Stimulus complexity shapes response correlations in primary visual cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Visual Perception

Background:

• Spike count correlations (SCCs) are common in sensory cortices and exhibit complex structure.
• Current visual perception models often overlook SCCs, focusing instead on average neuronal responses.
• Structured internal dynamics are believed to underlie SCCs, but their precise role in perception is unclear.

Purpose of the Study:

• To investigate how hierarchical probabilistic inference in visual perception shapes SCCs.
• To determine if top-down inferences modulate bottom-up processing and influence SCC structure.
• To predict and experimentally verify that stimulus complexity alters SCC fine structure.

Main Methods:

• Designed experiments using natural and synthetic stimuli to record SCCs in the primary visual cortex (V1) of macaques.
• Analyzed the fine structure of SCCs and their dependence on stimulus identity and statistical properties.
• Compared experimental SCC patterns with predictions from phenomenological models.

Main Results:

• The fine structure of SCCs in V1 was specific to the identity of natural stimuli.
• Changes in SCCs were independent of changes in average neuronal response (mean response).
• Stimulus specificity of SCCs was directly manipulated by altering high-order structure in synthetic stimuli.

Conclusions:

• The stimulus dependence of SCCs in V1 is not explained by simple models of neuronal activity.
• Structured internal dynamics, arising from hierarchical probabilistic inference, naturally explain observed SCC patterns.
• High-level feature inferences modulate low-level feature processing, creating structured dynamics and SCC patterns in V1.