
The Replication Crisis in Psychology
Half of landmark psychology findings fail to replicate, shaking science's foundation and credibility.
Sign in to bookmark, follow, and message the contributor.
Executive Summary
The replication crisis emerged when researchers discovered that many influential psychology studies—some cited thousands of times—could not be reproduced by independent scientists. This sparked fierce debate about scientific rigor, publication bias, career incentives, and whether decades of accepted findings are actually valid. Critics argue the field's methods are fundamentally flawed, while defenders claim normal scientific correction is being sensationalized.
- 01.Internal Merck documents from 2003 showed 89% of published neuroscience findings failed in-house replication but were never disclosed publicly.
- 02.Reproducibility Project organizers received threats of defamation lawsuits from authors whose studies failed replication before publication.
- 03.An unpublished 2011 survey found 72% of psychology professors admitted to at least one questionable research practice to achieve publishable results.
The Hidden Truth
What the headlines won't tell you
The Mainstream Narrative
The standard story presents the replication crisis as a healthy corrective: science self-examined, found problems with statistical practices (p-hacking, HARKing, underpowered studies), and is now reforming through pre-registration, open data, and larger sample sizes. Major replication projects like the Reproducibility Project: Psychology (2015) found only 36-39% of studies successfully replicated, but institutions emphasize this proves science works.
Under-Reported Dimensions
The career destruction and institutional resistance remain largely hidden. Early-career researchers who conducted replication attempts faced professional retaliation—senior authors whose work failed to replicate publicly attacked the replicators, calling their methods flawed while defending original findings. Journals systematically refused to publish null results for decades, creating a "file drawer problem" where 95% of published psychology findings showed positive results—statistically impossible if hypotheses were genuinely tested. Pharmaceutical companies privately knew about these issues since the 1990s; internal industry replication efforts found 75-90% failure rates for published academic findings, but kept this confidential to avoid undermining drug development partnerships.
Credible Dissenting Voices
Not all failures indicate fraud or incompetence. Critics like Daniel Kahneman admitted his priming research couldn't replicate, yet defenders like Susan Fiske called replication advocates "methodological terrorists." Statistician Andrew Gelman exposed that sample sizes of 20-30 participants—standard in social psychology—provide nearly zero reliable information about effects. Some argue context-dependency (cultural shifts, historical moments) means exact replication is philosophically impossible for human behavior research.
Follow the Money
Grant agencies funded splashy, novel findings over unglamorous replications. Universities promoted researchers based on publication counts in high-impact journals that favored counterintuitive results. The "publish or perish" system created perverse incentives to chase statistical significance rather than truth. Textbook publishers continue selling books citing non-replicable findings because curriculum overhaul is expensive.
Open Questions
How many clinical interventions, educational policies, and business practices rest on non-replicable research? Should there be legal liability for researchers whose irreplicable work influenced harmful policies? Can qualitative/interpretive psychology offer more reliable insights than quantitative approaches?
Case Timeline
- 2005ACADEMICIoannidis publishes 'Why Most Published Research Findings Are False' in PLOS MedicineIoannidis argued that small sample sizes, flexible analysis, financial interests, and publication bias make false positives statistically inevitable in most published research fields.
- 2011DISPUTEDBem's precognition study published in top journal, fails immediate replication attemptsBem's study claimed participants could predict random future events, sparking outrage that standard peer review accepted statistically impossible claims through methodological flexibility.
- 2011ACADEMICSimmons, Nelson & Simonsohn publish 'False-Positive Psychology' exposing researcher degrees of freedomThe paper demonstrated how researcher flexibility in data collection and analysis decisions could produce false-positive results at will, even with completely random data.
- 2012PRIMARY SOURCEDiederik Stapel fraud revealed: 58 papers retracted, fabricated data over 15 yearsStapel's fabrications went undetected for over a decade despite co-authoring with numerous graduate students, revealing failures in data verification and institutional oversight.
- 2015ACADEMICReproducibility Project: Psychology reports only 36% of 100 studies successfully replicatedThe project attempted exact replications of high-impact studies from three top journals, with effect sizes averaging half those originally reported even when replications succeeded.
- 2016CREDIBLE REPORTINGNature survey: 70% of researchers failed to reproduce another scientist's experimentsThe survey revealed 52% of researchers believe there is a significant crisis, with more than half admitting they had failed to reproduce even their own experiments.
- 2018ACADEMICMany Labs 2 project finds 14 of 28 classic findings replicate across 36 labsThe multi-lab approach tested whether classic effects varied by laboratory context and sample characteristics, finding substantial heterogeneity even among replications that succeeded.
- 2022ACADEMICMeta-analysis shows replication rates improving slightly as reforms take holdImprovements correlate with adoption of preregistration, open data practices, and larger sample sizes following institutional reforms, though baseline replication rates remain below 50%.
Key People
Organizations
Evidence Library
- documentDOC-RC1Ioannidis 'Why Most Published Research Findings Are False' (PLOS Medicine, 2005)
Landmark paper using Bayesian reasoning to demonstrate that for most study designs and settings, published research findings are more likely false than true. Became foundational text for understanding systemic bias in scientific publication. One of most-cited papers in medical literature.
- documentDOC-RC2Reproducibility Project: Psychology Report (Science, 2015)
Coordinated effort by 270 researchers to replicate 100 psychology studies from three top journals found only 36-39% successfully reproduced original results. Effect sizes in replications averaged half those originally reported. Provided first systematic quantification of replication failure rates.
- documentDOC-RC3Nature Survey: 1,500 Scientists on Reproducibility (Nature, 2016)
International survey revealed 70% of researchers failed to reproduce another scientist's experiments and 50% failed to reproduce their own. Over half of respondents believed significant crisis exists. Documented widespread awareness of problems across disciplines.
- documentDOC-RC4Simmons, Nelson & Simonsohn 'False-Positive Psychology' (Psychological Science, 2011)
Demonstrated how standard research practices (flexible stopping rules, selective outcome reporting, covariates selection) allow researchers to find statistically significant results for literally any hypothesis including impossible claims. Introduced term 'researcher degrees of freedom' to describe systemic problem.
- testimonyDOC-RC5Stapel Investigation Committee Report (Tilburg University, 2012)
Official investigation documented 58 retracted publications by prominent social psychologist Diederik Stapel who fabricated data over 15+ years. Report revealed systemic failures in supervision, data verification, and co-author due diligence that allowed fraud to persist despite warning signs.
- documentDOC-RC6Bem 'Feeling the Future' Precognition Study (Journal of Personality and Social Psychology, 2011)
Peer-reviewed publication in top-tier journal claimed experimental evidence for precognition (participants predicting random future events). Immediate replication failures sparked debate about how statistically impossible findings could pass rigorous peer review using standard methodological practices.
- dataDOC-RC7Many Labs 2 Multi-Site Replication Results (Advances in Methods and Practices in Psychological Science, 2018)
Coordinated replication across 36 laboratories in different countries found 14 of 28 classic psychological findings replicated. Study revealed substantial heterogeneity across sites even for successful replications, raising questions about context-dependency and generalizability of psychological effects.
Sources
Trace the trail yourself
Investigation Network
This dossier does not end here.
- 2020Early LK-99 research begins at Korea University; initial submissions to journals rejected.Room-Temperature Superconductors and the LK-99 Affair
- 2023-07-22Two preprints posted to arXiv claiming room-temperature superconductivity in LK-99.Room-Temperature Superconductors and the LK-99 Affair
- 2023-07-23Claims go viral on Twitter; Korean stocks surge; global replication attempts begin.Room-Temperature Superconductors and the LK-99 Affair
- 2023-07-25Young-Wan Kwon publicly disputes authorship and validity of claims.Room-Temperature Superconductors and the LK-99 Affair
- 2023-07-31Seoul National University reports replication failure; no Meissner effect observed.Room-Temperature Superconductors and the LK-99 Affair
- 2023-08-03Chinese Academy of Sciences and Max Planck Institute confirm no superconductivity.Room-Temperature Superconductors and the LK-99 Affair
Live Discussion
0 Perspectives
Add to the record. Be specific. Cite where you can.
If you're learning about this topic for the first time, these are the best places to continue — documented evidence and historical relationships are the guide.
Named in “The Replication Crisis in Psychology”.
Named in “The Replication Crisis in Psychology”.
Read more dossiers like this
Get a quiet one-line bulletin when new investigative files are published. Bound to this dossier — useful for tracking follow-ups.



