Team
Research Projects
Publications
Collaborators
Open Science
Main Page

Neuroscience of Cognition and Speech - NeuroCogS

Research interests: Speech comprehension is a dynamic and context-sensitive process. I use neuroimaging to understand how cortical analysis of language adapts to different contexts, goals and social environments. I am also deeply interested in using current neuroimaging methods for assessment, early diagnosis, and remediation of developmental and acquired disorders of language. All current projects

Goal oriented speech processing

The aim of this research strand is to understand how cortical speech processing at multiple levels (auditory, lexico-semantic and syntactic) is shaped by the listener’s communicative goals. Addressing this question is key to updating current models of cortical speech analysis. To address this question, we use EEG and naturalistic story listening. Project collaborators: Dr Anastasia Klimovich-Gray and Dr Agnieszka Konopka.

Music and Reading

The goal of this project is to relate reading performance in young readers (5-7 year olds) to their neural signatures of speech and music processing, music training and key executive skills. We will explore the hypothesis that musical training can facilitate cortex-to-music and cortex-to-speech entrainment and have beneficial effects for reading skills. In the long term this work will help to develop intervention tools for stimulating normal phonological awareness in children who struggle with reading. Project collaborators: Mari Garcia de Soria Bazan, Dr Anastasia Klimovich-Gray (supervisor), Dr Brian Mathias (co-supervisor), Dr Anne Keitel (co-supervisor). Funders: PhD Studentship from the School of Psychology, Aberdeen to Mari Garcia de Soria Bazan.

Human-Model-Priming

This is a collaborative project between the Computer Science and Psychology Departments in the University of Aberdeen. The key purpose of this study is to compare syntactic priming effects generated by human participants to that in AI systems that model language processing (such as Large Language Models - LLMs). We plan to understand the extent to which LLMs can learn to encode structure in language and whether resulting representations are similar to the human ones. Project collaborators: Dr Arabella Sinclair, Dr Agnieszka Konopka, Dr Anastasia Klimovich-Gray, Dr Nika Adamian. Funders: Aberdeen University, Internal Funding to Pump-Prime Research and Research Networks to all collaborators.

Neural2Speech

The goal of neural2speech is to build a multimodal contrastive neural network model trained on both neuroimaging input (fMRI and MEG) and speech output. Can we use tools of deep learning to accurately decode speech content from neural brain activity? Addressing this issue is critical for restoring communication in people with speech motor impairments. At the same time, assessing the quality of the perceived speech from brain recordings can provide better understanding of specific linguistic processes affected by a given impairment and provide clinical neuroscience with invaluable assessment tools. Project collaborators: PI, primary supervisor Prof Nicola Molinaro, second supervisor Dr Anastasia Klimovich-Gray, PhD student Vincenzo Verbeni. Funders: Predoctoral Researcher- IKUR Fellowship

BILENT :Pre-print, analysis scripts and EEG data (story listening) + behavioural measures dataset for 35 Basque-Spanish bilingual children (6 y.o.).

Spoken Word Processing in Dyslexia :Pre-print, analysis scripts and MEG data (single word listening) behavioural measures (reading tests) dataset for 14 matched typical and 14 dyslexic readers, adult Spanish native speakers.