A New Eye On Semantics: Probing the Spatiotemporal Brain Dynamics Underlying Natural Reading

In this thesis, I examined the influence of task and context on the semantic processing of single words in isolation and in a sentence context. In one study, I considered task effects on single-word semantic representations using multivariate pattern analysis. In two further studies, I tested whether sentence context affects the semantic processing of single words during natural reading. Overall, these studies aimed at testing predictions made by the Controlled Semantic Cognition (CSC) framework during natural reading, in order to better characterise the dynamic relationship between semantic control and representation mechanisms. To achieve this, I considered dynamic brain activity with good spatio-temporal resolution as recorded by EEG/MEG as well as behaviour as recorded by eye-tracking. In my multivariate pattern analysis in EEG/MEG source space data, I found that while semantic representations show some stability across tasks, even subtle task differences are associated with widespread differences. I found that stimulus and task effects are better assessed by different approaches: stimulus information is best investigated by considering the decoding performance in each individual ROI (by-ROI approach) while distributed, task effects are better understood by including all ROIs and assessing the back-projected model’s weights (across-ROI approach). In my behavioural study, I found additive effects of predictability and concreteness on early reading time measures, paralleling previous results on the relationship between predictability and frequency. An analysis on a larger and freely available corpus produced compatible results. For the first time, I performed a co-registration study combining EEG+MEG and eyetracking using a deconvolution approach to separate brain activity generated by overlapping fixations. Firstly, I tested some of the proposed procedures for ensuring a good data quality, i.e., removing artefacts generated by eye movements, and proposed some novel methodological improvements, such as the combination of the variance ratio and EOG correlation to detect artefactual components in independent component analysis. Analyses of the neural activity revealed an effect of predictability in sensor space and decoding analysis, which indicated that the left temporal lobe is crucial for integrating words into higher, sentence-level meaning. I found no effect of concreteness in the analyses conducted. Moreover, the behavioural results obtained from this co-registration analysis did not replicate the results obtained in the previous behavioural study. In summary, this thesis reveals novel insights into the dynamic interaction of control and representation mechanisms, by considering the effects of task and context. The methodological advancements contribute to the growing field of EEG/MEG and eyetracking co-registration.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2023

Author dc:creator

• Magnabosco, Federica

Advisors dc:contributor.advisor

• Hauk, Olaf
• Lambon Ralph, Matthew A

• cognitive neuroscience
• reading
• semantics
• EEG/MEG
• eye-tracking
• language comprehension

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/20f5edcf-ce87-49fb-928d-e8edbd0e944b/download
• https://creativecommons.org/licenses/by/4.0/

Language dc:language

eng

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.110758

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/371646