The brain constantly guesses what word or sound will come next during speech or reading, and this project will decode those predictions from electrical brain activity. Language processing is remarkably fast—around 250 written words or 200 spoken syllables per minute. The brain cannot wait passively for each word to arrive; it must anticipate what is coming. Yet scientists do not know whether predictions cause the brain to process expected input thoroughly or merely check that the guess was correct, nor what happens to a prediction when it is wrong. This project will use EEG recordings and machine learning to decode the brain’s predictions before the predicted stimulus even appears, testing these competing hypotheses directly. This is fundamental science. It will clarify the neural mechanisms that make humans such efficient language users. A deeper understanding of predictive processing could eventually inform therapies for language disorders where prediction may be impaired, such as aphasia or autism spectrum conditions. It may also reveal whether these predictive mechanisms are unique to language or generalise to other sequential tasks like music or motor planning—insights that could shape future cognitive training or communication technologies.
View original technical description
Project Description for Find-A-PhD Advert (max 400 words). This will be the text that will be advertised to prospective students. This project uses advanced analyses of electrical brain activity (EEG) to investigate the neural mechanisms of predictive processing in the human brain. Rather than passively waiting for input to arrive, the brain actively thinks ahead and continuously compares the input against internally generated predictions. Predictions likely underpin rapid processing of predictable input as well as learning from unexpected input. This is especially the case for human language, which unfolds on a particularly rapid time scale (around 250 written words or 200 spoken syllables per minute). However, it remains unclear exactly how predictions modulate the way stimuli are processed, and whether predictive mechanisms are specific to language or generalize to the processing of non-linguistic sequences. This project will systematically vary predictability in sentences and/or other stimulus sequences and examine its effects on temporal and spectral aspects of the EEG. In addition, the project will apply state of the art machine learning techniques to the EEG signal that make it possible to decode the predictions that people form about upcoming input before the input is perceived. This will allow for testing a range of hypotheses about predictive mechanisms. Is predictable input processed thoroughly, because predicted and actual input converge to yield a strong representation? Or is predictable input instead processed in a shallow fashion, as the brain merely verifies that the prediction was confirmed? And when a prediction gets disconfirmed, what happens to the original prediction: does it linger, or is it suppressed and updated? The results will help advance our understanding of the fundamental mechanisms of predictive processing and shed new light on the processes that make people such efficient language users.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know