How to Communicate Null Results From Your Research
A practical guide to sharing a study that found no effect without overclaiming or apologizing. It covers what a null result does and does not show, wording that stays accurate, where null results get published and how to explain one to people outside your field.
You ran the study properly. You preregistered it, recruited the sample you planned for and analysed it the way you said you would. The effect everyone expected is not there.
The paper is written and accepted, and now you are unsure what to do with it. Posting "we found nothing" feels like announcing a failure. Staying quiet means the one study that could save someone else a year of work disappears.
Why Null Results Need Saying Out Loud
The published record leans hard toward positive findings. In a 2010 analysis of 2,434 papers that declared they had tested a hypothesis, Daniele Fanelli found that the odds of reporting a positive result were around 5 times higher in Psychology and Psychiatry, and in Economics and Business, than in Space Science. For studies applying behavioural and social methods to people, the odds were 3.4 times higher than for physical and chemical studies on non-biological material.
That imbalance is why a well-designed null result is useful. It tells the next team which effect is unlikely to be large, and it gives meta-analysts a data point the literature would otherwise be missing. Your job when you share it is to say clearly what the study could and could not rule out.
What a Null Result Shows, and What It Does Not
This is where most null results get miscommunicated, often by their own authors. In his practical primer on equivalence tests, Daniël Lakens points out that when p is larger than the alpha level, the only formally correct conclusion is that the data are not surprising if the null hypothesis is true. The study may simply have lacked the power to detect a true effect.
He also gives a sense of how common the slip is: the words "no effect" appeared in 108 articles in Social Psychological and Personality Science up to August 2016, and on manual inspection almost all of them based that conclusion on statistical nonsignificance alone.
The alternative he describes is an equivalence test. You set bounds based on the smallest effect size you care about, and if the two one-sided tests reject effects outside those bounds, you can act as if the true effect is close enough to zero for practical purposes. If you planned one, say so. If you did not, describe the smallest effect your design had reasonable power to detect.
A Worked Example: Rewriting the Headline
Say you ran a preregistered study testing whether a ten-minute writing exercise before an exam reduces test anxiety in first-year students. The difference between the groups was not statistically significant. Here is how the same result can be described at different levels of accuracy.
| Draft wording | The problem | Accurate version |
|---|---|---|
| "Writing exercises do not work for test anxiety." | Generalizes from one design and one population to all exercises. | "In our sample of first-year students, we did not detect a reduction in test anxiety after a ten-minute writing exercise." |
| "We found no effect." | Treats nonsignificance as proof of zero. | "We did not find the difference we powered the study to detect." |
| "The study failed." | Confuses an unexpected result with a flawed study. | "The study ran as preregistered; the expected effect did not appear." |
| "Disappointingly, nothing was significant." | Apologizes, which invites readers to discount the work. | "The expected effect did not appear; the data do not rule out a smaller effect." |
The right-hand column is not spin. Each line says less than the left-hand one, and that is the point: it stops at what the data support.
Where Null Results Get Published
You do not need to file the study away. At the time of writing (October 2026), PLOS One's criteria for publication state that, in keeping with its mission to publish all valid research, it considers negative and null results. The same page says it considers Registered Report protocols and the research articles that report their outcomes.
Registered Reports change the order of events. The Center for Open Science explains that editors and reviewers assess the design at Stage 1, before data collection, and that in-principle acceptance virtually guarantees publication, provided the study is carried out as the approved protocol describes. It also notes that a PhD student using the format could publish every high-quality experiment from their PhD rather than selectively publishing the ones with positive results. If you are planning your next study, this is worth knowing before you collect a single data point.
How to Share It Beyond Your Field
Lead with the question, not the p value. "Does a short writing exercise calm exam nerves?" is something a teacher or student counsellor will stop to read. Then give the answer in one plain sentence and, in another, the smallest effect the study was designed to detect (or, if you ran an equivalence test, the effects it ruled out).
Three habits help:
- Say what the study was designed to detect before saying what it found.
- Name the population and the setting, so nobody stretches the result further than it goes.
- End with what this changes: which studies are worth running next, or which practice deserves a closer look.
The same rules apply in a lay summary as in a social post. The guide to writing a plain language summary covers the structure, and the LinkedIn post templates for your paper work just as well when the finding is a null one.
Where Amplify Fits
Null results are easy to distort in a summary, so the tool you use to explain one matters. Amplify works only from the PDF you upload, with no open-web filler, so the formats start from your own careful wording rather than a general claim about the topic.
Every format, the video abstract, the vertical short, the audio brief and the infographic, arrives as three variants that you can edit line by line. If a line slides from "did not find the difference we powered the study to detect" to "no effect", you change it. A subject-matter reviewer reads every draft before it can leave your workspace, every claim is traced to a page in the PDF, and nothing publishes until you approve it.
The embeddable widget then puts all four formats beside the paper on your own lab or profile page, and the analytics show reads, listens, questions asked and which format moved people. For a null result, the questions readers ask can tell you which part of the design still needs explaining.
Null results will always be a harder story to tell than a surprising effect. Told accurately, they are also one of the most useful things a field can read.
Quick answers
Does a nonsignificant result mean there is no effect?
No. As Lakens explains in his primer on equivalence tests, when p is larger than the alpha level the only formally correct conclusion is that the data are not surprising if the null hypothesis is true. The study may simply have lacked the power to detect a true effect, so "no effect" is usually stronger than the data allow.
How should I word a null result in my paper or summary?
Say what the study tested, how large an effect it could detect, and what it did not find. "We found no effect" overclaims; "We did not find the difference we powered the study to detect" stays within the data. If an equivalence test rejected effects outside the bounds you set in advance, you can say the effect was close enough to zero for practical purposes.
Where can I publish a study with null results?
Some journals say so explicitly. At the time of writing (October 2026), PLOS One's publication criteria state that it considers negative and null results. Registered Reports are another route: the Center for Open Science describes them as peer reviewed before data collection, with in-principle acceptance that virtually guarantees publication as long as the study follows the approved protocol.
Is it worth promoting a paper that found nothing?
Yes, if the study was well designed. The published record leans hard toward positive findings, and a careful null result tells the next team which effect is unlikely to be large while giving meta-analysts a data point they would otherwise miss. When you share it, say clearly what the study could and could not rule out, in plain words.
