大腦預期機制精準度加權的發展(5/5)

The predictive coding model of perception describes the brain as a hierarchical prediction system. It supports perception by constantly comparing predictions with sensory inputs. The sensory inputs that predictions fail to explain would be communicated forward to update predictions in the form of precision-weighted prediction errors. The processing of precision-weighted prediction errors constitutes a core aspect of predictive brain function. However, there is surprisingly little research investigating how the representation of prediction errors might develop in normal aging and how the minimisation of prediction errors might be modulated by external and internal factors affecting prior precision. This five-year project used electroencephalography (EEG) and magnetoencephalography (MEG) to examine how prediction errors are represented and minimised in the auditory cortex to offer a path to understanding these issues. We found that older brains show little decline in local (i.e., first-order) prediction errors but significant diminution in global (i.e., second-order) prediction errors. A brain-behaviour correlation analysis further demonstrated that whether and how lower-level prediction errors propagate to the higher-level might be affected by the age-related capacity to maintain information in working memory. We also found that both contextual regularity and selective attention can modulate the minimisation of prediction errors but in distinct manners. Overall, these findings advance our understanding of the predictive brain, providing precursors and indicators for clinical use pertaining to heterogeneous deficits in predictive function.