- -

TL;DR

Pulvinar allows hierarchy of cortical area to skip hierarchical organisation.
Like writing an angry mail to your CEO, it can be useful for efficient communications.
True, especially in the context of predictive coding.

Context

The brain is a distributed network - anything is a hub. Pulvinar is in the brain, and is also a hub. It is a critical one - losing the hub known as the pulvinar = good chance of schizophrenia.

Important fact : reciprocal connections throughout the cortical hierarchy. The pulvinar doesn’t care about hierarchy.

What is the point of this ? Broadcasting ? Skipping queues ? Balancing everyone’s activity ?
4 possibilities presented in the paper, 2 presented here.

Global workspace

Global Workspace Theory (GWT). Hypothesis on how consciousness arises from widespread information sharing across multiple brain regions.
Integration of information in a “workspace” involving neurons in multiples cortices.
Enables complex cognitive functions like decision-making, planning, and problem-solving.

Key idea: Conscious access results from information becoming globally available to various cognitive processes.

Pulvinar: access to most cortices - useful for this? An exampe: conscious detection of a subliminal visual stimulus. Process:

• Visual signal reaches V1, crosses a neural threshold.
• Activity spreads through the cortex.
• If strong enough, signal propagates to higher levels, reaching the prefrontal cortex (feedforward broadcasting).
• If too weak, signal decays.
• Strong signal causes widespread cortical activity (ignition), reverberating through recurrent interactions.
• Feedback communication relays response back to lower regions.
• Result: Bidirectional communication and long-range connections essential for conscious experience and integration in the nervous system.

Physiological evidence with electrical microstimulation of the thalamus evoking synchronized activity across multiple cortical areas.
Burst-like discharges in the pulvinar suggest that this activity enhances the likelihood of activating cortical neurons, helping them detect environmental changes.

Predictive coding

Predictive coding theory sees the brain as an inference engine. Brain builds an internal model to explain sensory input, making constant predictions about future states.
Model updated by prediction errors, discrepancies between predictions and reality. Neuroanatomically, feedback connections should carry predictions (top-down) and are compared to sensory inputs (bottom-up) to identify errors.

Brain selects representations (predictions) or new inputs (errors) to guide behavior accurately. Bayesian inference suggests the brain weighs information by precision, yet its implementation is debated.
Precision weighting, or inverse variance $ 1/\sigma^2 $ weighting, gain control mechanism, adjusts gain in predictions or errors.

Pulvinar may modulate precision in visual processing, fitting its connectivity and integration timescales.
Especially true when disrupting precision = seeing things that are not here = psychosis = symptom of schizophrenia = common result of pulvinar lesion.

Good sources on this: Kanai et al. [1] for pulvinar and predictive coding; Keller et al. [2] for psychosis in predictive coding terms.

Relevance

Important outstanding questions ! Another review coming on that, near future.

• What role does the pulvinar nucleus play in integrating and segregating cortical networks during complex cognitive tasks?
• Cortico-cortical organizational principles preserved in transthalamic pathways? Specifically through the pulvinar, during complex cognitive tasks?
• Do individuals with schizophrenia have fewer ignition events? How do ignition events in patients with pulvinar lesions compare to healthy individuals? Frequency or other characteristics changed?
• Mechanisms used by the pulvinar to adjust cortico-cortical connections when predicted visual stimuli do not match actual experiences? Influence on perception? Implications of malfunction, especially regarding visual disturbances in disorders like schizophrenia?
• Do transthalamic connections in biological neural networks resemble ‘skip connections’ in deep artificial neural networks? Skip connections bypass intermediate layers to reduce signal dampening and improve efficiency in complex networks. Pulvinar serves the same role in the brain?

References

[1] Kanai, R., Komura, Y., Shipp, S., & Friston, K. (2015). Cerebral hierarchies: predictive processing, precision and the pulvinar. Philosophical Transactions of the Royal Society B: Biological Sciences, 370(1668), 20140169.

[2] Keller, G. B., & Sterzer, P. (2024). Predictive Processing: A Circuit Approach to Psychosis. Annual Review of Neuroscience, 47.

On a personal note

Second pulvinar paper with Nelson, published almost exactly two years after the first one.