Pennebaker's protocol: fifteen to twenty minutes, on consecutive days, about the most upsetting thing you can write about, with no editing and no reader. The timer and the day count are here. The writing is not, because the protocol says it has no reader.
Before the first day. This asks you to write about the thing that upsets you most. For most people it is briefly hard and then useful. For some it is not: if you are in the middle of something raw, if writing about it has gone badly before, or if you have been told by somebody who knows you to leave it alone for now, leave it alone for now. It is fifteen minutes, not a treatment, and it works about as well as fifteen minutes can. Stop if you need to.
The session
15:00
Paper or a plain file. Write continuously. Do not stop to fix a sentence, and do not write it for anybody.
Days written
The honest limit, stated first. This is the best evidenced practice in the volume, and the honest news is how small that turns out to be. The 1986 paper that started the field found something real. The meta analysis twenty years later, across a large number of randomized studies, found an effect around a fifth of a standard deviation, far below what the first paper implied, and later trials in clinical populations often find nothing at all. The mechanism has moved three times in forty years, from inhibition to habituation to changes in the words people use, which is what a finding in search of a story looks like. Four days is the protocol as published, which is why this page counts consecutive days and nothing else. The writing never enters this page. That is not a privacy feature, it is the protocol: no editing, and no reader.