Warmth first, then one small step

Why warmth comes before the plan

By this point you know what you feel and what you need, and the fix is right there. The practice stops for warmth first, because that is where the felt payoff of the session comes from — and because a plan made in a spirit of self-correction tends to be a punishment with a deadline on it.

Kirschner and colleagues (2019) (opens in a new tab) compared two brief self-compassion exercises against three controls, including a positive-excitement condition, in 135 people. Self-compassion and excitement both raised warm, affiliative feeling and lowered self-criticism; so far, tied. What only the self-compassion conditions produced was the physiological pattern: lower heart rate, lower skin conductance, higher heart-rate variability, the signature associated with regulating well under difficulty. Rumination produced the inverse, in both the self-reports and the body.

Cheering yourself up and being kind to yourself are not the same move. Only one of them settles the body. Which is why this step is warm rather than upbeat — no fanfare, no you’ve got this.

What self-kindness sounds like when it works

In Kristin Neff’s framing (opens in a new tab), self-compassion has three moving parts: noticing that this is hard, remembering that being human is a shared condition rather than a personal defect, and speaking to yourself with some warmth. In the practice that arrives as an opening line and three more, slowly. This is the version for a feeling with heat in it; there are others for anxiety, for tenderness, and for good moments:

This one has heat in it. That heat came from somewhere that matters to you. Anyone who cared about this would have felt some version of it. You don’t have to be calm to deserve some kindness right now. May you be as fair to yourself as you’d be to anyone else here.

Note what is absent. No reassurance that it will be fine, no reframe, no silver lining. Warmth is not optimism, and it is not agreement that you were right. It is the tone you would use with someone you like who is having a bad day.

If self-directed warmth makes you flinch

For some people the warm prompt produces resistance rather than relief. This is documented: fear of compassion is a measured construct, with scales built for it by Gilbert and colleagues (2011) (opens in a new tab), and the people most likely to bounce off self-directed warmth are the highly self-critical — precisely the people the step was written for. If that is you, nothing has gone wrong, and you do not have to force it.

There is a second door, and it is not a lesser one: if a friend told you exactly this — same situation, same feeling — what would you say to them?

That question is usually easy to answer, and the answer is usually more generous than anything you would aim at yourself. Write it down as you would say it to them, in their words, out loud if you can. Then read it back and ask whether you would be willing to say it to yourself. Most of the resistance turns out to be to the aiming, not to the content — and if the answer is still no today, the honest move is to let the friend version stand and go on to the step. It does the same work.

Neff also describes a fiercer form of self-compassion, one oriented toward protecting yourself and getting your needs met rather than only soothing. That is the hinge into the second half of this step. Kindness here has some spine in it: honoring a need for rest means going to bed, not scheduling rest for Saturday; honoring autonomy means taking back a choice, not talking yourself out of wanting one.

One small step, in if-then form

Then the session asks for a single action that honors the need — offered in three directions, because the same need can be met in more than one way: something for yourself, something with another person, something in your body. Three options, one choice. A list of five good ideas is a list nobody does.

Except that some needs cannot be met. The person is gone; the job ended; the thing is simply done. Asking what small step would honor your need for connection, when the connection you mean is with someone who died, is the wrong question — and being asked a wrong question at your worst moment is its own small injury. So there is a fourth door on that screen, this one can't be met, and it changes the question rather than the person: not what would honor the need, but what would be kind, given that this part is not going to change. It is not a lesser version of the step. For some turns it is the only honest one.

That door is a design judgment, not a finding. The nearest evidence — that matching your strategy to whether a situation is changeable at all predicts how well it goes — is one study of 170 people, and it is not strong enough to build a practice on. What it is strong enough to do is stop the practice asking something impossible.

Insight is the satisfying part. It is not the part that shows up in the data. A 2025 preprint from Schleider and colleagues (opens in a new tab) dismantled a single-session intervention into psychoeducation, testimonials, and action planning, and tested all combinations in 889 people: only action planning significantly reduced depressive symptoms at two and eight weeks, and the effects were small (d = −0.18 and −0.12). That was a behavioral-activation intervention rather than this one, and it is a preprint, so hold it loosely. It is a reason to end on a plan — not a claim that ending on a plan treats depression.

The form of the plan is better established. Gollwitzer and Sheeran (2006) (opens in a new tab) meta-analyzed 94 tests across more than 8,000 people and found that implementation intentions — plans in if situation Y, then I will do Z form — produced a medium-to-large effect on goal attainment, around d = 0.65. A later meta-analysis of 642 tests (Sheeran, Listrom and Gollwitzer, 2024 (opens in a new tab)) put the effects in the .27 to .66 range, and found them larger when the plan used the contingent if-then form, when motivation was high, and when the plan had been rehearsed at least once.

Which is why the practice ends by pairing your step with a cue and reading it back to you, once:

If I want to fire off a reply, then I will list three things I handled well this week.

That plan will look like the wrong shape to anyone waiting for the confrontation. It is aimed at the need — evidence that you count, gathered from yourself first — rather than at the person who caused the feeling. The conversation may still be worth having tomorrow, and it will go better from a place where you are not proving anything.

Pick a cue you will actually notice: a body signal or a behavior, not a time of day. My jaw clenches. I catch myself checking my phone again. I start replaying it. The cue is what makes the plan fire without you having to remember it.

Then one skippable question — how’s the weather now? — and after that, nothing. No signup, no streak, no what’s next, and no scoring of how you did. The question sits before the last screen deliberately, so the warm moment is still the one you leave on: how a thing ends is most of what you remember about it, and a practice that ended on the problem would be remembered as the problem.

That last point has one good trial behind it rather than a literature. In a randomized study of 682 colonoscopy patients, Redelmeier, Katz and Kahneman (2003) (opens in a new tab) found that a gentler final minute made the procedure be remembered as less painful, and that those patients came back for repeat screening somewhat more often (OR 1.41). One trial, one marginal secondary result, and a procedure nothing like this — but it points somewhere, and the cost of ending warm is zero.

Plans that are really arguments

The step is a demand of someone else. Then I will explain to him why he was wrong is not a step, it is the argument with a plan attached. Aim at something that is yours to do.

The step is too big. If it will not fit in today, it will not happen today. Shrink it until it is nearly too small, then take that.

The step serves the story instead of the need. Check it against the need you named. A plan that vindicates you but leaves the need unmet will not settle anything.

The warmth gets skipped. It is the most skippable-looking half of the step and the one with the best in-session evidence behind it. Give it the extra thirty seconds.

Warmth first, so the plan is something you are doing for yourself rather than something you are doing about yourself.

That is the whole turn. It takes a few minutes, and it does not fix your life; it gets one feeling named, allowed, understood, and answered with something small. Then you close it and go do the thing. The next feeling will want its own turn, and the wheel will be where you left it.