Contents
Academy · Research Data, Analytics & AI · The 0.05 Threshold
The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance
The conventional p<0.05 threshold derives from Ronald Fisher’s 1925 statistics text, in which he described it as a convenient rather than uniquely correct cut-off. The American Statistical Association’s 2016 statement formally asserts that no single threshold should be used to separate ‘significant’ conclusions from those that are not.
The threshold most commonly used to describe a result as “statistically significant” is a p-value below 0.05, meaning that the observed data, or data more extreme, would occur less than 5% of the time if the null hypothesis of no effect were true. This threshold is a convention, not a law of statistics or of nature, and its originator described it in precisely those terms when he introduced it. Tracing the origin of the 0.05 cut-off, and what the professional body representing statisticians has since said about reliance on any single fixed threshold, is important because current methodological guidance treats a convention adopted for historical and practical reasons very differently from a fixed law.
Learning objectives
- The origin of the 0.05 convention, and how its originator described it
- What the American Statistical Association's 2016 formal statement says about using a single fixed threshold
- Why some methodologists have proposed lowering the default threshold, and what that debate implies for meta-analysis specifically
Common misconceptions
- “0.05 is the statistically correct cut-off for significance, derived from a formal mathematical property of hypothesis testing.” Ronald Fisher’s 1925 text, which popularized the convention, described a p-value of around 0.05 as a convenient point at which to judge a result deserving of attention, explicitly a practical convention rather than a precise or uniquely justified mathematical boundary, and the American Statistical Association’s 2016 statement formally recommends against using any single bright-line threshold to separate ‘significant’ from ‘not significant’ findings.
Historical origin of the 0.05 convention
Ronald Fisher's 1925 text, Statistical Methods for Research Workers, is the most frequently cited origin of the 0.05 convention; in it, Fisher suggested that a result with approximately a 1-in-20 chance of occurring under the null hypothesis was a convenient, practical point at which researchers might judge a result worthy of serious consideration1. Historical and methodological accounts consistently describe Fisher’s own framing as explicitly provisional and context-dependent, a working convention for the agricultural and experimental research of his era, rather than a universally correct boundary derived from first principles2. That a round number arising from one researcher’s practical suggestion in 1925 became the default threshold across almost all of empirical science nearly a century later is a documented example of a convention hardening into an unquestioned rule, which is the pattern that later methodological critiques, including the American Statistical Association’s own statement, challenge3.
The 2016 statement of the American Statistical Association
In 2016, the American Statistical Association (ASA) published a formal statement on statistical significance and p-values, the first such statement in the organization’s history, explicitly to address what it described as widespread misuse and misunderstanding of p-values in the scientific literature3. Among its six numbered principles, the statement asserts that scientific conclusions should not be based solely on whether a p-value passes a specific threshold, and that a conclusion does not immediately become “true” on one side of a threshold and “false” on the other, which directly challenges the practice of treating 0.05 as a bright line separating meaningful findings from meaningless ones3. Wasserstein and Lazar, who wrote the statement on behalf of the association, state that the ubiquity of the 0.05 threshold reflects historical practice and publishing convention more than any principled statistical justification for that specific number over, for example, 0.04 or 0.063. The companion guide to common statistical misinterpretations by Greenland and colleagues reinforces the same point, describing the treatment of 0.05 as a strict decision boundary as one of the most consequential and widespread misunderstandings in applied statistics4.
Proposed alternatives and implications for meta-analysis
Partly in response to these concerns, a group of methodologists proposed in 2018 that fields relying on claims of novel discovery lower the conventional threshold for statistical significance to 0.005, reserving the label “significant” for a smaller, more stringent category of findings and recommending that results between 0.005 and 0.05 be described as “suggestive” rather than significant5. That proposal remains debated rather than adopted as a new universal standard, but its underlying argument, that any chosen threshold should be treated as a deliberate, field-appropriate convention rather than an inherited default, applies directly to the reporting of meta-analyses and systematic reviews. The Cochrane Handbook directs review authors to emphasize the size of the pooled effect and its confidence interval over a single significant or non-significant verdict, because a fixed threshold, whatever its exact value, reduces a continuous measure of evidence to a binary label, a reduction against which both Fisher’s original framing and the later ASA statement caution6.
One-sided and two-sided thresholds and the importance of the distinction
The conventional 0.05 threshold is usually applied in a two-sided test, which allows for the possibility that the true effect runs in either direction, favoring the intervention or the control, and divides the 5% error rate between both tails of the distribution, typically 2.5% in each6. A one-sided test instead places the entire 5% in a single direction, which makes significance easier to reach when a result falls in the direction being tested, but at the cost of being unable to detect a significant effect in the opposite direction, however large that effect may be. Switching from a planned two-sided test to a one-sided test after observing the direction of the data is a documented means of making a marginal result appear more significant than the same data would under the originally planned two-sided analysis, which is why the Cochrane Handbook expects the choice between one-sided and two-sided testing to be pre-specified in the review protocol rather than decided once the direction of the observed effect is known7.
Implications for the statement of review conclusions
Taking the historical and statistical context together, a defensible approach for a systematic review is to report the exact calculated p-value, state which threshold convention (almost always 0.05, two-sided) was pre-specified in the protocol, and base the conclusion on the pooled effect size and its confidence interval rather than on a bare statement that the result “was” or “was not” significant7. This keeps the review consistent both with Fisher’s original framing of the threshold as a practical convention and with the ASA’s explicit 2016 recommendation against reducing a continuous measure of evidence to a binary verdict, while providing readers with the exact numbers they need to judge the strength of the evidence. A review that applies this approach consistently to every outcome, rather than selectively to favorable outcomes while reverting to dichotomous language for less favorable ones, is also less exposed to the appearance of selective reporting, because a consistent interpretive standard applied regardless of outcome is one of the more visible and verifiable markers of an impartially conducted review. Journal editors and peer reviewers can apply the same check: examining whether every outcome in a results section, favorable or unfavorable, receives the same depth of interpretation beyond a bare significance verdict is an efficient means of identifying a manuscript that may present its findings more favorably than the underlying numbers support. The same check applies to a research group’s body of work over time: a group that routinely reports significant findings with full effect sizes and confidence intervals, while mentioning non-significant findings only briefly with a bare p-value, applies an asymmetric reporting standard that careful readers, and increasingly careful reviewers, are likely to notice. One internal safeguard before submission is for a co-author who was not involved in the analysis to read only the results section and note, for each outcome, whether the depth of interpretation given to a favorable result matches that given to an unfavorable one, since an outside reader is often better placed to detect this asymmetry than the authors who produced the numbers. Including this check as a standing item on the internal pre-submission checklist, alongside more familiar checks such as reference accuracy and formatting compliance, prevents it from being omitted under the time pressure that typically accompanies final submission. None of this requires treating 0.05 as meaningless; it means that the number should appear in a results section together with the context a reader needs to use it responsibly, rather than substituting, on its own, for the conclusion that the data can support.
Key points
Key points
The 0.05 significance threshold is a historical convention traced to Fisher's 1925 text, not a derived statistical law, and the American Statistical Association's 2016 statement formally recommends against treating any single fixed threshold as a bright line between meaningful and meaningless results.
Practical application
Practical application
- The exact p-value and confidence interval should be reported, not only a significant or non-significant verdict. This preserves the continuous information that a fixed threshold discards, consistent with the guidance in the American Statistical Association (ASA) statement.
- A result just above or below 0.05 should not be described as categorically different. P-values of 0.04 and 0.06 reflect very similar evidence; treating them as qualitatively different outcomes overstates what the threshold itself can justify.
- The significance threshold should be stated in the methods as a deliberate, field-appropriate choice. Whether 0.05 or a different value is used, it should be documented as a chosen convention rather than an unexamined default.
Frequently asked questions
No. Fisher’s 1925 text described a result with approximately a 1-in-20 chance under the null hypothesis as a convenient, practical point for judging a result worthy of attention, explicitly a working convention rather than a derived mathematical boundary.
No. Its 2016 statement cautions against using p-values, or any single threshold, as the sole basis for a scientific conclusion; it does not argue that p-values have no legitimate role in statistical reasoning.
No. A 2018 proposal to lower the default threshold to 0.005 remains debated rather than broadly adopted, and the Cochrane Handbook and ASA statement both emphasize effect size and confidence intervals over any single fixed threshold regardless of its exact value.
Cochrane guidance favors emphasis on the pooled effect size and its confidence interval over a binary significant or non-significant verdict, although reporting the exact p-value alongside these remains standard and informative.
References
- Fisher RA. Statistical Methods for Research Workers. Oliver & Boyd; 1925. read 10 October 2026. View on Internet Archive
- Kennedy-Shaffer L. Before p < 0.05 to beyond p < 0.05: Using History to Contextualize p-Values and Significance Testing. The American Statistician. 2019;73(sup1):82-90. read 10 October 2026. View on Taylor & Francis Online
- Wasserstein RL, Lazar NA. The ASA Statement on p-Values: Context, Process, and Purpose. The American Statistician. 2016;70(2):129-133. read 10 October 2026. View on Taylor & Francis Online
- Greenland S, Senn SJ, Rothman KJ, Carlin JB, Poole C, Goodman SN, Altman DG. Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. European Journal of Epidemiology. 2016;31(4):337-350. read 10 October 2026. View on Springer
- Benjamin DJ, Berger JO, Johannesson M, et al. Redefine statistical significance. Nature Human Behaviour. 2018;2:6-10. read 10 October 2026. View on Nature Human Behaviour
- Schünemann HJ, Higgins JPT, Vist GE, et al. Chapter 15: Interpreting results and drawing conclusions. Cochrane Handbook for Systematic Reviews of Interventions. read 10 October 2026. View on training.cochrane.org
- Deeks JJ, Higgins JPT, Altman DG. Chapter 10: Analysing data and undertaking meta-analyses. Cochrane Handbook for Systematic Reviews of Interventions. read 10 October 2026. View on training.cochrane.org
- Goodman S. A dirty dozen: twelve p-value misconceptions. Seminars in Hematology. 2008;45(3):135-140. read 10 October 2026. View on Seminars in Hematology
- Sullivan GM, Feinn R. Using Effect Size-or Why the P Value Is Not Enough. Journal of Graduate Medical Education. 2012;4(3):279-282. read 10 October 2026. View on JGME
- Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd ed. Lawrence Erlbaum Associates; 1988. read 10 October 2026. View on routledge.com
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 10). The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance. Pubrica. https://pubrica.com/academy/threshold-used-for-statistical-significance/
Pubrica Academy. “The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance.” Pubrica, 10 Oct. 2026, pubrica.com/academy/threshold-used-for-statistical-significance/.
Pubrica Academy (2026) The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance. Pubrica. Available at: https://pubrica.com/academy/threshold-used-for-statistical-significance/ (Accessed: 15 October 2026).
Pubrica Academy. “The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance.” Pubrica. October 10, 2026. https://pubrica.com/academy/threshold-used-for-statistical-significance/
Pubrica Academy, ‘The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/threshold-used-for-statistical-significance/> accessed 15 October 2026.
Pubrica Academy. The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/threshold-used-for-statistical-significance/
Pubrica Academy, “The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/threshold-used-for-statistical-significance/
@misc{/la_statistical_significance_threshold_2026,
author = {{Pubrica Academy}},
title = {The Threshold Commonly Used to Determine Statistical Significance: Origins of the 0.05 Convention and Current Guidance},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/threshold-used-for-statistical-significance/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Significance Threshold Reporting Note
For use before finalizing how a report frames statistical significance.
[ ] Reported exact p-values and confidence intervals rather than only a significant/non-significant label [ ] Avoided describing results near the threshold (e.g. 0.04 vs 0.06) as categorically different [ ] Documented the chosen significance threshold explicitly in the methods section [ ] Emphasized effect size and confidence interval alongside, not instead of, the p-value [ ] Checked target journal/field conventions before deviating from the standard 0.05 threshold
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica's biostatistics team frames significance reporting around effect sizes and confidence intervals rather than a bare threshold verdict, consistent with current ASA and Cochrane guidance.