How to measure the information gained from one symbol.

Area of Science:

• Computational Neuroscience
• Information Theory
• Systems Neuroscience

Background:

• Information theory is widely used to quantify neuronal representation of stimuli and behavior.
• A key question is determining the information conveyed by neuronal responses.
• Current methods for measuring accumulated information in neural activity lack additivity.

Purpose of the Study:

• To identify and address the non-additive flaw in the commonly used measure of neuronal information.
• To propose and validate a novel, additive measure for information per neuronal observation.
• To re-interpret existing findings on neural coding using the new additive measure.

Main Methods:

• Theoretical analysis of information measures in the context of neural coding.
• Development of a new definition for information per observation that guarantees additivity.
• Empirical validation using electrophysiological recordings of visual responses from primate cortical neurons.

Main Results:

• The standard definition of accumulated information in neural activity is shown to be non-additive.
• A unique, additive measure for information per observation is derived and proven.
• Application to primate visual cortex data reveals distinct interpretations of neural coding efficiency.

Conclusions:

• The proposed additive measure is the only definition satisfying this critical property for analyzing population codes.
• The findings suggest that neurons, particularly those studied in primate visual cortex, operate substantially below their theoretical information capacity.
• This work provides a more accurate framework for quantifying information processing in the brain.