Bayesian surprise shapes neural responses in somatosensory cortical circuits

Area of Science:

• Neuroscience
• Computational Neuroscience
• Sensory Processing

Background:

• Bayesian inference is crucial for sensory decision-making, but its neural basis is not well understood.
• The specific neural circuits implementing probabilistic computations in the brain remain largely unknown.

Purpose of the Study:

• To investigate the neural circuitry underlying Bayesian inference in sensory decision-making.
• To identify how and where Bayesian surprise is encoded in the brain's neural activity.

Main Methods:

• Extracellular neural activity was recorded along the mouse somatosensory pathway.
• Specialized sensory stimulation paradigms were used to isolate responses to Bayesian surprise.
• Analysis focused on multiunit spiking activity, evoked potentials, and spectral power (gamma and alpha bands).

Main Results:

• Bayesian surprise is encoded in laminar cortical circuits of early sensory areas, specifically the primary (S1) and secondary (S2) somatosensory cortices.
• Layer 6 of S1 and S2 showed the highest sensitivity to surprise via multiunit spiking and evoked potentials.
• Gamma power in S1 (layer 2/3) and alpha power in S2 (layers 2/3 and 6) scaled with surprise in an NMDAR-dependent manner.

Conclusions:

• The somatosensory cortex precisely represents Bayesian surprise in both space and time.
• Bayesian inference appears to be a fundamental mechanism within cortical processing.
• The findings suggest a shift in understanding sensory processing from passive reception to active probabilistic inference.