| Project Detail |
Humans effortlessly process over 200 spoken words per minute in casual conversation. Speech recognition algorithms still fail at this challenging task. Our superior performance stems from our capacity to predict what the speaker may say next. Understanding how the brain uses these predictions to process the sensory input is crucial to understand perceptual function and dysfunction: dyslexia, autism and psychosis have all been linked to an impaired handling of predictions.
Hierarchical predictive coding (HPC) is the current leading framework to understand how predictions help us processing sensory inputs. However, HPC is only compatible with the function and organisation of the cerebral cortex. This is a decisive shortcoming: while only cortical stages have the foresight to perform conceptually accurate predictions, only subcortical stations have the temporal properties required to correctly process fast sensory inputs.
SynPrePro will reformulate HPC as an integrated theory explaining how cortical and subcortical stages work together to proficiently process fast and complex sensory inputs like speech.
I will use a unique experimental-theoretical approach to study the human auditory pathway as a model for sensory pathways in four work packages (WPs). WP1 will use cutting-edge human neuroimaging to unravel the implementation of HPC in the auditory pathway. WP2 will use innovative model-based neuroimaging to identify the mechanisms responsible for the generation of conceptually accurate and temporally precise predictions. In WP3 I will develop a ground-breaking computational model to identify the neural mechanisms implementing HPC in the thalamocortical loop. WP4 will use big-data analytics to disentangle how cortical and subcortical stages work together to swiftly process speech.
The outcomes will turn the cortical paradigm of HPC into an integrated theory of cortico-subcortical interactions, revolutionising our understanding of perceptual function and dysfunction. |
