Cold Control

CCT is built on David Rosenthal’s Higher-Order Thought (HOT) theory of consciousness. In HOT, a thought or mental state is conscious when there is a "higher order thought" about being in that state.. There are various levels of higher order thoughts, including second order and third order thoughts, but the important distinction is that there is an unbroken chain between higher order thoughts and the first order thought, also known as the intention.

So, in normal thought, there’s an unbroken line between conscious awareness and the unconscious intention to raise the arm.

In hypnosis, CCT posits that this connection is broken because the higher order thought either does not exist, or is suppressed or inhibited in some way.

This lack of intentionality is considered important enough in hypnosis that it’s called the "classic suggestion effect" and CCT attributes it to a lack of metacognitive awareness.

CCT also uses HOT to explain amnesia and hallucination. Amnesia is straightforward — the suggestion to forget the number "4" involves executive control over counting, but the subject will deny any such conscious awareness. Hallucination, especially positive hallucination, is a little more complicated. In CCT, a positive hallucination is an intention to imagine seeing something, and then an inaccurate second order thought so that instead of correctly recognizing the intention as mental imagery, it is recognized as a visual perception.

Hallucinations are recognized as more difficult for people, and the paper provides two reasons for this: it’s a deeper suppression, and it requires a high first order effort — generating a vivid mental image and simultaneously tweaking the higher order thought so it is perceived as reality is complicated for most people.

Rosenthal in Higher-order awareness, misrepresentation and function argues that higher order awareness (HOA) should be distinguished from classic metacognition, arguing that metacognitive judgements are typically conscious themselves and predict future recall, where as HOA simply make us aware of a current state and are rarely conscious themselves. Intriguingly, Rosenthal says that unlike metacognition, "it is unlikely that there is much, if any, utility to mental states' being conscious over and above the utility those states have when they are not conscious." I don’t think this can be true, as the core concept of hypnosis is that manipulating HOA can have great utility.

CCT makes a number of predictions, some of them confirmed, some of them not.

There is a follow up paper: Phenomenological control as cold control. The title is somewhat inaccurate, as it does not prove that phenomenological control is cold control; instead, this paper is a stress test: it attempts to find various things that should not be possible if Cold Control Theory is true.

The problem with negative predictions is that they are very hard to prove. If there were a positive counterexample, it could disprove it, but every negative result only fails to refute it. This means that a theory based on negative results is effectively unfalsifiable in practice.

A good overall summary of objections to the higher order approach is at the Stanford Encyclopedia of Philosophy. These are primarily philosophically based arguments, so I’ll just pick one out to give you the idea.

Ned Block argues The higher order approach to consciousness is defunct. His argument is the empty HOT problem: you can technically have a HOT without an actual first order state. HOT has two conditions — a sufficient condition, and a necessary condition. According to HOT’s sufficient condition, that HOT alone should be enough to produce a conscious experience. But according to HOT’s necessary condition, any conscious episode must itself be the object of a HOT. These can’t both be true.

Block also says that HOTs don’t matter if they aren’t connected to anything:

Rosenthal’s reply is that a theory of consciousness explains subjective appearance, not underlying mental reality. A HOT is sufficient for a state to seem to occur in one’s stream of consciousness — not for the first-order state to actually exist. Conscious agony in the empty HOT case is real as a subjective appearance; whether there’s actual agony underneath is a separate question.

The conversation between Block and Rosenthal continues, if you’re into that.

But this still doesn’t really ground HOT in anything. What does it actually mean, in practice, to have a higher order thought?

HOT occupies an interesting position, as it’s a theory based on philosophy that intentionally does not map to neurons. In Two concepts of consciousness, Rosenthal says:

As such, the work of providing a view of HOT through cognitive neuroscience falls to others. This is…​ awkward. Technically, any architecture that supports a "higher-order representation" qualifies, and so HOT’s predictions at the neural level are too broad to be falsifiable.

Lau in Empirical support for higher-order theories of conscious awareness says "this interpretation is accepted by many authors [2,6] including first-order theorists [12,13]" but when you look closer at the citations, it’s a self-citation, Kriegel (a philosopher), and Block (also a philosopher, who accepts the prefrontal interpretation but is arguing against HOT). Looking at the actual neuroscience citations in the paper, it’s Relative blindsight in normal observers and the neural correlate of visual consciousness (Lau again), Experimental and theoretical approaches to conscious processing (GWT, not HOT), Neural correlates of consciousness in humans (theory neutral), and Pattern of neuronal activity associated with conscious and unconscious processing of visual signals (also theory neutral).

Of these, the Blindsight paper is the most important. Using visual masking, they created conditions where subjects' ability to discriminate between two figures stayed the same, but subjects reported being consciously aware more frequently in one condition than the other. The higher awareness was associated with dorsolateral prefrontal cortex activity and not early visual areas. This is the most important result, because it directly distinguishes HOT from first-order theories.

The first order theories:

However, it doesn’t distinguish HOT from the other higher order competitors, which are:

In addition to the above, there’s also Global Workspace Theory (GWT) and Information Integration Theory (IIT), which are not higher order theories but are frequently cited. And then, finally, there’s predictive processing and active inference.

In fact, there are so many theories that there are theories on integrating the theories. From Just one moment: Unifying theories of consciousness based on a phenomenological “now” and temporal hierarchy there are two diagrams that say everything about how the theories interact, arguing that any or all of them could be correct.

But back to the HOT implementation. What actually is a higher order thought?

To Lau, it’s a higher order representation. This is a subtle shift, because you can’t measure a thought but you can measure a representation. This feeds into Lau’s Higher-order statistical inference view theory, which makes it a little confusing as to whether Lau is advocating for HOT or his own theory.

Lau’s theory is fully explained in A higher order Bayesian decision theory of consciousness. Instead of a thought, there’s a Bayesian decision process: consciousness arises when the DLPFC correctly sets a criterion based on learned higher-order representations of internal signal statistics.

Going down to a lower level, there are neurons which specifically light up at a metacognitive level, much like how there’s a neuron that lights up every time you think of Bill Clinton, or how you can tweak the Golden Gate Bridge neuron in a large language model.

Importantly for hypnosis, Lau’s theory is organized around the concept of dissociation and signal detection theory, using detection sensitivity (d') and the criterion for detection (c). When there’s a disconnect between the objective performance (d') and the reported consciousness (c), there’s dissociation.

He identifies three distinct forms of dissociation, each caused by a different failure mode of higher-order representation:

Lau’s framework is specifically about perceptual consciousness (criterion-setting for detecting external stimuli), and he doesn’t extend it to agentive consciousness (awareness of one’s own intentions), so it technically doesn’t bridge the gap to Cold Control Theory. Nevertheless, this is as close as I can get to grounding Cold Control Theory down to the neuron level, so I’ll take it.

CCT offers a coherent account of hypnotic phenomena grounded in Rosenthal’s HOT framework. The core insight is reasonable, philosophically defensible, and generates testable predictions.

But…​ there’s not a lot that points to CCT specifically as The One True Answer. The neural evidence for higher order processing is real and well established, but there are many theories to choose from amidst increasing friction between philosophers and scientists. Rosenthal’s reframing of HOT as a theory of subjective appearance doesn’t help, and HOT technically doesn’t even have a horse in the race if you count Lau’s theory as distinct from Rosenthal’s theory.

That being said, Lau’s Bayesian framework gets closest to a mechanistic account: consciousness as criterion-setting based on learned higher-order representations of internal signal statistics, localized to the DLPFC. This grounds the dissociation between objective performance and subjective awareness in concrete neural terms. But the framework addresses perceptual consciousness and has not been extended to agentive consciousness, which is what CCT specifically requires.

I think what I would want most from CCT is an explicit bridge to Lau and an exploration of how dissociation works on a neural level, especially comparing dissociation in hypnotic vs non-hypnotic contexts. Until then, CCT remains a plausible and unfalsified theory — which is not quite the same thing as a confirmed one.