Three Things I'm Grateful ForThe best-known exercise in positive psychology — and why the part almost nobody does is the part that works

Perhaps there is one on your bedside table. A small notebook in which you write three things you are grateful for each evening. Perhaps you kept it up for a few weeks. Perhaps it has been lying there blank for months, mainly producing a bad conscience.

The gratitude journal is the best-known exercise in positive psychology. And unlike affirmations, where it turned out that the exercise studied is a completely different one from the exercise people do, here the two match: what people do really is what was researched.

Which makes what comes out of it all the more interesting.

First the good news: it works

The largest review to date pools 163 studies with 24,804 people across 28 countries. It finds a small but reliable effect on well-being, one that survives correction for publication bias.

The authors themselves compare the magnitude to an antihistamine against a runny nose. It helps. It does not clear things up.

Something else speaks for gratitude too. One review took the five happiness strategies most frequently recommended in the media and then looked for adequately powered and preregistered experiments on each. Gratitude is among the best supported. For exercise, not a single study met both criteria.

In the largest single study — nearly eleven thousand people in 34 countries — the sample collected in Germany came fourth, well above average. In thirteen of the thirty-four countries the effect was not statistically distinguishable from zero; in Germany it clearly was. (In fairness: the survey was not conducted in German, and what was measured was the immediate effect of a one-off exercise.)

And now the comparison everything hangs on

To test whether an exercise works, you need a comparison group. And it makes an enormous difference which one you choose.

Compare a gratitude journal with doing nothing and it looks good. Compare it with an equivalent positive writing exercise — writing down in the evening what was nice, without the word gratitude appearing — and it looks like this:

• Meta-analysis 2016: no difference (d = −0.03)
• Meta-analysis 2017: no difference (d = 0.03)
• Meta-analysis 2025, the largest of them all: no difference (g = −0.15)

Three independent reviews, the same result. A systematic review from 2023 puts it soberly: the literature does not support the hypothesis that gratitude interventions perform better than positively valenced or psychologically active controls.

A replication of the famous Seligman study shows this most clearly. The researchers added a condition the original lacked: one group wrote down positive childhood memories each day. The result: that group improved to a similar degree as the gratitude group.

What appears to work is not the gratitude. It is sitting down in the evening and writing about something good.

The study that started it all

In 2003 the paper appeared that everything still rests on: “Counting blessings versus burdens.” Three studies, and it is worth looking into the first one.

192 students, noting up to five things once a week for ten weeks. There were effects — on how they judged their life as a whole, on their optimism for the coming week, even on hours of exercise. But the finding that is almost never quoted alongside: there was no measurable effect on mood. Neither on positive nor on negative feelings.

The second point matters more. What was it compared with? With a group asked to write down, for ten weeks, what annoyed them. That is not a neutral control but a presumably mood-lowering condition. The authors themselves write that the two conditions had “nearly equal and opposite” effects relative to neutral.

So the famous difference measures, for half of it, how people feel who spend ten weeks noting what bothers them. The large 2026 megastudy explicitly excluded this comparison condition because it produces distorted estimates.

A rule that never existed — and one that was tested

The most quoted practical advice in the gratitude literature is: once a week, no more often — otherwise it wears off.

That rule goes back to a review article from 2005. It describes a study in which students wrote down what they were grateful for either once or three times a week, and only the weekly group benefited. What is reported: no sample size, no measurement, not a single statistic. The result exists publicly as a curve in a figure. The reference in the bibliography reads, verbatim, “unpublished raw data.”

In more than twenty years it has not become a study. Two solid meta-analyses find no relationship at all between frequency and effect. And the original 2003 study found the daily version more potent than the weekly one.

With the second famous number the case is different — and nicer. Why three things, actually? Because there were three in the 2005 study. In the study before it there were five.

Here, though, someone did check. A preregistered study with open data had people note between one and ten things a day for two weeks, ten groups, the number held constant in each. The result: the number makes no difference. Not immediately afterwards, not after a week, not after six.

And one side finding from the pilot survey, which I consider the most useful sentence in this article: 52 per cent said it only came easily to them with two things or fewer. So if you sit in front of the third line in the evening and find nothing more, you have done nothing wrong. You are the majority.

What actually did work in the studies

So far this sounds like clearing away. It is not — something remains, and it is remarkably concrete.

1. 01Say it rather than list itIn the best study on this — 958 adults, active comparison group — six forms were compared. Long texts addressed to a person worked. Gratitude *lists*, by contrast, were indistinguishable from the control group. In the worldwide megastudy the message actually sent was the strongest of six forms — mere listing the weakest.
2. 02Praise them, not your own benefit370 video-recorded conversations, coded by independent raters. They distinguished two ways of thanking: praising the other person (“You really go out of your way”) or naming your own gain (“That helped me so much”). Only the first form landed: it predicted whether the other person felt seen, whether they felt positive emotions, whether they felt more affection. The second showed not one significant association. The difference is a single word: *I* or *you*.
3. 03You are wrong about how it landsPeople wrote a real thank-you letter and predicted how the recipient would react. Then the recipients were asked. The result: writers substantially underestimated how pleased the other person would be — and underestimated even more strongly how well written their letter came across. They overestimated the awkwardness. In one experiment the agreement between prediction and reality was essentially nil: they were not merely off, they had no idea.
4. 04The larger part of the effect isn't yours at allIn a diary study with 67 couples over two weeks: people who felt grateful felt somewhat better themselves the next day. But the effect on their partner was two to six times as large. Among the women the effect on themselves was not even statistically meaningful — the effect on the partner very much was.

Taken together that is a fairly clear instruction: the thing works better outward than inward. Anyone practising gratitude as self-optimisation is working on the smaller half.

What else happens

Two findings belong here that appear in no self-help book.

First: gratitude exercises measurably increase the sense of owing someone. The megastudy across 34 countries detected it; in the letter study it was even the strongest single effect of the whole investigation — stronger than any positive one. That fits one of the field's most stable findings: gratitude and indebtedness are not the same thing. Whether you feel grateful depends on whether you recognise the other person's goodwill. The sense of being in their debt arises independently of that — simply from having received something.

Second: it can backfire. In a month-long experiment, couples were asked to thank each other regularly for concrete things, compared with an active control group. Across 28 evenings it emerged that when the person expressing thanks was not experienced as responsive by their partner, the gratitude group had more negative emotions than the control group.

That, incidentally, is the best answer to the charge of forced positivity — and it comes not from critics but from gratitude research itself: thanks going through the motions is worse than none.

What the research does not say

• It does not say gratitude makes you healthy. After reviewing 19 controlled studies, the only repeatedly found result is better-rated sleep quality — in five of eight studies, and self-reported. For blood pressure or blood sugar there is a single study each.
• It does not say the effect lasts. Across the measurement points it shrinks steadily. The exercise with the largest immediate effect — reading a thank-you letter to someone — was back at baseline after three months.
• It does not say it works for everyone. The largest meta-analysis gives a range in which a new individual application is likely to land — and that range extends into the negative.
• It says almost nothing about German-speaking countries. The studies with German samples compared against waiting lists, that is, against doing nothing — precisely the comparison that inflates the effect. Austria and Switzerland do not appear in the worldwide study at all.
• It says nothing about people in real crisis. Anyone in grief or in a genuinely distressing situation will find no answer here — and should not look for one in a blog article.

And does it last?

This is where it gets uncertain, and I would rather say how uncertain.

In Seligman's 2005 study the long-term effect hung on a single factor: whether participants carried on of their own accord beyond the prescribed week. That sounds like a neat link to what we know about habits — it is not the technique that lasts, it is the habit.

Except the newer finding does not fit. In the preregistered study with the one-to-ten things, 17 per cent were still going voluntarily after six weeks — and no improvement could be detected in them.

In both cases the group that carries on is a group that selected itself. Neither finding establishes a cause. So the honest position is: we do not know whether sticking with it helps. What we do know is that the effect disappears without repetition.

And one more thing was shown that points beyond gratitude: whether such an exercise did anything at all depended on whether people had chosen it themselves and whether they put in effort. In the control condition neither made a difference. Prescribed gratitude is therefore not a sensible idea — not because it is morally questionable, but because it probably simply does nothing.

So if you take one thing from this, let it not be the number three or the question of morning versus evening. Let it be the person who came to mind while you were reading — and whom you never told.

The list in your notebook is the part that achieved least in the studies. The sentence you don't dare say is the one this is about.

This article reflects the state of psychological research and is intended as everyday orientation. The coaching methods described support general wellbeing and holistic balance. They do not constitute a medical diagnosis, promise no cure, and do not replace medical or therapeutic treatment.