Contents
Academy · Research Data, Analytics & AI · Origin of the Name
Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape
The chart is called a funnel plot because, in the absence of bias, large and precise studies cluster tightly near the top and small, imprecise studies scatter widely near the bottom, tracing the outline of an inverted funnel. Egger and colleagues built this shape directly into their 1997 diagnostic test.
The name “funnel plot” describes the shape that the chart is expected to take; it is not an arbitrary label. When study precision is plotted correctly on the vertical axis and effect size on the horizontal axis, a body of unbiased evidence produces a scatter that is wide at the bottom, where small, imprecise studies may by chance fall relatively far from the true effect, and narrow at the top, where large, precise studies cluster closely around it. Oriented with the wide end at the bottom and the narrow end at the top, this scatter traces the outline of an inverted funnel. Egger and colleagues built this expected shape directly into the diagnostic logic of their 1997 paper, from which point the name and the method became inseparable.
Learning objectives
- The statistical reason that the plot resembles a funnel in the absence of bias
- How the name relates to the diagnostic logic for which the chart is used
- Why a plot that does not resemble a funnel is itself diagnostically informative
Common misconceptions
- “The ‘funnel’ in funnel plot is merely a descriptive label with no bearing on how the chart is read.” Egger and colleagues' 1997 paper and Egger and Sterne's 2001 follow-up both tie the expected funnel shape directly to the statistical relationship between study precision and effect-size variability, which means a plot's resemblance, or lack of resemblance, to that shape is the actual basis for the method's diagnostic use.
Statistical basis of the funnel shape
The funnel shape follows directly from the behaviour of sampling variability as a study’s sample size increases. The effect estimate of a small study is based on relatively little data, so ordinary sampling error can move its result some distance from the true underlying effect in either direction, producing a wide spread of points at the bottom of the chart, where these imprecise studies are located1. The effect estimate of a large study is based on considerably more data, so the same sampling error has far less capacity to move the result away from the true effect, producing a tight cluster of points at the top of the chart, where these precise studies are located. When many studies of varying size are plotted together, this pattern, wide at the bottom and narrow at the top, traces two converging diagonal boundaries that meet at the pooled effect near the top, the visual characteristic from which the chart takes its name.
Egger and Sterne’s 2001 methodological analysis formalised this relationship by specifying standard error as the appropriate precision measure for the vertical axis, because it preserves the funnel-shaped relationship consistently across different effect measures, including mean differences, risk ratios and log odds ratios2. Had sample size been recommended instead, the resulting plots would not reliably form this shape, which is one of the specific reasons the paper gives for rejecting sample size as the axis variable; a chart that does not reliably resemble a funnel under the conditions that give it its name would be a poor foundation for a method designed to detect departures from that shape.
Relationship between the name and the diagnostic method
The original 1997 BMJ paper by Egger and colleagues, which introduced the now-standard graphical test for funnel-plot asymmetry, used the expected funnel shape as the direct basis for the test: a formal statistical assessment of whether the observed scatter deviates from the symmetrical, converging pattern that the funnel shape describes in the absence of bias1. The name is therefore not incidental to the method; the test is, in a precise sense, a test of whether the chart conforms to the funnel that its name describes. The consolidated 2011 guidance by Sterne and colleagues retains this shape-based framing throughout, describing asymmetry explicitly as a departure from the expected funnel outline rather than introducing a separate, shape-independent definition of what the test assesses4.
The contour-enhanced funnel plot introduced by Peters and colleagues in 2006 extends the same shape-based logic by overlaying statistical-significance contours directly onto the funnel outline, allowing a reader to see where within the expected funnel shape any gap or asymmetry falls, for example within the region in which non-significant small studies would be expected, which is the pattern that publication bias is expected to produce5. The contour-enhanced version therefore remains a more detailed means of addressing the same shape-based question that the name was coined to describe in 1997.
Interpretation of an absent funnel shape
Because the funnel shape constitutes the entire diagnostic premise of the chart, a plot that does not resemble one, showing a gap, an asymmetry, or an unexpectedly flat or irregular scatter in place of the expected converging outline, is itself the finding that the method is designed to identify. The 2011 guidance by Sterne and colleagues lists several specific causes of such a departure: reporting biases, differences in methodological quality correlated with study size, genuine heterogeneity, statistical artefacts associated with the chosen effect measure, and chance, particularly when few studies are included4. The earlier 2000 analysis by Tang and Liu had already identified heterogeneity as an underappreciated cause of such departures, one that a reader relying solely on the intuitive picture conveyed by the name may readily mistake for publication bias3. The 2006 editorial by Lau and colleagues sets out the broader caution that follows: a departure from the expected shape is a prompt to investigate which of these specific causes applies, not a self-evident confirmation of any one of them, including publication bias6.
The 2005 empirical study by Terrin, Schmid, and Lau adds a further practical caution of direct relevance to the name: researchers asked to judge from the shape of a chart alone whether it resembled the expected funnel outline or showed a biased departure from it were accurate only 53% of the time, essentially chance level for a binary judgement7. The name describes a real, mathematically grounded expectation, but determining whether a specific plot meets or departs from that expectation is considerably more difficult for the human eye than the idealised textbook version of the shape suggests. For this reason, the Cochrane Handbook and the guidance by Sterne and colleagues both pair visual inspection of the shape with formal statistical tests rather than relying on the intuitive picture conveyed by the name alone84.
Related plots based on the same principle
Several later refinements of the basic funnel plot retain the underlying funnel shape but add further visual information to it, rather than replacing the shape-based logic that the name describes. The contour-enhanced version by Peters and colleagues, discussed above, retains the funnel outline and overlays significance-threshold shading within it, so the chart must still resemble a funnel for the added contours to be meaningful5. The trim-and-fill method of Duval and Tweedie goes further by adding imputed data points, typically displayed in a different colour or shape from the observed studies and positioned to restore the expected symmetry of the funnel; the visual output of the method is thus a direct demonstration of the shape the plot is assumed to take once the suspected missing studies are restored9.
The same shape-based reasoning extends conceptually to other tools, including some outside the graphical display itself. The rank correlation test of Begg and Mazumdar, for instance, assesses whether the ranking of studies by effect size correlates with their ranking by precision, a numerical means of establishing whether the funnel shape holds without requiring a visual judgement8. Across all of these variations, whether contour-enhanced, trim-and-fill or rank-correlation-based, the reference point remains the inverted-funnel outline that Egger and colleagues described in 1997, and an understanding of that original shape and its statistical basis therefore remains the foundation for interpreting any of its later, more elaborate derivatives correctly.
The origin of the name is not merely of historical interest; it indicates what the chart claims to show whenever it appears in a published review. A figure labelled “funnel plot” implicitly asserts that the scatter it displays should be read against the specific, mathematically defined expectation that Egger and colleagues described in 1997: wide and dispersed at the bottom, narrow and clustered at the top, and converging on the pooled estimate. Any discussion of what the shape of a particular plot does or does not indicate is fundamentally a discussion of how closely that chart matches or departs from the shape its name implies. The name therefore warrants proper understanding rather than treatment as an arbitrary label attached to a generic scatter chart, and a reviewer who encounters such a plot in a manuscript is entitled to ask how closely it was checked against that expectation before any conclusion was drawn, rather than accepting the label as a substitute for that check, since the name denotes a specific statistical property rather than a superficial resemblance.
Key points
Key points
A funnel plot is named for the inverted-funnel shape that sampling variability produces in the absence of bias: wide and scattered for small, imprecise studies, and narrow and clustered for large, precise ones. Egger’s 1997 test and every subsequent refinement of it are built on assessing whether an observed plot matches or departs from that shape.
Practical application
Practical application
- Use the funnel shape as an initial, informal check before any formal test. A plot that does not narrow toward the top may indicate a construction error, such as an incorrect precision measure, before any question of bias arises.
- Recognize that the name describes an expectation in the absence of bias, not a guarantee. An observed plot that departs from the funnel shape is informative, but the accuracy findings of Terrin, Schmid and Lau indicate that this judgement is more difficult to make reliably than the idealised textbook shape suggests.
- Pair visual inspection of the shape with a formal test rather than relying on the shape alone. This is consistent with the approach recommended in the 2011 guidance by Sterne and colleagues and in the Cochrane Handbook.
Frequently asked questions
The shape and its diagnostic use were formalised in meta-analysis methodology through the widely cited 1997 BMJ paper by Egger and colleagues, which built the now-standard graphical asymmetry test directly on the expected funnel-shaped scatter.
Only under the idealised no-bias, no-heterogeneity scenario the name describes. A real plot can depart from that shape for several documented reasons, which is the basis for using the chart diagnostically in the first place.
No. The shape, and departures from it, relate to the broader concept of a ‘small-study effect’, of which publication bias is only one of several possible causes identified in current methodological guidance.
References
- Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315(7109):629-634. read 10 October 2026. View on The BMJ
- Egger M, Sterne JAC. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. Journal of Clinical Epidemiology. 2001;54(10):1046-1055. read 10 October 2026. View on Journal of Clinical Epidemiology
- Tang JL, Liu JLY. Misleading funnel plot for detection of bias in meta-analysis. Journal of Clinical Epidemiology. 2000;53(5):477-484. read 10 October 2026. View on Journal of Clinical Epidemiology
- Sterne JAC, Sutton AJ, Ioannidis JPA, et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials. BMJ. 2011;342:d4002. read 10 October 2026. View on The BMJ
- Peters JL, Sutton AJ, Jones DR, Abrams KR, Rushton L. Comparison of two methods to detect publication bias in meta-analysis. JAMA. 2006;295(6):676-680. read 10 October 2026. View on JAMA
- Lau J, Ioannidis JPA, Terrin N, Schmid CH, Olkin I. The case of the misleading funnel plot. BMJ. 2006;333(7568):597-600. read 10 October 2026. View on The BMJ
- Terrin N, Schmid CH, Lau J. In an empirical evaluation of the funnel plot, researchers could not visually identify publication bias. Journal of Clinical Epidemiology. 2005;58(9):894-901. read 10 October 2026. View on Journal of Clinical Epidemiology
- Sterne JAC, Egger M, Moher D (editors). Chapter 10: Addressing reporting biases. In: Higgins JPT, Green S, editors. Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0. The Cochrane Collaboration; 2011. read 10 October 2026. View on Cochrane Training
- Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics. 1994;50(4):1088-1101. read 10 October 2026. View on PubMed
- Duval S, Tweedie R. Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455-463. read 10 October 2026. View on Biometrics
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). Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape. Pubrica. https://pubrica.com/academy/why-is-it-called-a-funnel-plot/
Pubrica Academy. “Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape.” Pubrica, 10 Oct. 2026, pubrica.com/academy/why-is-it-called-a-funnel-plot/.
Pubrica Academy (2026) Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape. Pubrica. Available at: https://pubrica.com/academy/why-is-it-called-a-funnel-plot/ (Accessed: 15 October 2026).
Pubrica Academy. “Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape.” Pubrica. October 10, 2026. https://pubrica.com/academy/why-is-it-called-a-funnel-plot/
Pubrica Academy, ‘Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/why-is-it-called-a-funnel-plot/> accessed 15 October 2026.
Pubrica Academy. Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/why-is-it-called-a-funnel-plot/
Pubrica Academy, “Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/why-is-it-called-a-funnel-plot/
@misc{/la_why_called_funnel_plot_2026,
author = {{Pubrica Academy}},
title = {Origin of the Term ‘Funnel Plot’: The Statistical Basis of the Inverted-Funnel Shape},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/why-is-it-called-a-funnel-plot/},
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.
Funnel Shape Quick-Read Guide
This checklist supports the assessment of a funnel plot’s shape before a formal test is applied.
[ ] Confirmed the plot narrows toward the top (precise, large studies) and widens toward the bottom (imprecise, small studies) [ ] Checked whether the scatter looks roughly symmetrical around the pooled effect [ ] Noted any visible gap and which side/region of the expected shape it falls in [ ] Treated the visual read as a prompt for investigation, not a final answer [ ] Followed up with a formal test only if enough studies are included
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 meta-analysis team interprets each funnel plot against its underlying statistical expectation, not only its surface appearance, before drawing any conclusion from its shape.