What cognitive defusion is
A thought you are fused with is not experienced as a thought. It is experienced as the situation. “I'm going to embarrass myself” arrives with the authority of a weather report, and you start planning around it — rehearsing, avoiding, scanning the room. You are looking through the sentence at a world it has already coloured in.
Defusion is the shift to looking at the sentence. Same words, no rebuttal, no attempt to delete them — just the added recognition that this is language happening in a mind. In acceptance and commitment therapy the standard framing is that you stop trying to change what the thought says and change how tightly you are holding it.
That is the whole distinction, and it is worth being precise about, because the other obvious move — take the thought seriously as a claim, list the evidence for and against it, write a fairer version — is a different technique. That one is cognitive restructuring, it comes from classical CBT, and it lives in the CBT thought record on its own page. Both are useful. They are not the same operation, and doing one while believing you are doing the other tends to end in an argument with your own head.
Where it comes from
Defusion is one of the six processes in acceptance and commitment therapy (ACT), developed by Steven C. Hayes and colleagues from the 1980s onward and grounded in relational frame theory, an account of how humans learn to relate words to each other and then respond to the words as if they were the things. ACT groups defusion with acceptance, contact with the present moment, self-as-context, values, and committed action, and calls the whole set psychological flexibility.
The exercises themselves are older and stranger than the framework. Repeating a word until it turns to noise was described by Edward B. Titchener in A Beginner's Psychology in 1916; the name “semantic satiation” comes from Leon Jakobovits James's 1962 doctoral dissertation. This page describes that provenance and no more: it does not tell you what the exercise will do for you.
The five steps, and what each one is doing
Say the thought you typed is I'm going to embarrass myself in the meeting. Here is the sequence the tool above runs, and the mechanism each step is reaching for.
- Add a frame. You read it back as “I'm having the thought that I'm going to embarrass myself in the meeting.” Six words of grammar are inserted between you and the claim. The claim is untouched; its status has changed from fact to reported content.
- Step back once more. “I notice I'm having the thought that…” Now there is a noticer in the sentence — the vantage point the thought is passing through. ACT calls this self-as-context, and the point of naming it is that it is never the thing being noticed.
- Name the story. You give the pattern a short title: the everyone's-watching story. Minds run reruns, and a rerun with a name is easier to spot the next time it starts than one you have to re-derive from scratch at 3am. “Next time” only means something if the name outlives the tab, so if you saved a previous run this step lists the titles you have already used, with the date you named them, and one tap drops a title back into the box — recognising the rerun is the exercise.
- Wear out a word. You pick the most charged word — embarrass — and repeat it out loud for about thirty seconds. Meaning drains and the raw sound is left over. This is the semantic satiation step, and it is the one that feels ridiculous; saying it silently in your head largely defeats it.
- Check the hook. You re-rate 0–10. The number is your own read on how tightly you are holding the thought right now, nothing more — see the limits below before you treat it as a measurement.
You can move backwards and forwards between the steps, and re-run the whole thing on the same thought or a different one. Nothing carries over between runs unless you ask it to: step five has a save button, and the only thing saving changes about a later run is that step three shows you the story titles you have used before.
Honest limits
- The 0–10 rating is self-report, not data. It is your impression in one moment, taken minutes apart, by a rater who knows what the exercise is supposed to do. Read it as a prompt to check in, not as a result.
- The number sometimes goes up. Looking straight at a thought you have been half-avoiding can turn the volume up before anything else happens. That is a known and unremarkable outcome, not a sign you did it wrong.
- One pass is one rep. Defusion is taught as a practised skill, not a switch, and a single run on a thought you have been fused with for years is exactly that: one run.
- Some thoughts do not need defusing. If the thought is accurate and about a real problem — a deadline you will miss, a conversation you are dodging — loosening your grip on it is the wrong tool. That thought wants a decision, not a reframe.
- It is not treatment. This is a self-help exercise, not therapy, not a diagnosis, and not a substitute for care you are already getting. If a thought is frightening you or will not let go, that is a reason to talk to a professional, not to run the tool again.
What happens to what you type
Nothing typed here is ever uploaded. There is no server to upload it to: the exercise is a few hundred lines of JavaScript running in your tab, and the file it runs from — tools.js — is served unminified from this domain, so you can read it. By default the thought lives in the page's memory and nowhere else, so closing the tab is the delete button.
The one exception is a button you press yourself. At step five, Save this run to this device writes a single entry to your browser's localStorage under the key na.defusions — the story title, the thought, the two 0–10 hook numbers and the date, and nothing else. It is never automatic, never silent, and it never leaves the device: the same browser on the same machine is the only place that entry exists. It is there because step three asks you to name the story so you can spot it next time, and an instruction that spans sessions needs something that survives the tab. The newest 30 runs are kept; save a 31st and the oldest drops off.
Removing what you saved is one click and a confirmation, in both directions: every entry under Saved runs on this device has its own delete, and Delete all saved runs empties the key. You can also take the runs with you — Download my saved runs writes a JSON file, and Plain text renders them for copying or printing. If your browser has site data blocked or you are in a private window, saving is simply unavailable and the page says so; all five steps still run exactly as they do otherwise.
What the page deliberately does not do with those numbers: it never charts them, averages them, totals them, or tells you a run of falling scores means you are getting better. A 0–10 self-rating taken minutes apart by someone who knows what the exercise is for cannot support that claim, so the saved list shows titles and dates and leaves the reading to you. The one network request this page makes is a self-hosted page-view counter that records the URL, never your text; /privacy.html lists every key the site writes.
If the thought is about wanting to hurt yourself, please stop here and contact your local emergency number or a crisis line. This exercise is not built for that, and no exercise on a web page is.
Related tools and reading
The tools page has the box-breathing pacer and the Pomodoro focus timer; the CBT thought record has its own page too. The field guide covers the psychology underneath all four in plain language — predictive processing, the standard list of cognitive distortions, defusion and the observing self, and the physiology of slow breathing — including a section on where the evidence is thin.