Is there really a reproducibility crisis in science?

In a 2016 Nature survey of 1,576 researchers, about 90% agreed there is a reproducibility crisis (52% called it significant, 38% slight), and only 3% said there was no crisis. More than 70% of those surveyed had tried and failed to reproduce another scientist's experiments, and more than half had failed to reproduce their own. Source: Baker, Nature 533:452-454 (2016).

What percentage of studies successfully replicate?

It varies by field and remains debated, but in the largest systematic replication in psychology, the 2015 Reproducibility Project, 97% of the original studies reported a statistically significant result, yet only 36% of the replications did, and replication effect sizes were on average about half the originals. Source: Open Science Collaboration, Science 349:aac4716 (2015).

Home › Guides › Research by the Numbers

Research by the numbers

Q: How common is research fraud and questionable research practice?

Outright fabrication is relatively rare but questionable practices are common. A 2009 meta-analysis of survey data found that about 2% of scientists admitted to having fabricated, falsified, or modified data at least once, while up to 33.7% admitted to other questionable research practices such as selective reporting. Source: Fanelli, PLoS ONE 4(5):e5738 (2009).

Q: How many research papers are retracted each year?

Retractions are rising sharply: more than 10,000 papers were retracted in 2023, a record, though the bulk that year came from mass retractions at journals formerly owned by Hindawi over compromised peer review. Retraction totals are still a small fraction of all papers published. Source: Van Noorden, Nature 624:479-481 (2023).

By Dr. Rafiq Muhammad, MD, PhD · Updated June 2026 · every figure cited to its peer-reviewed source

Researchers study everything except, often, research itself. But there is now a substantial body of meta-research — research on research — and the headline numbers are sobering. Here is what the best-cited evidence actually says about how reliable published science is, with a real source behind every figure. We’ve kept the caveats in: these numbers are easy to sensationalise, and the honest version matters more.

1. Most scientists say there’s a reproducibility crisis

~90%

In a 2016 Nature survey of 1,576 researchers, about 90% agreed there is a reproducibility crisis — 52% called it “significant,” 38% “slight.” Only 3% said there was no crisis. More than 70% had tried and failed to reproduce another scientist’s experiments, and more than half had failed to reproduce their own.^[1]

Baker, M. (2016), Nature.

2. When studies are repeated, many don’t hold up

97% → 36%

The 2015 Reproducibility Project replicated 100 psychology studies with high-powered designs. 97% of the originals had reported a statistically significant result — but only 36% of the replications did, and replication effect sizes were on average about half the originals.^[2] Replication rates differ across fields, and this is one (large) study in one discipline — but it reframed how seriously the problem is taken.

Open Science Collaboration (2015), Science.

3. Outright fraud is rare; questionable practices are not

~2% / 34%

A 2009 meta-analysis of survey data found that about 2% of scientists admitted to having fabricated, falsified, or modified data at least once — but up to 33.7% admitted to other questionable research practices, such as selective reporting of results. Asked about colleagues, the figures were far higher (14% for falsification, up to 72% for other practices).^[3]

Fanelli, D. (2009), PLoS ONE.

4. Retractions are climbing fast

10,000+

More than 10,000 papers were retracted in 2023 — a record.^[4] The honest caveat: most of that spike came from mass retractions at journals formerly owned by Hindawi over compromised peer review, so it reflects a clean-up as much as a worsening. Retractions are still a tiny fraction of all papers published — but the trend is steeply upward.

Van Noorden, R. (2023), Nature.

5. The paper that started the conversation

2005

Much of this traces back to John Ioannidis’s 2005 essay, bluntly titled “Why Most Published Research Findings Are False.” Using a simple probability model, it argued that small studies, small effects, flexible analysis, and publication pressure combine to make a large share of published claims unreliable.^[5] It remains one of the most-cited papers in meta-science.

Ioannidis, J. P. A. (2005), PLoS Medicine.

What this means for your research

None of this says science doesn’t work — it says method discipline is what separates findings that last from findings that evaporate. Every weakness in the numbers above has a concrete defence, and they’re the things we teach:

Underpowered studies → run a real power analysis and justify your sample.
p-hacking & selective reporting → pre-specify your analysis in a protocol, and read what a p-value actually means.
Unreliable measures → check validity and reliability (and Cronbach’s alpha) before you trust a scale.
Bias in the evidence base → appraise it with risk-of-bias tools and synthesise it in a systematic review.
AI-fabricated citations → the newest failure mode; verify every reference (see AI hallucinations).

The antidote to the numbers above is rigor you build in from the start. Our free methods guides and 18 free tools walk every step — design, power, analysis, reporting — so your findings are in the share that holds up.

Get the free Research Design toolkit

A design-decision worksheet and rigor checklist — the things that keep your study out of the failure statistics — from Research Design Simplified. We’ll email you the download link.

Get Research Design Simplified

Sources

Every figure on this page is drawn directly from the peer-reviewed or editorial sources below. Follow the links to read the originals.

Baker, M. (2016). 1,500 scientists lift the lid on reproducibility. Nature, 533(7604), 452–454. doi:10.1038/533452a
Open Science Collaboration (2015). Estimating the reproducibility of psychological science. Science, 349(6251), aac4716. doi:10.1126/science.aac4716
Fanelli, D. (2009). How many scientists fabricate and falsify research? A systematic review and meta-analysis of survey data. PLoS ONE, 4(5), e5738. doi:10.1371/journal.pone.0005738
Van Noorden, R. (2023). More than 10,000 research papers were retracted in 2023 — a new record. Nature, 624, 479–481. doi:10.1038/d41586-023-03974-8
Ioannidis, J. P. A. (2005). Why most published research findings are false. PLoS Medicine, 2(8), e124. doi:10.1371/journal.pmed.0020124

A note on honesty: these figures are widely cited but also widely over-cited out of context. Replication rates vary by field and method; the 2023 retraction spike is partly a one-off clean-up; and “questionable practices” covers a wide range of severity. We’ve linked every source so you can judge for yourself — which is rather the point.

Frequently asked questions

Is there really a reproducibility crisis?

In a 2016 Nature survey, ~90% of 1,576 researchers agreed there is one (only 3% disagreed), and >70% had failed to reproduce another’s work.^[1]

What share of studies replicate?

In the 2015 Reproducibility Project (psychology), 97% of originals were significant but only 36% of replications were.^[2] It varies by field.

How common is fraud?

~2% of scientists admit fabricating/falsifying data at least once; up to ~34% admit other questionable practices.^[3]

How many papers are retracted?

Over 10,000 in 2023 (a record) — though mostly from one publisher’s mass clean-up.^[4]

← All guides How to do a systematic review →