What to do when it happens again

Updated 2 September 2026

• One occurrence changes nothing about whether the plan works. It tells you which situation the plan did not cover.
• The reaction is the thing with leverage. Marlatt's model identifies the self-blaming reading of a lapse as an immediate determinant of what follows.
• “Not one, ever” is a good rule to hold beforehand and a bad sentence to hear afterwards, because afterwards it makes the second one follow from the first.
• A large response — a ceremony, a reset, a talk about what this says about you — manufactures the thing it is trying to prevent.
• The question worth answering is not why you did it. It is where you were, what had just happened, and what was within reach.
• If the same situation produces the same result twice, that is a plan problem and it is fixable. Once is not yet a pattern.

What the model actually says

Alan Marlatt's cognitive-behavioural model, set out in the 1980s and reviewed many times since, separates two things that ordinary language runs together. There is the event — a single occurrence after a period without one — and there is the response to it. The model names the second as an immediate determinant of what happens next, alongside the high-risk situation itself and whether the person had a way of coping with it.

The mechanism is not complicated. Someone who has held a rule for three weeks breaks it once and reads that as proof about themselves: that they cannot do this, that the three weeks were pretence, that the rule was never really theirs. Having reached that conclusion, the rest of the evening has no reason to be different from the first ten minutes of it. The conclusion does the damage, and the conclusion is optional.

Larimer, M. E., Palmer, R. S., & Marlatt, G. A. (1999). Relapse prevention: An overview of Marlatt's cognitive-behavioral model. Alcohol Research & Health, 23(2), 151–160. PMCID: PMC6760427.

Why a big response makes it worse

The instinct after an occurrence is to do something proportionate to how it felt. Start again on Monday. Set a harder rule. Write down what it means. Every one of those treats the event as significant, and significance is exactly what you do not want to give it, because the reading is the risk.

This is also the argument against the mechanic that most apps in this area are built on. A total that counts up is a total that can be sent to zero, and sending it to zero is a ceremony whether or not anybody meant it as one. It says, in the clearest way an interface can, that what you had is now gone. That is the reading you were trying not to reach, delivered by the thing that was supposed to help.

There is a longer version of that argument on why a day counter can make stopping harder.

The three questions that are actually useful

Not why did I do that. That question has no checkable answer and every answer it produces is about your character. These three have answers you can act on the same evening.

Where were you? A room, a car, a walk home, a chair. The situation is the unit that a plan can attach to, and it is almost always more specific than you would guess: not “evenings” but the twenty minutes after sitting down.

What had just finished? Most of these happen at a seam — the end of work, the end of a meal, the end of what you were watching, the end of a conversation. The seam is where the plan has to be, because that is the last point at which you are still deciding anything.

What was within reach? This is the least interesting question and the most productive one. If it was in the house, in the car, on the home screen or in the drawer, then the plan has a physical version that costs nothing to try.

What changes, and what does not

Change the plan for that one situation. Leave every other situation alone — tightening the rule everywhere after one occurrence is the large response wearing a sensible coat, and it produces a plan that describes an ideal week rather than yours.

If the same situation has now produced the same result twice, that is different. Twice is a plan that does not fit, and it wants a real change: a different place to keep the thing, a different thing to do there, or that situation left out of scope until the others are quiet. Leaving one out is not giving up on it. It is refusing to spend the next month failing at it daily.

What to say to yourself, if anything

As little as possible, and nothing about you. “That situation had no plan” is true, is actionable, and is not a verdict. It also has the advantage of being what you will still think in a week, which is more than can be said for most of what gets said in the hour after.