Cardinal mechanisms of perception: Prediction, Valuation, Categorization

Perception is arguably the most basic and most important function of our mind, because it provides the sole source of information about the environment. Our senses present us with a window into the world, enabling us to take up information from the surrounding world. Perception, by contrast, is the process by which this information is interpreted, a “making sense of the senses”. Research on sensory processing has been immensely successful, but we want to understand how the human brain extracts meaning from these basic sensory signals. Previous attempts to understand perception have emphasized specific solutions to specific perceptual problems. Here we propose to understand perception in terms of a set of three underlying principles: Prediction, Valuation and Categorization. These cardinal mechanisms create and maintain sophisticated internal models of the world. The brain is the organ that continuously optimizes these internal models, enabling us to predict the future state of the environment and the consequences of actions, evaluate the potential risks and benefits of different stimuli and responses, and categorize a complex continuous world into discrete mental concepts and behaviors. Accordingly, we organize our proposal into three research areas:

A. Prediction. Project group A will investigate how perceptual predictions actively guide our sensors to acquire information optimally. We seek to understand how predictions allow us to discount the sensory consequences of our own actions and how they enable robust and efficient information uptake.

B. Valuation. Project group B will investigate how valuation processes weigh different sensory signals and action outcomes to maximize information gain and reward. We seek to understand how valuation both optimizes the immediate behavioral consequences of an action and continuously corrects internal models.

C. Categorization. Project group C will investigate how categories are inferred from regularities in the environment, across different domains, like perception and language. We seek to understand the advantages that effective categories entail for perception by emphasizing relevant information.

To obtain a comprehensive understanding of prediction, valuation and categorization, we deploy a unique combination of human behavioral experiments, physiology and modeling.  Our goal is to delineate the cardinal mechanisms behaviorally, to identify their underlying neural substrates and to explain their function with a computational model. In the long run, we seek to extend our investigation of the development of the cardinal mechanisms throughout the entire life span, and to uncover the functional role of their impairment in neurological and psychiatric disorders.