Relating Population-Code Representations between Man, Monkey, and Computational Models.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Neural activity patterns across neuronal populations are believed to encode perceptual and cognitive content.
• Quantitative methods are needed to compare neural representations and test computational models of brain function.

Purpose of the Study:

• Introduce Representational Similarity Analysis (RSA) as a method to quantitatively relate population-code representations.
• Demonstrate RSA's utility in testing computational models and comparing neural codes across species (e.g., humans and monkeys).
• Illustrate how RSA can reveal representational transformations and links to behavioral data.

Main Methods:

• Characterize neural population codes using representational dissimilarity matrices (RDMs).
• An RDM captures pairwise dissimilarities between neural activity patterns evoked by a set of stimuli.
• Analyze correlations between RDMs derived from different data sources (e.g., experimental recordings, computational models, behavioral data).

Main Results:

• RDMs effectively summarize the information emphasized or deemphasized by a neural representation.
• Correlations between RDMs allow for the validation of computational models against empirical data.
• RSA facilitates cross-species comparisons of neural representations, such as between human and monkey object vision.

Conclusions:

• Representational Similarity Analysis provides a powerful framework for understanding neural coding and bridging gaps in systems neuroscience.
• The method enables quantitative comparisons of brain representations, aiding in the development and testing of computational theories.
• RSA's flexibility allows its application to diverse neuroscience questions, including representational dynamics and behavioral relevance.