The Open Science Collaboration replicated each of the 100 psychology studies only once using roughly the original sample sizes, and when researchers applied that same low-powered design to a data set where each study had been replicated with more than 30 times the original number of participants, the resulting replication rate was even lower than the OSC's.
Notes on verification
Core mechanism and key figures (34% vs 47%) are well corroborated by the primary Gilbert et al. 2016 Science comment and multiple independent summaries. The 'more than 30 times' sample size multiplier is a reasonable approximation consistent with Many Labs 1 data but not a verbatim figure from primary sources. The broader interpretive debate (whether this undermines OSC's conclusions) remains contested among methodologists, though that dispute doesn't affect the specific empirical claim here.
Sources
- Study that undercut psych research got it wrong (news.harvard.edu)
- https://www.science.org/doi/10.1126/science.aad7243 (science.org)
- https://thehardestscience.com/2016/03/03/evaluating-a-new-critique-of-the-reproducibility-project/ (thehardestscience.com)
- https://www.bps.org.uk/psychologist/replication-glass-half-full-half-empty-or-irrelevant (bps.org.uk)