Many scientific studies cannot be replicated, questioning their validity.
The Open Science Collaboration's 2015 attempt reproduced 36 of 100 psychology findings. Cancer biology and economics have shown comparable rates, so the problem is not confined to one field.
In practice
- A textbook chapter rests on four studies, three of which have failed to reproduce, and the chapter is unchanged in the next edition.
- A company builds a feature on a widely cited finding and cannot reproduce the effect in its own test at ten times the sample size.
- Further reading
- Wikipedia
- Subject
- Science & Epistemology
- Essay
- The ways evidence misleads