CBT on the thought you never noticed thinking

Your mood has been bad since about four o'clock and you could not say why. Nothing happened. The afternoon was ordinary.

The question cognitive behavioural therapy asks here is a strange one the first time you hear it. Not why do you feel like this, which invites a theory. It asks: what went through your mind just before it dropped?

Most people, asked that, can find the sentence. It was short. It was there for about a second. It did not feel like an opinion, it felt like noticing something true, which is exactly why nobody argued with it.

CBT is built on the claim that the sentence can be retrieved, written down, and checked.

In one sentence

What you feel follows from how you read a situation rather than from the situation itself, and the reading is a thought like any other: catchable, writable, and open to being tested against what actually happened.

The problem it tries to solve

The problem is not that you feel bad. It is that the feeling arrives with no visible cause, so there is nothing to disagree with.

Someone does not reply and you are already certain of what it means. You get one piece of feedback in an hour of praise and it is the only piece you can remember at midnight. You know, in some other part of your head, that you are being unfair to yourself, and knowing it changes nothing, because the unfairness never presented itself as an argument you could take up.

Beck's name for the sentence is automatic thoughts, and the word doing the work is automatic. Not irrational, not unconscious, not hidden. Just fast enough, and familiar enough, that you never stopped on it. You can usually report it a second later if somebody asks. Nobody asks.

That is the specific gap this tradition works in: between a thought being available and a thought being examined. Everything else it does follows from the claim that the gap is small, and that crossing it is a skill rather than an insight.

How it sees the mind

The model is a sequence, and it is deliberately plain: a situation, a thought about the situation, and then the feeling and the behaviour that follow from the thought rather than from the situation. Two people, same meeting, different sentences, different evenings.

Under the automatic thoughts the model puts two deeper layers. Intermediate beliefs are the rules and assumptions you operate by, usually in the shape of a conditional or an obligation: if I ask for help, they will think I can't cope; I should be able to do this by now. Core beliefs are the flat statements about yourself, other people and the world that the rules protect you from having to face: I am incompetent, I am unlovable, people leave. In Beck's vocabulary the organised version of these is a schema, a term he took from Piaget: not a thought you have but a filter that decides which thoughts you get.

The three level naming needs attributing correctly, because it does not all come from the same place. Schema and automatic thought are Beck's. Intermediate beliefs, as a layer with its own name sitting between the other two, is the teaching formulation of Judith Beck, whose textbook is how most clinicians now learn the model. In the 1979 manual the assumptions and the core beliefs are not yet pulled apart this way.

The last piece is the list of recurring shapes. Beck called them cognitive errors. The list that circulates outside clinical settings is David Burns's, from 1980, and it has ten entries: all-or-nothing thinking, overgeneralization, mental filter, discounting the positive, jumping to conclusions, magnification or minimization, emotional reasoning, should statements, labeling, and personalization. Two of those are containers rather than single items. Jumping to conclusions splits into mind reading and fortune telling, and catastrophizing is the alternative name Burns gives to magnification, not a separate entry.

That detail matters, because almost every list you will meet since, in books and in software, has promoted the sub-items to entries of their own and quietly dropped others. The ten cognitive distortions names a family, not a fixed set. What every version of it has in common is the only part that matters: these are named shapes of inference, and that is all they are.

Two things this model asserts, and one it does not ask you to believe. It asserts that thoughts are accessible to report, and that changing the reading changes the feeling. It does not ask you to accept schema theory to use any of it. You can keep a thought record for a month without having a view about whether core beliefs are stored structures or a useful metaphor, and the record does the same thing either way. That separation matters more here than in most traditions, because this one makes empirical claims and can therefore be wrong in public. Where that has happened is further down, in its own literature, not here.

The central move

Treat the thought as a hypothesis rather than a report.

That is one move with two halves, and CBT says so in its own name.

The cognitive half tests the thought on paper. You write down the situation, the sentence your mind produced, and a number for how much you believe it. Then the evidence that supports it, and, separately, the evidence that does not fit. Then a reading that accounts for both. Then the number again.

The behavioural half tests it in the world instead. If the thought is if I say I don't understand, they'll think less of me, that is a prediction, and predictions have outcomes. A behavioural experiment is the deliberate act of finding out. The therapy's other behavioural components, exposure and activity scheduling among them, are variations on the same logic: put the belief where reality can answer it.

What holds the halves together is the conversion. A conviction is something you carry. A hypothesis is something that can come out wrong, and the whole apparatus exists to make coming out wrong possible.

Which of the two halves is actually responsible for people getting better is a live question inside the field.

A case where the move is hard

Not a case of a patient. The hard case for CBT is that the strongest challenges to its central move have come from inside its own literature, from people running its own trials.

The dismantling study. In 1996 Neil Jacobson, an American clinical psychologist, and colleagues took the standard treatment for depression apart and tested the pieces separately: behavioural activation alone, behavioural activation plus work on automatic thoughts, and the full package including work on core beliefs. The prediction was ordinary: more components, more effect. The result was not. The behavioural component on its own did as well as the complete treatment, both at the end of therapy and at the six month follow-up, and a separate follow-up two years on reported the same (Gortner and colleagues, 1998). The part that looks most like the central move, examining and testing the thought, added nothing the study could measure.

The mediation question. If the model is right, symptoms should improve because the thinking changed, and in that order. Richard Longmore and Michael Worrell, two clinical psychologists, reviewed what the evidence actually showed and put the question in the title of their 2007 paper: do we need to challenge thoughts in cognitive behaviour therapy? Their answer was that the component and process evidence did not establish it.

That one did not stand unanswered either. Stefan Hofmann, a clinical psychologist who has spent a career on both sides of this question, replied in 2008 that the review rested on misconceptions about what the cognitive model actually claims, and Longmore and Worrell answered him in the same issue. Either paper read without the other settles the question, in whichever direction it is read first.

This is the same structural problem the ACT literature has, from the other direction, and both traditions have it because both made a claim specific enough to check.

The effect that shrank. In 2015 Tom Johnsen and Oddgeir Friborg, two Norwegian psychologists, pooled decades of trials of cognitive therapy for depression and reported that the measured effect had been falling steadily since the late 1970s. The finding was contested immediately, on the grounds of how the trials were selected and combined, and the argument has not closed. Alongside it sits a separate and better established problem: Pim Cuijpers, a Dutch researcher who specialises in the meta-analysis of depression treatments, and colleagues have shown repeatedly that once you adjust for studies that were never published and for trials at high risk of bias, the effects across psychotherapies for depression come down substantially.

And the answer from the other side, because leaving it out would misrepresent where the argument stands. Behavioural activation has since been tested at scale and has held. Sona Dimidjian and colleagues, in a trial of 241 people in 2006, found it did better than cognitive therapy among the more severely depressed. David Richards and colleagues established its non-inferiority in 2016 in a trial of 440, published in the Lancet.

None of that refutes the cognitive model, and the reason is inside Jacobson's own numbers rather than in anybody's commentary. Both of the shorter conditions, the behavioural one on its own and the one that stopped before the work on core beliefs, changed negative thinking as much as the complete treatment did. Something in there was altering how these people thought. It was not the component designed to work on their thinking directly.

So what is unsettled is not whether the treatment works. It is whether it works for the reason the model gives.

That is the honest position of this tradition today, and it is a strange kind of vulnerability. That CBT is the most researched form of psychotherapy there is has been put in print by Daniel David, Ioana Cristea and Stefan Hofmann (2018); the commentary published against them disputed what that volume of research entitles anyone to conclude, not the volume (Fernández-Álvarez and Fernández-Álvarez, 2019).

And the instruments doing the finding are its own. The dismantling study, the meta-analyses, the corrections for bias: every one of them was built by people inside this tradition, published in its journals, using the method it brought into the room in the first place. A field that makes itself checkable gets checked, and there is no comfortable version of that.

What you actually practice

The unit is the exercise done between sessions, not the session, and the stated aim is that the person ends up doing it without the therapist. In its own phrase, becoming your own therapist.

Cognitive, on paper:

• Catching an automatic thought at the moment the feeling shifts, which is what the what went through your mind just then question is for
• Keeping a thought record: situation, thought, belief rating, evidence for, evidence against, alternative reading, belief rating again
• Naming the pattern, matching the sentence against the list of cognitive errors
• Socratic questioning, the one item here that happens in the session rather than between them: the therapist asking rather than telling, so that the answer is arrived at rather than received
• Following the downward arrow, taking a surface thought and repeatedly asking what it would mean if it were true, to reach the rule underneath

Behavioural, in the world:

• Running a behavioural experiment: stating the prediction, doing the thing, recording what happened
• Scheduling activity and tracking what it does to mood
• Approaching what is avoided, gradually and by agreement, which is the anxiety half of the treatment

What these are for, according to the tradition: not relief in the moment. The point is to change what you expect to be true, and to leave behind a habit of checking that outlasts the course of treatment. Relapse prevention is stated as an aim, not a side effect.

What people often misunderstand

That it is positive thinking. It inverts the thing completely. The stated target is accuracy, not comfort, and an honest test can land on the worse answer. Sometimes the evidence for the thought is good. Sometimes the alternative reading is this went badly and it is fixable, which is not a nicer thought, only a truer one. A procedure whose whole design is that you might be wrong is not an affirmation exercise, and the difference is not cosmetic: replacing the test with reassurance produces the version of CBT that people try, find hollow, and abandon.

The reason this particular misunderstanding attaches to CBT and not to its neighbours is that CBT is the only tradition here whose declared aim is to find out whether the thought is true. Where that aim goes missing, all that is left is the tone, and the tone sounds like cheerfulness.

That a distortion means the thought is false. The list names the shape of the inference, not the truth of the conclusion. You can catastrophize about a real catastrophe and read the mind of someone who does in fact dislike you. The label says only that the route to the conclusion skipped a step, which is why the procedure carries on to the evidence instead of stopping at the name.

That CBT is the thought record. Half of it has always been behavioural, and the dismantling literature above suggests that half may be carrying more weight than the name it shares suggests. The paper form is the part that travels well, into books and apps, which is why the public version of this tradition is the cognitive half almost alone.

Where it sits next to what came before and after

It was born out of a break with two things at once, in the 1950s and 1960s.

With psychoanalysis, from inside, and it took two attempts. Beck trained as a psychoanalyst and set out, in his own word, to validate the analytic account of depression as anger turned against the self. The dreams of depressed patients did not deliver it: he found less hostility rather than more. He did not drop the theory at that point. He rescued it, reading the dreams as masochism and publishing them that way (Beck and Hurvich, 1959; Beck and Ward, 1961).

What broke it was a different experiment. Depressed patients given tasks they succeeded at did better afterwards, not worse, which is not what a need to suffer predicts (Loeb, Feshbach, Beck and Wolf, 1964). The conclusion he drew from the wreckage was that the meanings were not hidden and did not need excavating: patients could report them if asked directly. That is the founding move, and it is methodological before it is clinical. Albert Ellis had arrived at a related position from another direction a few years earlier with rational emotive behaviour therapy.

The detour is worth keeping rather than tidying into a single moment of insight, because it is an instance of the thing this page is about. A belief met a piece of evidence that did not fit, absorbed it by being reinterpreted, and lasted five more years until it met one that left it nowhere to go.

With behaviourism, which it also came out of. The behaviour therapy of the period treated inner events as outside the scope of a science of behaviour. CBT kept the method, measurement, protocols, defined targets, and readmitted the thought as a legitimate object of it. The name holds the compromise.

Next to Stoicism, the lineage is documented, in the founders' own words, and this library covers it in the Stoicism piece rather than here: Ellis and the 1979 manual both trace the principle back to Epictetus, and the difference between asking whether a judgment is well-made and asking whether a thought is accurate is set out there.

Next to ACT, which defined itself against this tradition, the argument about whether the acceptance-based approaches are a new generation or a repackaging is in the ACT piece, along with what the comparisons found. From this side, the disagreement is clean and states itself in one line: CBT asks whether the thought is true because it holds that believing an untested sentence is what does the damage. ACT sets the truth question aside as the wrong one.

Next to the contemplative traditions, the meeting point is a specific graft. Mindfulness-based cognitive therapy, built by Zindel Segal, Mark Williams and John Teasdale in 2002, was designed for a problem CBT had not solved, relapse in people who had recovered, and it works by a different mechanism: not changing the content of the thought but changing your relation to it.

The Zen piece names the price of that arrangement. It is talking about a different programme, MBSR rather than MBCT, but the sentence transfers without adjustment: a practice that has to answer to a trial has to say in advance what will count as better, and saying it is what makes the claim capable of being shown wrong. The difference is that CBT has never been anywhere else. It was not carried into that arrangement from somewhere older. It was assembled inside it, out of a psychoanalytic theory that an experiment had just falsified, which is why the pages in this library that carry a review date are this one and ACT, and not the other four.

After: this is the institutional default. Delivered in public health services, taught from manuals, and the form most often adapted into self-help books and software.

Recommended turns out to be a weak word for it, and looking at an actual guideline shows why. England's national guidance on depression in adults (NICE NG222, 2022) offers eleven treatments for less severe depression and ten for the more severe, and says in as many words that any of them can be used first. What it does tell you is a ranking, because it orders those tables by what its committee judged most effective clinically and for the money. For moderate and severe depression, CBT combined with an antidepressant is first and CBT on its own is second. For less severe depression it is fourth, and the guidance sends you to the least intrusive option before reaching it.

One line below CBT in that same table, third of ten and ahead of medication, sits behavioural activation on its own. The guidance of 2022 lists as a treatment in its own right the component the dismantling study of 1996 said was carrying the weight. That is the closest thing to an institutional verdict on the argument in the section above, and it went the way the awkward result pointed.

Go deeper

• Beck, Rush, Shaw and Emery, Cognitive Therapy of Depression (1979): the manual that fixed the method. The first chapter is where the Stoic line is quoted, and the middle of the book is the thought record before it became a worksheet.
• Judith Beck, Cognitive Behavior Therapy: Basics and Beyond (third edition, 2021): the current teaching text, and the source of the three level statement of belief that most clinicians now learn.
• Loeb, Feshbach, Beck and Wolf, on the effect of success and failure on mood and performance in depression (1964): the experiment that ended the psychoanalytic account for Beck, worth reading next to Beck and Ward on the dreams of depressed patients (1961), which is the one that did not.
• David Burns, Feeling Good (1980): where the list of distortions entered general circulation. Read it as the source of the popular version, knowing that it is the popular version.
• Jacobson et al., component analysis of cognitive behavioral treatment for depression (Journal of Consulting and Clinical Psychology, 1996): the dismantling study, and the single most uncomfortable paper in this tradition's own literature.
• Longmore and Worrell, Do we need to challenge thoughts in cognitive behavior therapy? (Clinical Psychology Review, 2007), with Hofmann's reply on common misconceptions about cognitive therapy (2008) and their answer in the same issue: each one on its own closes a question that is open.
• David, Cristea and Hofmann, Why cognitive behavioral therapy is the current gold standard of psychotherapy (Frontiers in Psychiatry, 2018), read with the commentary against it by Fernández-Álvarez and Fernández-Álvarez (Frontiers in Psychiatry, 2019): the source of the claim about how much research exists, and the argument about what having it entitles anyone to conclude.
• Johnsen and Friborg, on the falling effect of cognitive behavioural therapy as a treatment for depression (Psychological Bulletin, 2015), read with the published replies: Cristea and colleagues (2017), Ljótsson and colleagues (2017), and Waltman, Creed and Beck (2016). Together they are the argument, not the answer to it.
• Cuijpers and colleagues on publication bias (British Journal of Psychiatry, 2010, where the pooled effect falls from 0.67 to 0.42 once the unpublished trials are accounted for) and on study quality (Psychological Medicine, 2010, where 0.74 across all trials becomes 0.22 among those at low risk of bias): the reason to distrust any single effect size, including the flattering ones.
• Dimidjian and colleagues, on behavioural activation against cognitive therapy and medication in major depression (2006), and Richards and colleagues, the COBRA trial (Lancet, 2016): the two large tests of the behavioural half on its own, and the reason the 1996 dismantling result did not stay an oddity.
• Hofmann et al., review of meta-analyses of cognitive behavioural therapy (Cognitive Therapy and Research, 2012): the map of where the support is strongest, cited in the Stoicism piece for the same reason.
• NICE guideline NG222, Depression in adults: treatment and management (2022): read the tables rather than the summaries. They are ordered, the guideline says what by, and the order carries more information than the recommendation does.
• Segal, Williams and Teasdale, Mindfulness-Based Cognitive Therapy for Depression (2002; second edition 2013): the graft, and the clearest account of why they went looking for a different mechanism.

---

What this changes in how you'd use Kitehold

One disclosure first, and it is larger here than anywhere else in this library. The Thought Record in Examine is this tradition's instrument, field for field: it runs Beck's worksheet in Beck's order, and the nine patterns it offers you are that same list of cognitive errors, arrived at by the ordinary route described further up. Nine and not Burns's ten, because mind reading and fortune telling have been promoted out of jumping to conclusions and catastrophizing out of magnification, while mental filter, labeling and minimization are missing.

So this is a page about the tradition that supplied an entire exercise and most of a vocabulary to the app it is published on. Read the rest knowing that.

Three things to do differently.

The second rating is the exercise. How much did you believe it at the start and how much do you believe the original thought now at the end are not two questions, they are one measurement taken twice, and the number that means anything is the gap. Skipping the first one, which is easy to do because at that moment you are certain, leaves you with an ending and nothing to compare it to.

Read the app's own caution on the alternative literally. Under is there another way to read this? it says: this is not necessarily truer, it is another possibility. That is a deliberate step back from the classical version, whose target is accuracy. Taken as written, the alternative is not the answer you are supposed to reach. It is a second reading placed next to the first so that the first stops being the only one available.

Take the line under the patterns at face value. It reads: recognising a pattern is not the same as being wrong. When you tick catastrophizing you have named the shape of the inference and nothing else. Whether you were right is what the two evidence sections after it are for.

And a boundary with the rest of the library, because it is the sharpest one here. This is the only exercise in Kitehold that can honestly end on a higher number than it started with. You test the thought, the evidence holds, and you believe it more at the end than at the beginning. That is not the exercise failing. It is the difference between testing a thought and talking yourself out of one, and it is this tradition's entire answer to the question this library keeps asking: a thought is owed as much authority as it can earn under examination, and not a point more.

The other pages here answer that question differently. And the pull is not the belief: they are separate measurements in separate units, for reasons the comparison piece sets out.