Stimulus-choice (mis)alignment in primate area MT
Yuan Zhao1, Jacob L Yates2, Aaron J Levi3
1Department of Neurobiology and Behavior, Stony Brook University, Stony Brook, New York, United States of America.
Plos Computational Biology
|May 19, 2020
Summary
Neural activity in sensory areas correlates with perception. Unexpectedly, choice information in the middle temporal (MT) area is misaligned with stimulus encoding, suggesting feedback rather than optimal feedforward readout.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Sensory Perception
Background:
- Neural activity in sensory areas often correlates with an animal's perceptual report, especially near the perceptual threshold.
- This phenomenon is typically explained by feedforward readout of neural activity by downstream decision-making circuits.
- However, the interpretation of choice-correlated activity is ambiguous and requires understanding population-wide neural correlations.
Purpose of the Study:
- To investigate the nature of choice-correlated activity in the middle temporal (MT) area during perceptual decision-making.
- To differentiate between sensory-driven and choice-correlated activity within the population code.
- To understand how neural population activity relates to stimulus information and decision outcomes.
Main Methods:
- Utilized a statistical nonlinear dimensionality reduction technique on single-trial ensemble recordings from the MT area.
- Extracted low-dimensional latent factors representing population-wide neural fluctuations.
- Analyzed the alignment of choice-correlated signals with stimulus encoding dimensions.
Main Results:
- Extracted latent factors strongly encoded stimulus direction with a temporal signature similar to single MT neurons.
- Found a significant component of choice information residing in a subspace orthogonal to the stimulus representation.
- Observed that this misaligned choice information is inconsistent with an optimal feedforward readout model.
Conclusions:
- The misaligned choice information suggests a mechanism where feedforward sensory information coexists with decision-making processes.
- The temporal dynamics of these signals indicate potential feedback from downstream areas to the sensory population.
- Hypothesize that this non-corrupting choice-correlated feedback may play a role in learning or reinforcing sensory-motor relationships.


