Replication Crisis

More than half the findings in some scientific fields can't be reproduced. Let that sink in.

In 2005, John Ioannidis published a paper arguing that most published research findings are false. It sounded extreme. Then researchers started checking. The Open Science Collaboration tried to replicate 100 psychology studies — only 36 produced the same results. Cancer biology fared no better: Amgen scientists attempted to reproduce 53 landmark studies and succeeded with just 6. These aren't fringe papers. These are the studies that textbooks are built on.

The causes are structural. Careers depend on publishing novel, positive results. Journals prefer exciting findings over boring replications. Funding rewards quantity, not verification. So researchers face constant pressure to produce impressive-looking data, whether or not it reflects reality. Add in small sample sizes, flexible analysis methods, and publication bias, and you get a system that manufactures confidence out of noise.

This doesn't mean all science is broken. It means the self-correcting mechanism — replication — has been neglected for decades. When a finding hasn't been independently reproduced, treat it as a tentative claim, not an established fact. The crisis isn't that science failed. It's that people forgot which part of science actually works.

---

References

• John Ioannidis — Why Most Published Research Findings Are False (2005)
• Richard Harris — Rigor Mortis (2017)
• Ben Goldacre — Bad Science (2008)