Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Roles for alpha/beta and high gamma power in feedback and feedforward signalling during sentence reading?
Poster Session D, Saturday, October 26, 10:30 am - 12:00 pm, Great Hall 3 and 4
Ashley Lewis1,2, Floris de Lange1, Peter Hagoort1,2; 1Donders Institute for Brain, Cognition and Behaviour, 2Max Planck Institute for Psycholinguistics
Predictive coding theories provide a general organizing principle for information processing in the brain. The canonical cortical microcircuit proposes a domain-general role for alpha/beta and gamma neural oscillations in feedback and feedforward signalling respectively within cortical hierarchies. Whether these ideas apply in the context of language processing is an open question. In this MEG study (N=32) we investigated the role of alpha (7-13 Hz), beta (12-22 Hz) and high gamma (HG; 55-95 Hz) in the context of sentence comprehension. Feedback information was manipulated at the level of semantics, with high vs low constraint sentence contexts making the sentence-final target word (TW) more or less predictable respectively. Feedforward information was manipulated by prediction errors at both the level of semantics (TW congruent or incongruent with the preceding sentence context), and visual input (TW visually degraded – blurred – or not). This was designed to illicit prediction error signals at different levels of the language processing hierarchy, and in the case of visually degraded blocks to encourage participants to make increased use of feedback information when it was available. Cluster-based permutation statistics revealed an overall difference in HG power between visible and blurred TWs (p = 0.0018) in the left hemisphere visual cortex, visual word form area, superior temporal gyrus, and supramarginal and angular gyri. Neither congruency nor contextual constraint modulated HG at the TW, nor were there any interactions between these factors, which suggests that if anything HG plays a role in feedforward propagation of visual, but not semantic prediction errors. On the other hand, alpha (p = 0.0006) and beta (p = 0.0002) power in left posterior superior temporal gyrus (STG) prior to the onset of the TW decreased more for blurred compared to visible TWs. These pre-TW alpha and beta effects in left posterior STG likely reflect increased reliance on feedback information when the TWs were not clearly visible. Predictions based on any available information provided by the sentence context may be fed from this higher-order language region back to lower levels of the brain’s language processing hierarchy (e.g., visual word form area and primary visual regions in occipital cortex) in order to ‘fill in’ the missing visual word form information. There were however no effects of semantic constraint on pre-TW alpha/beta power, nor was there any interaction between visibility of the TWs and the degree of semantic constraint, making it difficult to definitively conclude that the semantics of the sentence context is the relevant feedback information being used to predict the blurred TWs. Our results suggest that HG may reflect visual prediction error signals, but more likely indicate delayed visual processing when TWs are blurred. Alpha/beta power prior to the onset of the TW does appear to reflect predictions about the form of the TW, but probably not based on the degree of semantic constraint established by the preceding sentence context. Follow-up work will probe the direction of information flow (connectivity profile) between left posterior STG regions exhibiting alpha/beta power effects and left visual regions exhibiting HG effects.
Topic Areas: Syntax and Combinatorial Semantics, Control, Selection, and Executive Processes