Related Experiment Video
Updated: May 9, 2026
06:22
Stochastic Noise Application for the Assessment of Medial Vestibular Nucleus Neuron Sensitivity In Vitro
Published on: August 28, 2019
Stimulus-dependent variability and noise correlations in cortical MT neurons.
Adrián Ponce-Alvarez1, Alexander Thiele, Thomas D Albright
1Theoretical and Computational Neuroscience Group, Center of Brain and Cognition, Universitat Pompeu Fabra, 08018 Barcelona, Spain. adrian.ponce@upf.edu
Summary
Neural population codes use firing rates, but variability and correlations matter. This study reveals that these second-order statistics are tuned, carrying information that enhances neural encoding accuracy.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Information Theory
Background:
- Neural systems encode sensory input via neuronal firing rates.
- Trial-by-trial variability and noise correlations can impact neural code information capacity.
- Stimulus presentation may reduce variability and correlations, potentially improving encoding.
Purpose of the Study:
- To investigate whether second-order neural statistics (variability and noise correlations) are tuned to stimuli.
- To determine if these tuned statistics carry information.
- To assess the impact of tuned second-order statistics on population coding accuracy.
Main Methods:
- Analysis of rate variability and noise correlations in directionally selective middle temporal (MT) neurons.
- Modeling a stochastic recurrent neural network to simulate the emergence of tuned statistics.
- Information theoretic analysis to quantify information carried by second-order statistics.
Main Results:
- Rate variability and noise correlations systematically vary with stimulus direction in MT neurons, exhibiting tuning curves.
- Tuned second-order statistics emerge in the modeled stochastic recurrent network.
- Information theoretic analysis demonstrates that second-order statistics contain information that improves population code accuracy.
Conclusions:
- Second-order neural statistics, specifically rate variability and noise correlations, are not constant but are tuned to stimulus properties.
- This tuning is reproducible in computational models of neural networks.
- Tuned second-order statistics play an informative role, enhancing the overall accuracy of neural population codes.
