← All posts

What expressive writing research actually found

The original 1986 study and a meta-analysis of 146 experiments. The effect is real, small, and easy to oversell. Here are the numbers the papers actually report.

920four · October 13, 2026 · 4 min read

Search for "journaling benefits" and you will meet a genre. Writing for a few minutes heals trauma, boosts immunity, and changes your life. Some of that genre is pointing at a real research program. Almost all of it skips the size of the effect and the warning the original authors printed on the work.

This is that warning, with the citations.

The 1986 study

James W. Pennebaker and Sandra K. Beall published "Confronting a traumatic event: Toward an understanding of inhibition and disease" in the Journal of Abnormal Psychology in 1986. The American Psychological Association's record of the paper describes the design:

  • 46 introductory psychology students took part.
  • They wrote on 4 consecutive days.
  • They wrote either about personally traumatic events or about trivial topics.
  • The researchers looked at short-term physiological arousal and mood, and at health-center visits over the following months.

The finding that gets quoted, and that the abstract actually states, is specific. Writing about both the emotions and the facts of a traumatic event was associated with relatively higher blood pressure and more negative mood right after the essays, and with fewer health-center visits in the 6 months after the experiment.

The authors' own conclusion, in that same abstract, is that the findings should be considered preliminary. They situate the result among questions of catharsis, self-disclosure, and a theory of psychosomatics based on behavioral inhibition. They do not claim a treatment you can prescribe from a blog post.

Two things are easy to drop when this study is retold. The writing that lined up with fewer clinic visits was the writing that included emotions and facts together, not a gratitude list and not a log of the day's errands. And the immediate experience was often worse, not better. People felt it. Blood pressure rose. The interesting result showed up later, in health-center records, in a small sample of students.

What 146 experiments add up to

Twenty years later, Joanne Frattaroli published "Experimental disclosure and its moderators: A meta-analysis" in Psychological Bulletin (2006, Vol. 132, No. 6, 823–865). She collected 146 randomized studies of experimental disclosure: people asked to disclose thoughts and feelings about something personally meaningful, usually in writing, compared with a control.

Using a random-effects model, she found a positive and statistically significant average effect of r = .075. That is the number in the abstract. It is a real effect across a large set of experiments. It is also a small correlation. A product page that turns r = .075 into "journaling will change your health" is not summarizing the paper.

Frattaroli also looked at what varied between studies. In the discussion she reports that studies with at least three disclosure sessions had a marginally larger overall effect (r = .082) than studies with fewer than three (r = .040), and that sessions of at least 15 minutes (r = .080) differed from sessions under 15 minutes (r = .007) on the overall effect. She is careful about what that means. Studies were not randomly assigned to those conditions, moderator variables were correlated with each other, and she describes the moderator results as a starting point for later tests rather than proof of cause.

She also notes that Pennebaker's own public guidance at the time recommended writing for at least 15 minutes across at least three sessions. That is advice attached to the paradigm, not a law the meta-analysis proved.

What we will not claim

WriteFlow is a place to write. It is not a clinical intervention, and using it is not the Pennebaker protocol. The protocol, when researchers run it, usually asks people to write about something emotionally significant for a set number of sessions, often with a control group writing about trivial topics. A calm diary, a project note, and a trauma narrative are different tasks. The research does not say they do the same thing.

If you want to try the paradigm itself, the useful reading is the method section of the papers, not a thread that borrowed the citation. If you want a private page to think in writing, that is a different, simpler promise. Both can be true without one pretending to be the other.

The result worth keeping is the modest one. Across many experiments, brief expressive disclosure beat doing a control writing task by a small average amount. In the first study, the benefit that showed up in clinic visits came with a worse mood on the days people actually wrote. Anyone selling a guaranteed transformation is selling past the data.

Sources

Related: attention residue, and why the page should stay quiet.