Is most published research wrong? What the replication debate does and does not show

In the early 2010s, researchers in several fields reported that many published findings did not hold up when other teams tried to repeat them. This became known as the replication crisis. The video "Is Most Published Research Wrong?" from the Veritasium channel is a popular introduction to the argument.

What the debate showed: some findings, particularly small studies with flexible analysis, are less reliable than their publication suggested. Incentives that reward novelty and positive results make that worse, and publication bias means that negative results often go unseen.

What it did not show: that science as a whole is broken, or that every field is equally affected. Replication rates differ by field, by method and by how the question is framed, and some of the most robust results in science have been confirmed many times.

Practical consequences: preregistration, sharing data and code, larger samples and replication studies are all responses.

How to read a single new finding: treat it as promising, not settled, until it has been repeated by independent teams.