What This Changes in the Therapy Room
Cognitive behavioral therapy already works. Nothing here asks anyone to stop doing what they're doing or to add a technique. What the control reading offers is a layer underneath, and one thing the standard diagram can't show.
Before this framework goes anywhere near application it has to line up with CBT, because CBT is the model I like, it's the model most people reading this relate to, and a framework that can't be stated in its terms isn't worth much.
It lines up pretty well. Term for term, with one piece left over that turns out to matter.
What is the cognitive model, in one diagram?
An event happens. You run a cognitive assessment of it, which surfaces as automatic thoughts. That assessment produces an emotion. The emotion produces behavior. Underneath the assessment sit the schemas, the core beliefs, and those are what the assessment is running on. Not all core beliefs are maladaptive. It's the maladaptive ones that get in the way.
Here's the diagram I learned from my mentor Ryan Sheade, LCSW, and have put in front of students and clients for years.
When the schemas are maladaptive, five show up over and over.
I'm not good enough. That one arrives as shame.
I'm not worthy of love.
I'm insignificant, or unimportant.
I'm unsafe, or I can't keep myself safe.
I'm defective.
They form early, roughly five to ten years old. From the child's point of view, at that age adults know everything. The child has just become cognitive enough to build general rules out of specific events, and not yet cognitive enough to consider that a rule might be about somebody else.
Everything that happens is therefore about them. Bad things happening means something is wrong with them, the rule gets written down and filed, and it keeps running for decades.
How do the CBT terms map onto a control loop?
One to one, with nothing left standing. The event is sensory input. The cognitive assessment is the estimator, and the schemas are the priors it runs on. The emotion is the error signal, the gap between the estimate and the set point being defended. The behavior is the controller's output. That mapping is not a metaphor and it is not a rename. Each pair is doing the same mechanical job.
Then the piece the standard diagram doesn't have.
The behavior produces a result, and the result is the next event. The chain isn't a chain. It's a loop, and it runs continuously.
Two things follow, and the second is why the redraw was worth doing.
The step from assessment to emotion stops being a mystery. In the standard teaching that arrow is the part that gets waved at, and students accept it because the sequence obviously works even when nobody can say how. Under the control reading the emotion is a comparison and its magnitude is the size of a discrepancy.
Which is why the same thought produces a different feeling in two people, and a different feeling in one person on two different days. The thought didn't change. What it was being measured against did.
That also gives four of the five core beliefs a job and separates out the fifth. Not good enough, not worthy of love, insignificant, defective are all priors, biasing what the estimator concludes from thin evidence.
But I'm unsafe isn't doing that. It sets how much estimated danger is tolerable before the system calls it an error. That's a set point rather than a prior, and it produces a different presentation and needs a different intervention. Under the standard model all five sit in the same box, which is a real cost of the generic cognitive model's economy rather than a flaw in it (Beck & Haigh, 2014).
Why does a wrong belief survive contact with reality?
Because it runs a loop that manufactures its own evidence. A left-to-right chain ends at behavior, so reality never gets to answer. Close the loop and you can see the belief producing the very result that confirms it. That's what a maladaptive core belief is mechanically: not a wrong idea sitting in storage waiting to be corrected, but a prior generating the data that keeps it accurate.
Work an example. The prior is I'm not worthy of love.
The event is a partner who's quiet at dinner. The estimator, running on that prior, reads quiet as withdrawal, and withdrawal as the front edge of leaving. The set point is being securely connected. The gap is large, so the emotion is large, somewhere between fear and shame.
The controller has high gain and no damping, so the reaction is to press. What's wrong. Are we okay. You'd tell me, right. Three hours of it.
The result is a partner who was tired at the start of the evening and is now irritated and genuinely pulling back.
That result is the next event.
Nobody in that loop did anything irrational. Every step follows correctly from the step before it. The prior was wrong when the evening started and accurate by the time it ended, because the loop went and made it accurate.
Which is also why arguing with the belief so often does nothing. You're disputing a conclusion the system re-derives from fresh data every week.
DON'T say: "You're being irrational." SAY: "I think your instrument is reading off."
The anxious person usually isn't over-reacting to their situation. They're reacting accurately to an estimate of their situation that happens to be wrong.
Does the cognitive assessment do two jobs?
Probably, and the standard model treats them as one. An assessment tells you what's happening and how much it matters in the same breath. "He's pulling away" is an estimate. "He's pulling away and this is the end of everything" is an estimate with a gain setting attached to it. Appraisal research has said something close to this for decades, holding that appraisal determines intensity and not only content.
This is the question I keep coming back to, and I don't have an answer to it.
Smith and Ellsworth found appraisal dimensions that predict which emotion arises and how strongly (Smith & Ellsworth, 1985). If that's right, the assessment box is doing two mechanically different jobs.
Which opens something the diagram doesn't show. A maladaptive core belief may be able to enter the loop in three different places.
As a prior. I'm defective biases what the estimator concludes from thin evidence. Ambiguous feedback gets read as criticism.
As a set point. I'm unsafe changes what counts as an error at all. Less estimated danger is tolerable, so the same situation opens a gap that wouldn't open in somebody else.
As a gain. I'm not worthy of love may not distort the reading whatsoever. It may set how hard the system acts once an error exists, because it reads the stakes as total. The estimate is accurate. The response is maximum.
Same five beliefs. Three mechanisms. Three different treatments, which is the argument of this whole series arriving from a direction I hadn't planned on.
I can't settle it from here and as far as I can tell nobody has. What's worth saying is that it's separable in principle. A belief operating as a prior predicts biased interpretation of ambiguous evidence and roughly normal handling of unambiguous evidence. A belief operating as a gain predicts accurate interpretation across the board and an outsized response to both.
Hand the same person ambiguous and unambiguous versions of the same situation, and the two accounts come apart. That's a study, and it's a small one.
What four questions does this hand a practitioner?
Is this belief inaccurate, or accurate and acted on too hard? Is it wrong, or held with too much confidence? Is the person failing to update, or being prevented from updating? And what environment taught this setting, does it still exist? None of these require a new technique. They sort people who look identical at the symptom level into different repairs.
Take them one at a time.
Inaccurate, or acted on too hard. Cognitive restructuring addresses the second and does very little for the first. If the reading is right and the response is disproportionate, disputing the thought is aimed at the wrong dial.
Wrong, or held too confidently. Content and confidence come apart. Behavioral experiments work on confidence, which is why the experiment beats the argument about the experiment (Craske et al., 2022).
Failing to update, or prevented from updating. Safety behaviors block disconfirmation, which Salkovskis worked out for panic decades ago (Salkovskis, 1991). The prior isn't stubborn. It's starved. Craske's inhibitory learning account of exposure makes the same claim in different words: exposure works by expectancy violation rather than by habituation (Craske et al., 2014). That's the closest existing bridge between CBT and this framework, and it was built without anyone needing the control vocabulary.
What environment taught this. Usually answerable in a session, and it changes the conversation from correction to recognition. The setting was right where it was learned.
Next, the hardest case I know of, and the one I worked this out on first.
References
Beck, A. T., & Haigh, E. A. P. (2014). Advances in cognitive theory and therapy: The generic cognitive model. Annual Review of Clinical Psychology, 10, 1–24. https://doi.org/10.1146/annurev-clinpsy-032813-153734
Craske, M. G., Treanor, M., Conway, C. C., Zbozinek, T., & Vervliet, B. (2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, 58, 10–23. https://doi.org/10.1016/j.brat.2014.04.006
Craske, M. G., Treanor, M., Zbozinek, T. D., & Vervliet, B. (2022). Optimizing exposure therapy with an inhibitory retrieval approach and the OptEx Nexus. Behaviour Research and Therapy, 152, 104069. https://doi.org/10.1016/j.brat.2022.104069
Salkovskis, P. M. (1991). The importance of behaviour in the maintenance of anxiety and panic: A cognitive account. Behavioural Psychotherapy, 19(1), 6–19. https://doi.org/10.1017/S0141347300011472
Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in emotion. Journal of Personality and Social Psychology, 48(4), 813–838. https://doi.org/10.1037/0022-3514.48.4.813
The full documents, free and ungated