Only Half of Published Findings Reproducible, Massive SCORE Study Shows

**TL;DR** A huge collaboration of 865 scientists found that only about half of nearly 4,000 published studies can be reproduced. The SCORE project, reported in Nature, tested reproducibility, robustness and replicability across social, biomedical and financial research.

## Context The Center for Open Science led the effort, sampling claims from 3,900 papers in 62 journals between 2009 and 2018. Fields included political science, education, finance and health. Researchers and automated tools tried to repeat the original analyses using the same data (reproducibility), alternative methods on the same data (robustness), and full experiment repeats (replicability).

## Key Facts About 50 percent of the claims held up when another team re‑analyzed the original data. Similar success rates appeared for robustness and replicability, showing each metric captures a different slice of reliability. The project involved 865 contributors, making it the largest coordinated check of research credibility to date.

## What It Means The results suggest that a substantial share of published findings may not survive independent verification, prompting journals and funders to consider stricter validation steps. Going forward, watch whether replication attempts become routine in grant reviews and whether new tools emerge to predict study reliability before publication.