Project description
Perceptual prediction theories united
Previous experiences play a large role in determining what we perceive (see, hear, taste, feel). Paradoxically, two main theories provide conflicting explanations for this. The first claims that we are more likely to perceive what we expect because it was more likely to be there. The second proposes that we are more likely to perceive what we do not expect because it provides more information. The EU-funded ConflictedPrediction project will address this paradox through pioneering, interdisciplinary research. It will test a new theory using electroencephalography (EEG), magnetoelectroencephalography (MEG), and 7 Tesla functional magnetic resonance imaging to explain how learning could render both veridical and informative perception. Project findings will shed light on underlying processes and deepen our understanding of perception and learning.
Objective
Our sensory receptors are bombarded with noisy, continuous streams of information. From these streams, our brains must construct percepts that are (1) veridical representations of the world, and (2) informative – i.e. highlighting what we did not already know. Cognitive science has suggested that our brain meets these challenges by using probabilistic expectations to shape our perceptual experiences. However, there are currently two broad classes of theory concerning how expectations shape perception, that are both supported by large bodies of evidence and mutually incompatible: some theories propose that we upweight what we expect to generate veridical representations, whereas others propose that we downweight what we expect to privilege the most ‘newsworthy’ information.
ConflictedPrediction will test a new theory addressing and solving this paradox for the first time, to determine how perception can be rendered both veridical and informative. I propose that probabilistic knowledge pre-emptively biases perception towards what is likely, to generate largely veridical experiences rapidly. However, if the input is particularly surprising, catecholamine release – acting to aid learning – reactively enhances perception of these inputs by modulating sensory gain. This perceptual enhancement will generate a clearer estimate of these highly unexpected events to guide model-updating.
To test the theory, ConflictedPrediction will use temporally- and spatially- sensitive neural measures (EEG, MEG, 7T fMRI), in combination with computationally derived parameters of perception and unexpectedness. This interdisciplinary project therefore will unify understanding of perception and learning across typically isolated scientific domains. Its findings will chart a new research frontier for understanding how the brain surmounts key computational challenges, enabling us to survive and thrive in a challenging sensory world.
Keywords
Programme(s)
Funding Scheme
ERC-COG - Consolidator GrantHost institution
WC1E 6BT London
United Kingdom