What predictive processing expects to find in the brain
Predictive processing treats the brain as a hierarchical prediction machine: higher levels generate expectations about the causes of sensory signals, and lower levels send back the mismatch, or prediction error, between what was expected and what arrived. Consciousness, on this view, is not produced by one special structure but by this ongoing inferential exchange. So the neural correlates it predicts are distributed, hierarchical, and dynamic, tied to how predictions and errors are balanced rather than to a single anatomical hotspot [1][3].
This is a meaningfully different prediction from theories that nominate a specific mechanism. Global workspace theory points to a system that broadcasts information widely across the brain; higher-order theories point to representations of one's own mental states; integrated information theory points to a measure of how unified a system's causal structure is; recurrent processing theory points to feedback loops in sensory cortex [1][2]. Predictive processing instead predicts correlates spread across the hierarchy, which makes it harder to pin down but also harder to falsify.
Where the theories actually disagree
The clearest disagreement is about what kind of neural event should track consciousness. If a global broadcast is the key, then widespread ignition-like activity should mark conscious access. If recurrence is the key, then local feedback loops should. Predictive processing sits closer to the recurrence family, since it depends on bidirectional prediction and error signals, but it adds the claim that what matters is the content of the prediction, not just the presence of a loop [1][3].
A second disagreement concerns whether consciousness can be separated from the mechanisms that support report and behavior. Some frameworks treat conscious states and their neural correlates as potentially coming apart, while others tie consciousness tightly to the machinery of access and report; this remains actively contested [3]. Predictive processing's emphasis on inference rather than broadcast makes it more compatible with the idea that experience and report can diverge, though that is a matter of interpretation rather than settled result [3].
Why the evidence is hard to read off the brain
The practical obstacle is that the theories often predict overlapping patterns. A burst of recurrent activity could be read as recurrent processing, as part of a global broadcast, or as prediction error being resolved. Because the theories make some of the same neurobiological commitments, a single experiment rarely isolates one [1]. Reviewers of the field note that the theories need further development before they can be cleanly tested against each other [1].
A 2022 public debate among proponents of five major theories, including predictive processing, made the gap explicit: there was more controversy than agreement on what consciousness is, how to identify conscious states, and what any theory must deliver [2]. That does not mean the theories are interchangeable, but it does mean that a clean neural-correlate comparison is still a goal rather than a finished result [2].
What other frameworks add to the picture
Some frameworks go beyond cortical prediction altogether. The Default Space Theory, for instance, proposes that consciousness rests on synchronized bioelectric oscillations across neural and visceral systems, and it explicitly contrasts this with predictive processing's hierarchical inference account [5]. That is a different kind of prediction: it points to body-wide synchrony rather than cortical prediction error as the correlate to look for [5].
Computational treatments of predictive processing also try to specify the conditions under which neural correlates appear, rather than just naming brain regions [4]. Taken together, the theories agree that consciousness has neural correlates but disagree sharply about what those correlates are and how to recognize them, which is why the question remains open rather than answered [1][2][4].
Sources used in this answer
Theories of consciousness
Reviews four major theories of consciousness, including predictive processing, and argues they must be developed and tested against each other before they can be empirically distinguished.
Unpacking the complexities of consciousness: Theories and reflections
A 2022 debate among proponents of five theories, including predictive processing, found more controversy than agreement on basic questions about what consciousness is and how to identify conscious states.
Predictive processing and consciousness
Examines predictive processing and consciousness, noting that conscious states and their neural correlates may come apart, a view that remains contested.
Computational Explanation of Consciousness: A Predictive Processing-based Understanding of Consciousness
Offers a computational explanation of consciousness based on predictive processing and discusses the conditions under which neural correlates of consciousness appear.
Synchrony in Neural and Visceral Systems: A Bioelectric Basis of Consciousness
Proposes the Default Space Theory, grounding consciousness in synchronized bioelectric oscillations across neural and visceral systems, and contrasts this with predictive processing's hierarchical account.