Mind

How to Start Again When You Have Stopped Before

Cameron Usher · Founder, PX

15 September 2026 · 9 min read

12 years in competitive sport; trained alongside the Australian Olympic squad; coached 100+ people one-on-one.

The first time you started, you did not know what would happen. That is a real advantage and nobody appreciates it at the time.

Looking for the motivation to start working out again after you have already failed at it two or three or five times is a different problem, and it is harder in a specific way. You are not asking whether you can do it. You are asking it while holding evidence that you cannot. Every previous version of this plan is sitting there in your memory, and each one ended the same way.

This post is not going to tell you that this time is different. It is going to tell you why the comparison you are making is unfair, what actually predicts whether a restart holds, and why the thing you have lost is smaller than you think.

For the mechanics of the first few weeks, how much to do and how fast to build, getting back into training after a break has the numbers. This post is about the weight you are carrying into it.

You are comparing yourself to a filtered sample

Here is the structural unfairness, and it comes from an unlikely place.

Everybody has a rough sense of how many attempts a behaviour change is supposed to take. Somewhere around three or four, maybe five, and if you are past that, something is wrong with you specifically. That number comes from asking successful people how many attempts it took them.

Which is the problem. Ask people who have already succeeded and you have systematically removed everybody for whom it took longer, because they have not finished yet.

Researchers did it properly for smoking cessation, following 1,277 people in a longitudinal cohort every six months for up to three years rather than surveying survivors. They modelled the number of attempts four different ways. The estimates came out at 6.1, 19.6, 29.6 and 142, and the authors identified 29.6 as the method with the lowest expected bias. Their own conclusion was that understanding it may take 30 or more attempts for many people could help set clinical expectations.

Two honest caveats before you take that anywhere. The estimates span a twenty-three-fold range depending on assumptions, so 29.6 is a best estimate and not a fact. And this is smoking, which involves a physical dependency training does not. It is an analogue, not a transfer.

But what it licenses is enough: the fifth attempt is not evidence of anything. The number you have been measuring yourself against was produced by a survey design that could not see people like you, and the moment somebody looked properly the number went up by a factor of five. You are not an outlier. You are in the part of the distribution nobody bothered to count.

What actually predicts whether a restart holds

There is one systematic review that asked this question prospectively rather than retrospectively: what, measured in advance, predicts who lapses and who relapses in physical activity.

It covered 37 studies. Self-efficacy, meaning your specific belief that you can do this particular thing in these particular circumstances, was the only consistent predictor of both lapse and relapse. For most other variables, there was not enough evidence to conclude anything at all.

Two things to take from that, one useful and one sobering.

Useful: self-efficacy is not general confidence and it is not self-esteem. It is task-specific and situation-specific. Believing you are a capable person does not help. Believing you can do a thirty-minute session on a Wednesday when work has overrun does.

Sobering: the review authors are careful to say most predictors had been studied once or not at all. This is a thinner literature than its confidence suggests. We know a great deal about why people stop and remarkably little about what makes them start again, which is worth knowing when anyone tells you confidently what will work.

The uncomfortable finding about how confidence gets built

If self-efficacy is the thing, the obvious question is how you raise it. The evidence here is genuinely counter-intuitive and it is why this post does not contain a pep talk.

A meta-analysis examined 27 physical-activity interventions covering 5,501 people, all of them explicitly designed to increase self-efficacy. The overall effect was small. More interesting was which techniques were associated with which outcomes. Feedback on past performance was associated with the highest self-efficacy. Persuasion and barrier identification were associated with lower self-efficacy.

That is an observational moderator analysis, not a randomised comparison, so it is a strong signal rather than proof. But sit with the direction of it. Being told you can do it did not appear to help. Being talked through everything that might get in your way did not appear to help. Being shown what you had already done did.

Which means the most useful thing available to you right now is not a reason to believe in yourself. It is a record. Not of what you planned. Of what you actually completed, this week, including the sessions that were shorter than intended, because those count and the plan you abandoned in March does not.

This is also why the version of PX that would have been easier to build is the wrong one. A system that pushes you is a persuasion machine, and persuasion is on the wrong side of that finding. A system that shows you what you did, and sizes the next fortnight from it, is on the right side.

You are not starting from zero, and here is the honest version

The claim you have heard is muscle memory: your body remembers, so a comeback is faster than the first time. The popular version rests on cellular research that has not translated cleanly to humans. The permanence of extra cell nuclei from previous training has been confirmed, and it did not produce faster regrowth in the studies that tested it, and in older adults the advantage disappeared entirely.

So here is the version that survives scrutiny, and it is better than the myth because you can rely on it.

A break did not cost you the endpoint. In a randomised controlled trial, untrained adults were split into a group that trained for twenty weeks straight and a group that trained ten weeks, stopped for ten, then retrained for ten. At the end, the two groups were level on leg press, biceps curl, muscle cross-sectional area and jump performance. No significant differences on any measure. During the first five weeks of retraining, the returning group gained faster than the continuous group did over the equivalent stretch.

Half the training time. Same finishing position.

And skill does not decay the way size does. After ten weeks of training and twenty weeks of doing nothing, one study found muscle thickness and cross-sectional area returned to baseline while strength remained elevated by around 60%, which the authors attributed to a long-lasting motor learning effect. The classic reviews of detraining agree that force production declines slowly and typically stays above untrained levels for long periods.

What you should not expect: aerobic fitness to hold. In someone who had only recently built it, recently-acquired aerobic gains are lost after a few weeks off, along with much of the muscle capillarisation. That part goes, and it comes back the way it came the first time.

So the accurate ledger is: you lost the engine, you kept the skill, you kept a good share of the strength, and the finishing line has not moved. That is not nothing to start from. It is most of the way to an argument that the fifth attempt has better raw material than the first one did.

Plan for the collapse, not the start

Every restart is planned for the good version of the week. That is the design flaw, and there is a clean trial showing how much it costs.

Participants were randomised into three groups. One received standard care. One made action plans, specifying when, where and how they would exercise. One made action plans and coping plans, specifying what they would do when the obstacle they expected actually arrived.

Two months later: 95 minutes of weekly exercise in the control group, 113 in the action-planning group, 179 in the group that had planned for the obstacles.

And the middle number is the one that should stop you. Action planning on its own was not significantly better than doing nothing.

Writing "Tuesday, 7am, gym" is the thing everybody does and it is, on this evidence, close to inert. What worked was the sentence after it: and if Tuesday's meeting overruns, it becomes twenty-five minutes on Wednesday morning.

The caveat, kept in plain sight: this trial ran in cardiac rehabilitation patients with a mean age of 59, and the broader literature on if-then planning for exercise is more modest, with pooled effects that are small and only reliably significant when the plan is reinforced rather than made once. So use the shape of the finding and not the size of it. The shape is: the plan for the bad week is the plan.

Ask yourself which specific thing ended the last four attempts. It is usually the same thing every time, and you almost certainly know what it is. That thing, and what happens when it arrives, is the only part of the plan that has been shown to matter.

What a plan should do with a history like yours

Three things follow from all of the above.

It should size from what you did, not what you meant to do. In PX that is the capacity dial, one to ten, driven by adherence. Complete 80% or more over a fortnight and it steps up. Fall below 50% over a week and it steps down. It recalculates every Sunday. It is deliberately unflattering, and stepping down is the part that does the work, because a comeback built on the level you had two years ago is a comeback designed to fail in a fortnight.

It should already know what happens when the week collapses. Not a reminder that you missed a session. A recut session, of the size that fits what is left, with the reason attached: Tuesday did not happen and Wednesday has a forty-minute gap, so this is a forty-minute version and the long session moves to Saturday.

It should show you what you completed rather than tell you that you can do it. That is the feedback finding, built into a product rather than into a slogan.

One last thing, and it is the point of this post more than any of the numbers. You have been treating your previous attempts as evidence about your character. They are evidence about the plans. Four plans failed. Each of them was sized for a week you did not have, made no provision for the thing that always happens, and asked you to supply the difference out of willpower. That is a design problem with a known solution, and it has been misdiagnosed as a you problem for years.

If part of what ended it was guilt about the sessions you missed, that is worth reading on its own, because the standard response to it makes the next attempt less likely rather than more: guilt about skipping a workout. And if the question underneath this one is how any of it becomes automatic, how to build an exercise habit has the honest timeline, which is longer and more forgiving than you have been told. Both are part of the same argument about staying consistent with training over the long run.

Start from where you actually are. PX steps your capacity down honestly and rebuilds from what you complete, with the reason on every call. Try PX.