Relapse is data, not a personality test

The word relapse comes from clinical work on returning to a problem behaviour after a period of change. In a habit tracker it gets used more loosely: you were doing the new thing, then you were not. Either way, the episode contains information. Time of day. Place. People. Sleep. Hunger. A streak that had become the real goal. If you skip the information and jump to "this is who I am," you will not have a next experiment. You will have a mood.

This is not medical advice. Clinical relapse in addiction, eating disorders, or other diagnosed conditions is a care issue. What follows is about ordinary behaviour change and about the consumer version of Marlatt's warning: the story after the slip often does more damage than the slip.

Reading the miss as a log, not as a mirror

Wood and Neal treat habits as learned mappings from context to response. A "relapse" into the old mapping is often just the old cue winning because it was stronger, more practised, or more available than the new one. That is not a personality. That is a stronger association. The data question is which cue won. Kitchen at 10 p.m.? Unlocking the phone in bed? The pub on Friday because weekday rules never had a Friday version?

Write four lines, not a memoir. Where were you. What time. What happened in the hour before. What rule you had written (if any) for that situation. Gollwitzer's implementation intentions exist because people leave those situations unspecified. "I will do better" is not an if-then. "If I am in the kitchen after 9, then I put the kettle on and leave the room" is an if-then. Relapse data is how you find the missing if.

Lally's 2010 results are a check on drama. Everyday automaticity was a curve that could survive a miss. A personality test cannot survive a miss by design. If your inner examiner only offers pass/fail, fire the examiner. Hire a lab notebook.

What counts as data, what counts as noise

Data: the cue, the competing reward, the size of the first slip, whether the rest of the day was abandoned (that last one is the abstinence violation effect). Noise: zodiac, "I lack discipline," a story about your childhood that you cannot test this week. Shame produces noise. Guilt, in the ordinary research split, produces a repair item.

Fogg would look at whether the new behaviour was too big to survive that context. Tiny it. Duhigg's popular loop would ask what reward the old behaviour still pays. If the old behaviour pays relief, the new one needs a relief channel that is legal, or the old one will keep winning at 10 p.m. That is analysis. It is not an excuse factory. Excuses skip the next instance. Analysis schedules it.

A week of returns is a different dataset from one evening. Patterns across a week tell you the cue is still in the environment. One evening tells you a high-risk hour exists. Do not flatten those into the same moral.

Restart without making it a saga

The next instance should be smaller than the one you failed, in the same place if the place is safe, or in a replaced place if the place is the problem. Log it. If your tools only understand perfect days, they will fight you. Partial credit, a bronze session, a rate: these keep the notebook usable. Habit AI is a concrete example of a log that can hold a reduced day. Use any notebook that does not demand a redemption arc.

Adriaanse and colleagues, among others, have tested if-then plans in eating contexts; the broader Gollwitzer line is that specifying the situation plus the action beats a vague wish. Relapse is often just an unspecified situation showing up on time.

Webb and Sheeran documented how weakly a change in intention predicts a change in action. You can intend sincerely and still lose to a cue. That finding is merciful if you let it be. It means you need better situation design, not a harsher self-concept.

Question after a return	Personality answer	Data answer
Why did this happen	"I am weak"	Which cue, which hour
What next	Wait to feel like a new person	Tiny action at the next cue
What to log	Nothing, hide	The episode plus the next stub
What to change	Everything, Monday	One if-then for that situation

You are allowed to dislike the episode. Dislike is not data. The four lines are. Keep the four lines. Throw away the verdict.

Memoir is delay

A long journal about why you failed can feel like work. Sometimes it is hiding. Five minutes of prose that ends without a next cue is rumination with better spelling. Fogg would rather you did the tiny action, then wrote one line. Wood and Neal would rather you changed the kitchen at 10 p.m. (light on, food out of sight, kettle as the competing cue) than that you understood yourself more poetically.

If you like writing, put it after the stub, as a reward, not as a gate. Duhigg's loop can hold that: cue, tiny routine, then the paragraph. The paragraph does not get to cancel the routine. Lally's repetitions do not accumulate in a memoir. They accumulate in the room.

Read next

The spiral has a name: the abstinence violation effect. For how one miss sits on a learning curve, see missing one day. Habit AI if you want a log that can hold the stub.