A normalization model suggests that attention changes the weighting of inputs between visual areas.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Divisive normalization models accurately predict neuronal responses across sensory, association, and motor areas.
• Visual attention modulates neuronal gain and trial-to-trial variability, including correlations between neurons in different visual areas like V1 and MT.
• Previous work showed attention increases V1-MT correlations, and a normalization model extension explained this.

Purpose of the Study:

• To provide causal evidence for the nonlinear relationship between V1 and MT activity.
• To investigate the mechanisms by which attention influences correlations between V1 and MT neurons.
• To test whether attention modulates connection weights or individual neuronal responses.

Main Methods:

• Electrical microstimulation in V1 paired with recordings in MT.
• Utilizing a divisive normalization model.
• Conducting recording and microstimulation experiments to analyze V1-MT correlations under attention.

Main Results:

• Causal evidence confirmed a nonlinear V1-MT relationship well-described by divisive normalization.
• Attention-dependent V1-MT correlations are better explained by changes in connection weights between V1 and MT.
• A mechanism modulating connection weights, not individual area responses, underlies attention's effect on V1-MT correlations.

Conclusions:

• Divisive normalization can explain inter-areal neuronal interactions.
• Attention modulates connection weights between visual areas, influencing neuronal correlations.
• This framework enables using multi-area recording and stimulation to probe neural computation mechanisms.