Does Journaling Actually Work? What the Evidence Shows
Yes — journaling has two kinds of evidence behind it: centuries of practitioners who swear by it, and modern trials showing real, modest average gains from even brief first attempts. The gap between the two is mostly dose and subject — the trials measured strangers writing three times; the practice its advocates describe is chosen, personal, and kept. Write about what matters to you and keep going, and you are on the right side of both.
Yes. The people who keep journals have been saying so for as long as there have been journals, and the research — read for what it actually measured — agrees more than its headline numbers suggest. What decides your outcome is not a lab average: it is what you write about, and whether you keep going.
Start with the people who practise it
Brother David Steindl-Rast, the Benedictine monk whose TED talk carried the idea to a mainstream audience, puts the case in one sentence: it is not happiness that makes us grateful, but gratefulness that makes us happy. He is not reporting a trial. He is reporting a monastic lifetime of the practice — and a fair answer to “does it work” keeps that kind of testimony on the table alongside the data, because the data, as we are about to see, measured something much smaller.
Robert Emmons, the researcher who founded the scientific study of gratitude, made the same bet in book form: Thanks! argues that gratitude is not a temperament you are born with but a practice you can train — and that training it makes people measurably happier. When the person who ran the foundational trials chooses to argue the practice that warmly, that is information too.
So the practitioners are not hedging. The numbers below are smaller than their testimony — and the gap between the two has an explanation worth reading before you take the numbers as the final word.
The number
The most complete answer comes from Frattaroli’s 2006 meta-analysis in Psychological Bulletin: 146 randomised studies, analysed with random-effects models, average effect r = .075.
That is small. It is also the most trustworthy figure in the field, and here is why that matters. Two earlier syntheses — Smyth in 1998 and Frisina and colleagues in 2004 — reported larger effects using fixed-effects models, which assume every study estimates one identical underlying effect and therefore limit how far you can generalise. Frattaroli’s analysis was bigger, better specified, and landed lower.
When a better analysis produces a smaller number, the smaller number is usually closer to the truth.
For gratitude journaling specifically, the equivalent figure is Choi and colleagues’ 2025 preregistered meta-analysis in PNAS: 145 papers, 24,804 participants, 28 countries, Hedges’ g = 0.19.
Two different literatures, two decades apart, both landing on “small but real”. That convergence is worth more than any single dramatic study.
What those numbers measure matters as much as their size. Both pool trials that asked randomly assigned participants — most of whom had never chosen to journal — to write for a handful of sessions, often across a single week. That is a fair test of a first attempt. It is not a measurement of somebody who decided to keep a journal and then kept it for a year, and no meta-analysis in this field has measured that person. Frattaroli’s own moderator analysis points the same way: how much and how long people wrote changed the result.
What it seems to do
Reduce mental load. Klein and Boals ran two semester-long experiments and found students who wrote about a stressful experience showed larger working memory gains seven weeks later than students who wrote about a trivial topic, alongside declines in intrusive thinking. The proposed mechanism is straightforward: unresolved things occupy attention, and writing stops you carrying them.
Help with processing a specific event. Qian and colleagues pooled eight randomised trials of writing for pregnancy-related distress. It worked for post-traumatic stress. It did not produce the expected effects on anxiety or stress, and results for depression were inconclusive. That split is the clearest illustration of what this tool is: something for working through a particular difficult thing, not a general-purpose calming technique.
Modestly support mental health as an adjunct. Sohal and colleagues screened 3,797 articles and meta-analysed 20 randomised controlled trials. Journaling arms improved more than controls — a gap of roughly 5%, a small-to-moderate Cohen’s d. Their own summary is unusually blunt: the data are highly heterogeneous, methodological flaws limit firm conclusions, and the recommendation is B-level, not A.
Their practical conclusion is the right one to carry: given the low risk and low cost, it is worth considering as something that complements evidence-based care. Not as a substitute for it.
What it does not do
It did not beat control for cancer patients. Zachariae and O’Toole pooled 16 randomised trials and found no statistically significant effect on psychological health (g = 0.04), physical health (g = 0.08), or quality of life (g = 0.09). Every confidence interval crossed zero, and the result held regardless of control condition, setting or cancer type. Zhou and colleagues, pooling an overlapping set, did find a short-term reduction in physical symptoms that faded past three months — and no psychological effect at any point.
Two careful reviews of similar literature reaching different conclusions is the honest state of this field.
It does not reliably change your body. Boggiss and colleagues reviewed 19 randomised gratitude studies for physical health outcomes. Subjective sleep quality improved in 5 of 8 studies. Everything else — blood pressure, inflammation, physical symptoms — was understudied, mixed, or rested on a single study, and most included studies carried risk-of-bias concerns.
It can make you feel worse first. Smyth’s 1998 synthesis found writing reliably increased distress immediately afterwards — and that this short-term distress was unrelated to whether people benefited. Feeling raw after writing about something painful is a documented part of the process, not a sign it is going wrong.
The failure mode worth knowing before you start
Writing about yourself is not automatically good for you.
Takano and Tanno separated self-reflection — curious, problem-solving — from self-rumination, the anxious loop that revisits the same hurt without resolving it. Reflection predicted lower depression; rumination predicted higher. But because people who reflect also tend to ruminate, the total effect of self-reflection came out at almost zero.
An open page hosts either mode. That is the strongest practical case for prompts over blank pages: a prompt points attention somewhere, and an empty page lets it circle.
If journaling consistently leaves you worse, that is a recognised pattern, not a personal failing. If you are struggling with your mental health, please see our safety page for crisis resources — a journal is not treatment and this page is not medical advice.
Why a small effect can still be worth it
Read r = .075 and the reasonable reaction is “why bother”. Three reasons:
- The cost is minutes and the risks are minimal. Small effects are worth having when they are nearly free. Sohal and colleagues make exactly this argument.
- The average hides the range. Frattaroli identified moderators — conditions under which writing works better or worse. Some people benefit considerably more than the average.
- Dose is under your control. Komase and colleagues found interventions asking for four entries or fewer produced no significant change on any outcome. A practice that is too small to work is the most common version of “journaling didn’t work for me”.
What actually decides it
Not the notebook, not the prompt wording. Whether you keep doing it.
That is the least glamorous conclusion in this literature and the best supported. Most journals die of forgetting rather than failure, which makes the delivery question more interesting than the format question.
The clearest evidence on that point is Gold and colleagues’ randomised trial of three-good-things prompts sent by text message, where positive affect was still significantly improved at three months and retention through the study was high. Komase and colleagues’ review points the same way from the other direction: interventions asking for four entries or fewer changed nothing measurable.
So the practical question is not which notebook to buy. It is what will still be putting the question in front of you in week six.
