Why neurons mix: high dimensionality for higher cognition.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Neurons often exhibit mixed selectivity, responding to various task-relevant variables simultaneously.
• This mixed selectivity is crucial for the computational capacity of neural representations.
• Understanding neural representations' dimensionality is key to decoding brain function.

Purpose of the Study:

• To review the conceptual and theoretical framework of mixed selectivity in neural representations.
• To present experimental evidence supporting high-dimensional neural representations.
• To discuss implications for future experimental designs in neuroscience.

Main Methods:

• Review of existing literature on mixed selectivity and neural representation dimensionality.
• Analysis of theoretical models explaining the computational role of mixed selectivity.
• Synthesis of experimental findings on neural population activity.

Main Results:

• Mixed selectivity leads to high-dimensional neural representations.
• High-dimensional representations with mixed selectivity allow for complex computations via linear readouts.
• Specialized neurons lead to low-dimensional representations, limiting computational flexibility.

Conclusions:

• Mixed selectivity is fundamental for the brain's ability to process complex information.
• High dimensionality of neural representations is supported by experimental evidence.
• Future experiments should consider the implications of mixed selectivity and dimensionality for neural coding.