Effect sizes shrink on replication

In a coordinated attempt to replicate 100 psychology studies, only about 36% reproduced statistically significant results and the average effect size was roughly half the original. The project used standardized protocols and open materials to reduce garden‑of‑forking‑paths and showed that headline findings are often inflated. — It warns media and policymakers to demand replication and preregistration before building policy or public narratives on single, striking studies.