Stop Trusting the First Version of a Feeling or a Guess

What each one is actually for

90-Second Rule

The 90-second rule asks you to do something specific the moment you notice a strong emotional surge — not to suppress it, not to figure out why you're feeling it, but to physically locate it: where in your body is this happening? Heat in the face, a tight chest, a clenched jaw, a racing pulse. You stay with that physical sensation, as plainly as you can, without adding a sentence to it — no “because he always does this,” no “this is so unfair.” The moment you add that sentence, you've handed your body a reason to produce another hit of the same stress chemistry, which is what actually stretches a ninety-second wave into a bad afternoon. Done without the added story, the physical intensity genuinely peaks and starts clearing within about a minute and a half, which you can feel directly if you've ever actually tried it and paid attention.

Superforecasting

Calibration training works by making you put a number on things you'd normally just have a feeling about. Instead of thinking “this project will probably go well,” you write down an actual percentage — say, seventy percent — for a specific, checkable version of the claim, like “we'll ship by the deadline we set.” You do this repeatedly, across many different predictions, big and small, and you keep a record. Later, once the outcomes are known, you go back and check: of everything you called seventy percent likely, did roughly seven out of ten of them actually happen? If your seventy-percent predictions come true nine times out of ten, you're systematically underconfident and should be saying higher numbers; if they come true four times out of ten, you're overconfident and saying numbers too high. The training isn't really about any single prediction — it's about tracking dozens of them over time until you can trust that when you say seventy percent, it actually means roughly seven-in-ten.

What you get out of pairing them

The 90-second rule's payoff is fast and physical: you get a real, working mind back in under two minutes instead of losing an hour, an afternoon, or a whole interaction with someone to a wave of feeling that has, biologically, already run its course but is being kept alive by the story you keep telling about it. Calibration training's payoff is slower and shows up over months, not minutes: your actual judgment about uncertain things — will this deal close, will this treatment work, will this person follow through — gets measurably more accurate, because you're no longer relying on a felt sense of confidence that was never checked against anything. Both give you the same underlying thing, on different timescales: a more accurate read of what's actually true, instead of the first automatic answer your body or your gut handed you.

What 90-Second Rule rests on

• Interoceptive attention randomised trials
  Directing sustained attention to internal bodily sensations increases awareness of the body's state and changes how that state is experienced.
• Acceptance reduces struggle randomised trials
  Allowing an internal experience to be present without resisting or acting on it reduces the psychological struggle that resistance would otherwise intensify.
• Physiological arousal self resolves if unfed plausible mechanism
  An emotional chemical surge has a limited biological duration and will subside on its own if not renewed by continued thought.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Set next to each other, the shape is the same even though the material is completely different. Both practices identify a fast, automatic, deeply convincing internal output — an emotional surge that feels like an accurate verdict on the situation, or a gut sense of confidence that feels like an accurate read on the odds — and both insert a deliberate pause between that output and your acceptance of it. The 90-second rule's pause is physical: notice the raw sensation, don't add the story. Calibration training's pause is numerical: notice the raw hunch, don't accept it until it's been forced into a checkable number. Neither practice trusts the first, unexamined version of what your mind hands you.

The principle underneath both

A fast, automatic internal signal — whether it's an emotional surge or a confident hunch about how something will turn out — arrives with more conviction than accuracy, and the conviction itself is not evidence that the signal is correct. Left unexamined, both kinds of signal get accepted as fact simply because they feel certain from the inside: an emotion feels like a true verdict on the situation, and a strong hunch feels like real knowledge about the future. The only way to find out whether either one deserves that confidence is to deliberately interrupt it — hold the raw sensation without the story, or hold the raw hunch long enough to force it into a number — before treating it as settled.

What that makes the real goal

If the risk is mistaking a convincing internal signal for an accurate one, the real target both practices are aimed at is closing the gap between how certain something feels and how accurate it actually is — in your reading of your own emotional state in the moment, and in your reading of an uncertain future event. Neither practice claims to eliminate emotion or eliminate uncertainty; both are aimed specifically at not letting the felt intensity of a signal stand in for its actual accuracy.

Other practices that serve the same goal

Cognitive behavioral therapy's thought record — writing down a triggering situation, the automatic thought it produced, and the evidence for and against that thought — chases the emotional-accuracy half of this objective through a more structured, written process than the 90-second rule's purely physical pause, and it's aimed specifically at thoughts that repeat rather than single acute surges. Prediction markets, where many people place real bets on an outcome and the aggregate price reflects a collective probability estimate, pursue the calibration half of the objective from a different angle: instead of training one person's individual accuracy over time, they average many individually biased guesses into something that's often more accurate than any single participant's gut call. And Daniel Kahneman's distinction between fast, automatic thinking and slower, deliberate thinking, from his research on judgment and decision-making, describes the exact same interruption this page is describing at the level of the whole mind — the recurring finding that the fast system produces convincing answers that are often wrong, and that deliberately engaging the slower system is what catches the difference.

A tradition that already does this

The Stoics were already making this exact move, in both directions, two thousand years before either modern version existed. Their central practical teaching was to withhold agreement from your very first impression of a situation — whether that impression arrived as an emotional reaction (“this is a disaster”) or as a snap judgment about what would happen (“this will surely go badly”) — and to examine it before treating it as true. Epictetus's instruction to test an impression before accepting it applies without modification to both a surge of anger and a confident guess about the future; the 90-second rule and calibration training are, in a real sense, two separate modern, more mechanical versions of a single older instruction: don't believe the first thing your mind hands you just because it arrived with conviction.

Using the 90-Second Pause Before You Commit to a Number

Calibration training asks for an honest probability, but a strong emotional reaction to the question itself — dread about a bad outcome, excitement about a good one — will quietly distort the number before you've even consciously chosen it; fear tends to pull an estimate down, hope tends to pull it up, regardless of the actual evidence. Before writing down a probability on anything you're emotionally invested in, run the ninety-second pause first: notice where the feeling about this specific prediction is sitting in your body, let it crest and clear without narrating it, and only then commit to a number. The number you write down after the wave has passed is measurably more likely to reflect what you actually know about the situation, rather than how badly you want, or fear, a particular outcome.

Turning “This Feeling Will Last Forever” Into a Testable Number

Part of what makes an emotional surge feel unbearable in the moment is an unspoken, unexamined prediction riding along with it — the sense that this feeling, whatever it is, is going to last, that it will never actually pass. Calibration training gives you a direct tool against that specific distortion: when a wave hits, ask yourself explicitly, as a genuine forecasting question, “what's the probability this feeling is still this intense in ten minutes?” and commit to an actual number. Most people, asked this honestly, will say something like twenty or thirty percent, which is already a different relationship to the feeling than believing, unexamined, that it's permanent — and checking back in ten minutes to see whether your own forecast was right builds, over repeated practice, real trust that the ninety-second window is a fact about how you work, not just something you were told.

The Email That Ruins Your Afternoon Before You've Even Reread It

You open an email from your manager that's shorter and colder than usual, and within seconds your chest is tight and your mind is already three sentences into a story about being fired. You notice the sensation — the tightness, the heat behind your eyes — and instead of finishing the story, you just stay with the physical wave for what feels like a long minute and a half, watching it actually crest and start to loosen. Once it's cleared enough to think, instead of immediately deciding the story is true, you ask the calibration question directly: what's the real probability, given everything you actually know about this manager and this job, that this email means what your first read of it assumed? Named honestly, it's maybe fifteen percent, not the certainty your body handed you thirty seconds earlier. You reread the actual email. It's short because your manager is in back-to-back meetings, and it's about scheduling a call for Thursday, nothing else. Nothing in the email required the story your body had already written for it — the tightness was real, but it was ninety seconds of chemistry, not information about Thursday.

When this is worth reaching for

A tense email arrives right before you have to make a decision. A hard conversation lands minutes before you need to judge someone's intentions. A stretch of anxiety hits right before you have to estimate how a project is actually going. This isn't about daily maintenance — it's about the specific junctions where an unexamined emotional read is about to get baked into an unexamined guess about the future.

Where this does not apply

If the emotion in question is grief, or a proportionate response to a genuinely serious loss or threat, treating it as a ninety-second chemical event to be waited out and then converted into a probability estimate badly misreads what's actually happening — some feelings are appropriately long, and turning them into a forecasting exercise can feel, and be, dismissive of something real. And if a decision genuinely has no meaningful uncertainty — the outcome is already known, or entirely outside anyone's control regardless of the odds — assigning it a calibrated probability adds a layer of false precision to something a number doesn't actually help with.

What the evidence does and does not support

The 90-second rule's physiological claim — that an acute stress-hormone response has a limited biological duration if not actively renewed — is grounded in real neuroscience of the stress response, though the specific “ninety seconds” figure is closer to a memorable approximation from Jill Bolte Taylor's own account than a precisely measured universal constant; actual duration varies by person and by the intensity of the trigger. Calibration training's core claim has much stronger, more direct support: Philip Tetlock's Good Judgment Project, a large multi-year forecasting tournament, found that people who tracked and refined probability estimates over time became measurably better forecasters than both untrained participants and, in some categories, professional analysts with access to classified information. What hasn't been studied is the specific combination this page describes — using an emotional pause to improve the accuracy of a subsequent probability estimate — which is a reasonable inference from what's known about emotion distorting judgment, not a finding from the calibration research itself.