs: 593; range: 485-684). rERP predictors included prestimulus amplitude, log-transformed word frequency and GPT2-based lexical surprisal, taking the entire preceding session as the context window. We estimated individual peak alpha frequency and aperiodic activity (slope and offset) from the resting-state recordings. rERP data from 20 children (audio transcriptions are ongoing for the remaining 15 participants) revealed word frequency and surprisal effects in the N400 time window (300–500 ms post word onset), with most pronounced amplitudes over left-anterior channels. Linear regression analyses revealed larger surprisal effects for participants with higher aperiodic offsets even when age was controlled for. The current findings indicate that it is possible to examine predictive processing in preschool-aged children under highly naturalistic conditions. In addition, our results suggest that individual differences in the maturity of predictive language processing may correlate with resting-state metrics. The aperiodic offset is associated with aggregate population spiking activity and correlates with the fMRI BOLD response. Interestingly, previous developmental research has shown that aperiodic offsets decrease over the course of childhood development, in parallel to a flattening of the aperiodic slope (Hill et al., 2022). However, in younger and older adults, higher resting aperiodic offsets and steeper aperiodic slopes correlate with a higher adaptability of predictive models in response to novel language input (Bornkessel-Schlesewsky et al., 2022). Tentatively, in spite of the general age-based flattening of aperiodic activity, children with higher aperiodic offsets in comparison to their age cohort may be more sensitive to the predictability of language input. If this relationship is confirmed in future research, aperiodic metrics could potentially serve as age-independent predictors for the maturity of language processing.
Topic Areas: Language Development/Acquisition,