Electrophysiology

Prior expectation reduces the trial-by-trial variation of pursuit direction and interneuronal correlations in macaque area MT (2019 SfN abstract)

Behavioral Results

Trial-by-Trial correlation between neurons

Trial-by-Trial correlation between neuron and behavior

When we interact with the environment and make appropriate behavioral responses, our brain relies not only on the incoming sensory information but also on prior knowledge based on recent experience. The influence of prior knowledge becomes more prominent when the sensory evidence is weak or ambiguous. In this study, we investigated the competition between the sensory information and prior knowledge of incoming motion direction, using smooth pursuit eye movements in macaque monkeys. We controlled the monkey’s prior expectation of motion direction and the strength of visual motion (100% vs 8% luminance contrast) independently. We found that the trial-by-trial variation of pursuit direction was significantly reduced when prior knowledge for motion direction was strong, especially if the sensory evidence for the motion was weak.

To understand the neural mechanisms of the effect of prior expectation on the sensory-motor behavior, we recorded responses of neurons in the middle temporal visual area (area MT), to the same visual stimulus in two different prior expectation conditions. We found that the firing rate, the direction tuning properties, and Fano factor of each neuron were not affected by prior expectation. However, the trial-by-trial correlation between activities of area MT neurons was significantly reduced by prior expectation in a 100% contrast condition. In agreement with the change in the inter-neuronal correlations, prior expectation also reduced the trial-by-trial correlation between single neural activity and pursuit direction variation in a 100% contrast condition. This result suggests that the reduction of behavioral variation by prior expectation can be partly explained by the reduction of correlated inter-neuronal variability in population activity of area MT neurons.