Introduction
It's said with authority, usually by someone who means well, and often to shut down a conversation. Put your own oxygen mask on first. You can't pour from an empty cup. Sweep your own doorstep.
There's something in it. There's also a substantial body of research suggesting that, applied to addiction recovery, the maxim gets the order wrong — and that waiting until you're fixed before helping anyone may be costing you the thing you're waiting for.
What the Research Shows
The idea has a name, coined in 1965 by Frank Riessman: the helper therapy principle. The observation that when a person helps someone with a shared affliction, they help themselves.
For decades this was folklore. Then it was tested.
Using data from Project MATCH — one of the largest clinical trials in alcohol research — Pagano and colleagues examined people who helped other alcoholics during treatment, specifically by sponsoring someone or working the twelfth step.
Those who helped others were roughly twice as likely to be sober twelve months later.1 Concretely: 40% of the helpers avoided drinking in the year following treatment, against 22% of those who did not help.
The effect held after accounting for how many meetings people attended, which matters — it wasn't simply that helpers went to more meetings.
The Rest of the Evidence
A later analysis found that 94% of those who helped others at any point during a fifteen-month study continued doing so, and that they experienced lower levels of depression.
Other work has found sponsorship associated with substantially higher odds of abstinence, and helping associated with higher self-efficacy — the belief that you can do this, which is itself a predictor of whether you do.
And crucially, the helping that shows benefits is not heroic. It includes making coffee, setting up chairs, calling someone to remind them about a meeting. Anyone can do it, on day four, with nothing sorted out.
Why It Might Work
Several explanations, none of them mutually exclusive.
Helping displaces self-preoccupation, which addiction reliably intensifies — Pagano frames it explicitly as diminishing the egocentrism she treats as a root cause. An hour spent attending to someone else is an hour not spent inside your own head.
It generates evidence about yourself. A person who has been useful today has something to set against the account of themselves as a burden, and that account is what shame runs on.
It also raises the cost of relapse in a way that isn't shame-based. Someone is expecting you to make the coffee.
And it produces the thing early recovery most conspicuously lacks: a role, a reason to be somewhere, and people who notice when you're absent.
The Honest Caveat
This is observational, not experimental. Nobody randomly assigned people to help others.
So it remains possible that people who were already doing better were more inclined to help, rather than helping causing them to do better. The Project MATCH analyses did control for meeting attendance and used methods designed to separate these effects, which is reassuring but not conclusive.
What can be said: the association is robust, it appears across multiple studies, and there is no evidence pointing the other way — no finding that helping others while still early in recovery harms the helper.
So Why Does the Maxim Feel True?
Because it's protecting against something real.
There is a version of helping that is genuinely a problem. Taking on someone else's recovery as a project. Making yourself responsible for whether another person stops. Attaching your own stability to their outcome, so that their relapse becomes your collapse.
That is not helping. It's a way of not looking at yourself, dressed as generosity, and recovery communities see it constantly.
The maxim is aimed at that. It has simply been over-applied, until it now prevents the modest, bounded, mutual helping that the evidence says is beneficial.
Where the Line Actually Falls
Not "have you fixed yourself." Something more useful.
Are you taking responsibility for their outcome? You cannot make someone stop. Believing you can is grandiosity, and it will hurt you when they don't.
Does the helping have boundaries? Turning up for a shift is helping. Being available at 3am indefinitely to someone actively using is not.
Are you avoiding your own work? If the sponsee's crisis is a reliable reason not to attend to your own life, notice that.
Is it costing your stability? Not discomfort — stability. These are different.
If those answers are good, the evidence is on the side of helping now rather than later.
The Version That Applies to Family
One important exception, because this article shouldn't be misread.
If the person you want to help is the partner you deceived, or the child you frightened, or the parent who is still afraid of you — that is a different situation. Making the people you hurt into the site of your service is not the helper therapy principle. It asks them to hold you while you practise.
Help strangers. Help peers. The people you damaged need repair, which is a different activity, and it runs on their timeline.
The Bottom Line
"Fix yourself first" guards against a real failure mode — making someone else's recovery your project, and thereby avoiding your own work. But the research points the other way for ordinary, bounded helping: people who sponsored others or worked with them during treatment were about twice as likely to be sober a year later, and the helping that counted included making the coffee and setting out the chairs. The evidence is observational rather than experimental, so hold it loosely. But nothing suggests that helping early harms the helper. You do not have to be fixed to be useful, and being useful appears to be part of how people get fixed.
Sources
- 2x more likely sober at 12 months — Pagano ME, Friend KB, Tonigan JS, Stout RL (2004). Helping other alcoholics in Alcoholics Anonymous and drinking outcomes: findings from Project MATCH. Journal of Studies on Alcohol, 65(6):766-773. See also review: Pagano et al (2010), Alcoholism Treatment Quarterly. View source ↗ (observational, not randomized)