You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.

Poster A29, Thursday, August 16, 10:15 am – 12:00 pm, Room 2000AB

Inter-individual differences in predictive coding during language processing: the role of individual alpha frequency and idea density

Ina Bornkessel-Schlesewsky1, Caitlin Howlett1, Isabella Sharrad1, Erica Wilkinson1, Matthias Schlesewsky1;1University of South Australia

Predictive coding constitutes a possible unifying theory of information processing in the human brain (e.g. Friston, 2005), including language processing (e.g. Pickering & Garrod, 2007). It is also a highly promising framework for examining the neurobiology of language, given the level of neurobiological detail proposed as part of hierarchical cortical predictive coding architectures (Bastos et al., 2012). However, little is known about whether and how predictive coding differs between individuals (see Moran et al., 2014, for lifespan changes). Here, we examined inter-individual differences in predictive coding during language processing in young, healthy adults, using individual alpha frequency (IAF) and idea density (ID) as predictors. IAF, the peak frequency of the EEG alpha band, correlates with processing speed, memory and intelligence (Klimesch, 1999; Grandy et al., 2013). High-IAF individuals also tend to have more fine-grained perceptual representations (Samaha & Postle, 2015; Cecere et al., 2015). ID measures the number of ideas expressed relative to the number of words used and thus provides a measure of linguistic information encoding efficiency (Kemper et al., 2001). Low ID in young adulthood has been associated with poor cognitive function in older adulthood (Snowdon et al., 1996). In the present study, native speakers of English (n=42; 28 women; mean age:22.8, SD:3.7) listened to 150 passages (4-5 sentences) while their EEG was recorded. Ninety passages contained 2 two-adjective noun phrases (e.g. "huge grey elephant") at varying positions. In 50% of cases, adjective order was non-canonical ("grey huge"; not examined here). We examined N400 responses and pre-stimulus alpha power at IAF at two positions: passage-initial words (low predictability; only the 35 passages starting with proper nouns were used for this analysis) and nouns in the two-adjective NPs (high category predictability due to preceding adjectives). While pre-stimulus alpha power is thought to reflect the precision of predictions (e.g. Bauer et al., 2014), we assume that N400 amplitude indexes predictive model updating. IAF was estimated from eyes-closed resting EEG and ID was calculated from a text sample written in the lab. Data were analysed using linear mixed-effects models, with IAF and ID as continuous predictors. Across both positions, ID-related N400 differences emerged across the course of the experiment. N400 amplitude was inversely correlated with ID, particularly at the end of the experiment and for participants with a low IAF. High ID participants also showed lower pre-stimulus alpha power than low ID participants, again particularly at low IAF. These results indicate that high ID individuals are better able to adapt their predictive linguistic models to the current input, as they show neurophysiological correlates of higher predictive precision and less model updating during later stages of the experiment. High ID may thus be indicative of a more adaptable language model. Intriguingly, ID differences were particularly pronounced for low IAF. We conclude that the quality and flexibility of an individual's language model (ID) as well as biological constraints on the resolution of one's perceptual input (IAF) jointly determine inter-individual differences in the efficiency of predictive coding during language comprehension.

Topic Area: Grammar: Syntax