Decoding the Brain: Neural Representation and the Limits of Multivariate Pattern Analysis in Cognitive Neuroscience

Area of Science:

• Cognitive Neuroscience
• Neuroscience
• Philosophy of Mind

Background:

• Multivariate pattern analysis (MVPA), also known as neural decoding, has become a transformative tool in cognitive neuroscience.
• The field widely accepts the 'decoder's dictum': successful decoding implies that the decoded information is represented by neural patterns.

Purpose of the Study:

• To critically evaluate the philosophical underpinnings of the decoder's dictum.
• To argue that decodability is an insufficient basis for inferring the content of neural representations.
• To propose refinements for justifying inferences about neural representations using MVPA.

Main Methods:

• Philosophical analysis of the decoder's dictum.
• Critical evaluation of the inference from decodability to neural representation.
• Exploration of alternative frameworks for interpreting MVPA results.

Main Results:

• The decoder's dictum, as currently formulated, is false.
• Decodability alone does not provide strong evidence for what information neural patterns represent.
• Existing interpretations of MVPA may overstate their conclusions about neural content.

Conclusions:

• The reliance on the decoder's dictum in cognitive neuroscience warrants re-examination.
• Improved methodologies and interpretive frameworks are needed to draw valid conclusions about neural representations from MVPA data.
• Future research should focus on developing more robust methods for linking decoding success to representational content.