CBT worksheets: how to fill in a thought record so it actually works
The worksheet filled in for show
A client brings in a completed CBT worksheet. Under "situation": "argument with my wife". Under "thought": "everything is bad". Under "emotion": "bad, 100%". Three lines, written — by the look of them — in the lift on the way to the session. Formally, the homework is done. There is nothing here to work with: no moment, no thought, no measure.
The first explanation that comes to mind is that the client is not motivated. Even where that is true, it is of little use: what a therapist can act on is not motivation but how the worksheet was introduced and how it gets reviewed. A CBT worksheet is not a report on the work; it is the work itself, carried out beyond the consulting room.
In Stone et al. (2003), the paper diaries of patients with chronic pain were fitted with a sensor that logged every opening of the binding. Patients handed in entries for 90% of the assigned moments. The binding was actually opened 11% of the time, and on 32% of days it was never opened at all — yet those days were "reported" too. Reminder signals lifted the figure only to 29% (Broderick et al., 2003); an electronic diary, which timestamps each entry, reached 94%.
What we know about homework
A caveat is needed here, without which the figure above becomes a sleight of hand. Stone studied pain diaries, not CBT worksheets. No study has ever measured when a thought record is actually filled in. What transfers is not the result but the mechanism: paper carries no timestamp, and so it cannot tell the researcher — or the therapist — when it was written.
The evidence on homework itself is firmer. When the same therapy is compared with and without homework, the difference is d = 0.48 (Kazantzis, Whittington & Dattilio, 2010; 46 studies, N = 1072). This is not a correlation of the "diligent clients do better" kind, but a comparison of protocols — an argument about the contribution of the procedure itself.
A correction is due straight away, or this turns into a sermon. Dismantling studies offer little evidence that cognitive techniques specifically add anything on top of behavioural ones (Longmore & Worrell, 2007), and the argument is far from settled. So the precise claim is more modest — and more useful — than "worksheets heal": a worksheet makes the intervention reproducible and testable. Without a record, therapist and client discuss the memory of an episode rather than the episode.
The thought record, column by column
The same sheet travels under several names. Thought record is the everyday term; in Beck's manuals it is the Dysfunctional Thought Record, or DTR; in Greenberger and Padesky it runs to seven columns. The column count differs between traditions, the logic does not — and the naming muddle costs a therapist nothing, right up until the moment they go looking for a blank copy online.
The columns, and what actually belongs in them
- Situation — a moment, not a story. Not "argument with my wife", but "9:40 p.m., kitchen, she says: 'You're not listening to me again'".
- Automatic thought — verbatim, first person, present tense: "I ruin everything." Not a paraphrase of the thought, but the thought.
- Emotion — one word and a number from 0 to 100. "Shame, 80", not "it was hard".
- Bodily reaction — where and what: heat in the face, clenched jaw. This is the anchor by which the client will recognise the state next time.
- Behaviour — what was done: left the kitchen, went silent, started making excuses. This is where avoidance becomes visible.
Four mistakes that stop a worksheet from working
- The situation is written as a story. "We argued all evening" is a generalisation with no moment and no thought in it. There is nothing to take hold of.
- An emotion has landed in the thought column. "I felt anxious" is not a thought but its consequence. The thought sounds different: "I won't cope", "she will leave".
- Intensity is not measured. With no number before, there is no number after. The shift in belief stops being visible — to the client and to the therapist alike.
- The sheet is filled in from memory, in one batch, before the session. A thought reconstructed a week later is an edited version, rewritten by how the story ended.
These are not quibbles about neatness. Each of the four mistakes strips out exactly what the record exists for: a point in time, an object to work on, and a measure. What remains is not the episode but an account of it — and that account has already passed through the very cognitive processing the work is meant to examine. To be clear: no study of fill-quality against outcome exists. This is an argument from the mechanics of the work, not a finding.
The behavioural experiment
A behavioural experiment is built differently from cognitive restructuring: the belief is not argued with, it is tested. Prediction — test — outcome — conclusion. A client convinced that "if I say no, people will stop respecting me" does not debate the thought across the desk. They say no, and watch what actually happens.
How much stronger this is than verbal disputation is a question CBT cannot yet answer with confidence, and pretending otherwise does no one any favours. There is no meta-analysis comparing experiments with thought records. There is one RCT in a non-clinical sample (McManus et al., 2012; N = 91): belief shifted earlier and generalised more widely, though the authors themselves call the advantage small. Bennett-Levy's well-known work (2003) studied trainees practising on themselves, not patients. The direction is there; the evidence is weaker than the training-course rhetoric suggests.
What reliably ruins an experiment is an untestable prediction. "It will go badly" cannot be checked: any outcome either confirms it or says nothing at all. A testable prediction is concrete: "I won't be able to finish my sentence", "my anxiety will hit 90 and stay there for ten minutes", "he will turn away". Such a prediction has an outcome — and the outcome can be written down beside it.
Exposure hierarchy and SUDS
The exposure hierarchy was traditionally built like this: list the avoided situations, rate each on the Subjective Units of Distress Scale (SUDS) from 0 to 100, work upwards, and stay on each step until anxiety drops by roughly half. The logic was habituation: anxiety must subside here and now, and that was taken to be the condition of the effect.
Neither the degree by which fear reduces nor the ending fear level predict therapeutic outcome.— Craske et al., Behaviour Research and Therapy, 2008
The inhibitory learning model turns this picture around. Exposure delivered in a random order, with no regard for the level of fear, provokes higher anxiety that does not habituate within the session — and produces a better long-term outcome (Craske et al., 2008, 2014, 2022). It is still usual to begin with the least distressing item — not for the sake of habituation, but to reduce the risk of dropout. The point, then, is not to wait for anxiety to fall, but to obtain an experience that contradicts the client's expectation.
The hierarchy is not abolished — its purpose changes. From a ladder climbed as SUDS come down, it becomes a menu of tests: each step checks one specific expectation. For the worksheet this implies a concrete edit: the column "what I expect, and how likely, %" matters more than the SUDS column. What needs comparing is expectation against outcome, not anxiety against anxiety.
Making worksheets part of the session
All of the above rests on one condition: the worksheet is introduced during the session, on live material, rather than handed over at the door with "fill this in at home". This is not merely common sense: among the therapist behaviours that increase engagement with homework, a written summary of the task and its rationale is singled out (Ryum, Bennion & Kazantzis, 2023), and handing over a written sheet at all is itself associated with better completion (Helbig & Fehm, 2004 — survey data, so an indication rather than proof).
- Fill in the first worksheet together, on an episode from this week. The client should see a completed row, not guess at what is wanted.
- Review homework at the start of the session, not in the last five minutes. Where it sits on the agenda tells the client what the task is worth, more accurately than any words.
- An unfinished task is material, not a failure. "What got in the way?" often opens the most substantive conversation of the session.
- Record intensity and expectations as numbers. Without a number the shift is invisible — and there is nothing to show the client.
Ready-made worksheets — the thought record, the behavioural experiment, the exposure hierarchy and a homework form — are available in the Soveria library as Word documents: download them, print them, or send them to a client. The library is open to registered specialists; a free account is required.
A worksheet works exactly to the extent that it is reviewed in session. A sheet handed over without review measures the client's compliance, not their thinking.
The difference between "the worksheet is filled in" and "the worksheet worked" does not lie with the client. It lies in how the therapist introduced the sheet, what exactly they asked to be written down, and what they opened the next session with. What the client brings back on paper is not a report — it is the only access anyone has to an episode the therapist did not witness.