Ubiquitous predictive processing in the spectral domain of sensory cortex

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Canonical microcircuits in the granular cortex suggest specialized computations within six-layer columns.
• Predictive processing theories propose that neuronal firing rates signal surprise, but recent studies challenge this.
• An alternative model, predictive routing, posits that neuronal oscillations encode surprise.

Purpose of the Study:

• To test predictive processing hypotheses in the spectral domain using multi-area laminar recordings in mice.
• To investigate the role of cortical laminae in processing sensory information during a visual oddball task.
• To determine whether neuronal oscillations or firing rates encode surprise signals.

Main Methods:

• Multi-area laminar recordings were performed in mice during a visual oddball task.
• Local field potentials (LFPs) and single-neuron spike rates were analyzed.
• Histological labeling identified cortical laminae for fine-grained analysis of their roles.

Main Results:

• Gamma-band (ɣ) LFP oscillations conveyed feedforward prediction errors in lower sensory cortical areas.
• Alpha/beta-band (⍺/β) oscillations decreased in unpredictable conditions.
• Theta-band (θ) oscillations signaled slower, longer-scale temporal prediction errors.

Conclusions:

• Neuronal oscillations, particularly gamma rhythms, encode surprise signals in the mouse visual cortex.
• The findings support the predictive routing model, where rhythmic timing of spikes, not firing rate, encodes surprise.
• Predictive routing offers a more comprehensive explanation than ubiquitous predictive coding or feedforward adaptation.