Start here

01Three kinds of one-person experiment

Researchers run one-person experiments in two traditions, and people run a third kind on their own.

Clinical n-of-1 trials come from medicine. One person goes through several pairs of periods, with and without the thing being tested, in an order set at random, and often without knowing which period is which. A 1986 paper set out this design for everyday medical practice (Guyatt et al., 1986). A later review counted 108 such trials, with 2,154 people, published between 1985 and 2010 (Gabler et al., 2011).

Single-case designs come from education and behavioural research. A baseline stretch is followed by a change, and the pattern is repeated and read from a chart. Their standards ask for an effect to show up at least three times, at three different points in time (Kratochwill et al., 2010).

Everyday self-experiments are what people do with a notebook or an app. You usually know what you changed, and you choose when. Researchers have proposed borrowing single-case methods to make these sturdier (Karkar et al., 2016).

02Where each sits on the evidence ladder

In one widely used table of evidence, the top level for the question “does this intervention help?” is either a randomised n-of-1 trial or a systematic review of randomised trials (Oxford CEBM, 2011). That standing belongs to repeated trials in a random order, usually with the person kept unaware of which period is which.

A single open test, at a time you choose, is a long way from that. Single-case research calls it an AB design: one stretch before, one after. Its standards say a design like this makes it hard to draw a sound conclusion, because with no repeat it can’t rule out other explanations (Kratochwill et al., 2010).

03What a careful test can and can’t tell you

A careful test can show whether your ratings differ with and without a change, in your own life, at a size that matters to you, but not whether the change or your expecting it made the difference. With a baseline and repeats, it can make a hunch stronger or weaker.

It can’t, on its own, show what was behind a change, or say anything about other people. Results are often unclear even in formal trials: in that review of published n-of-1 trials, 38% of the people with follow-up information had results that were ambiguous (Gabler et al., 2011). In one randomised trial of 215 people, those whose one-person trials were set up with their clinicians and run through an app did no better on its main measure than people having their usual care (Kravitz et al., 2018).

04The ingredients of a careful test

Each page takes one ingredient:

1. Why test on yourself: people differ, but much of what looks like difference is everyday noise.
2. Write it down on the day: why notes made the same day beat memory.
3. Choose what to measure: a short, fixed set of ratings, and one main measure named in advance.
4. Baseline first: a stretch of ordinary days, and a change threshold set before you start.
5. Design the test: one change at a time, repeated, in a balanced or random order.
6. Bringing a food back: how research structures a reintroduction, and what belongs with a doctor.
7. Record what else was going on: sleep, stress and other new foods.
8. Expectation: the weak point of a test where you know what you changed.
9. Reading your result: how researchers read a chart, and why “inconclusive” is an answer.
10. Safety, and when to involve a clinician: when a home test is the wrong tool.

Every source is listed on the References page. Research words are explained in the Glossary.

05Who made this guide

This guide was written by Sean, who makes n1lab and is not a clinician, with an AI research assistant. Every source was checked against its PubMed or DOI record, or the publisher’s own page. No clinician has reviewed it. How this guide was made says how the sources were found and checked, what was left out and why, and how to report a mistake.

06What n1lab is, and isn’t

n1lab is a notebook for one-person experiments: it keeps the record. It doesn’t hide what you’re testing, choose when a test runs, or repeat, randomise or blind a test for you. It isn’t medical advice: see what n1lab is not, in the terms.

07Sources on this page

Each source is listed once, with links to its record. The reference list says what was read for each, and when it was checked.

1. Gabler et al., 2011Gabler NB, Duan N, Vohra S, Kravitz RL. 2011. N-of-1 trials in the medical literature: a systematic review. Medical Care 49(8):761–768.DOI 10.1097/MLR.0b013e318215d90d · PubMed 21478771
2. Guyatt et al., 1986Guyatt G, Sackett D, Taylor DW, Chong J, Roberts R, Pugsley S. 1986. Determining optimal therapy—randomized trials in individual patients. New England Journal of Medicine 314(14):889–892.DOI 10.1056/NEJM198604033141406 · PubMed 2936958
3. Karkar et al., 2016Karkar R, Zia J, Vilardaga R, Mishra SR, Fogarty J, Munson SA, et al. 2016. A framework for self-experimentation in personalized health. Journal of the American Medical Informatics Association 23(3):440–448.DOI 10.1093/jamia/ocv150 · PubMed 26644399 · PMC6095104
4. Kratochwill et al., 2010Kratochwill TR, Hitchcock J, Horner RH, Levin JR, Odom SL, Rindskopf DM, et al. 2010. Single-case designs technical documentation. What Works Clearinghouse, US Institute of Education Sciences. Version 1.0 (pilot), June 2010.ies.ed.gov
5. Kravitz et al., 2018Kravitz RL, Schmid CH, Marois M, Wilsey B, Ward D, Hays RD, et al. 2018. Effect of Mobile Device-Supported Single-Patient Multi-crossover Trials on Treatment of Chronic Musculoskeletal Pain: A Randomized Clinical Trial. JAMA Internal Medicine 178(10):1368–1377.DOI 10.1001/jamainternmed.2018.3981 · PubMed 30193253 · PMC6233756
6. Oxford CEBM, 2011OCEBM Levels of Evidence Working Group. 2011. The Oxford 2011 Levels of Evidence. Oxford Centre for Evidence-Based Medicine.cebm.ox.ac.uk