What would disprove this analysis? (Criterion 1)

Large-scale replication projects find that 80%+ of published studies replicate successfully

What would disprove this analysis? (Criterion 2)

Pre-registration and open science practices become standard without changing replication rates

What would disprove this analysis? (Criterion 3)

Public trust in science remains stable despite awareness of the replication crisis

When should you avoid replication crisis trust erosion?

You're making major decisions based on a single study, no matter how prestigious the journal. You're citing research without checking if it's been replicated

What are alternatives?

Pre-registration: Researchers declare hypotheses and methods before collecting data — prevents p-hacking. Registered reports: Journals accept papers based on methodology before results are known — eliminates publication bias. Open science practices: Open data, open code, and pre-prints enable verification and replication

Catalog

H003

Science & Health

Replication Crisis Trust Erosion

HIGH(84%)

February 2026

4 sources

Context

Science is built on replication — if a finding is real, other researchers should be able to reproduce it. But when researchers tried to replicate landmark studies across psychology, medicine, and economics, the results were devastating. 50-70% of studies failed to replicate. The causes are systemic: publish-or-perish incentives, p-hacking, small sample sizes, and journals that only publish positive results. The replication crisis doesn't just affect academia — it undermines public trust in science at a time when that trust is desperately needed.

Hypothesis

What people believe

“Published peer-reviewed research is reliable and can be trusted as the basis for decisions.”

Actual Chain

→

50-70% of landmark studies fail to replicate(Reproducibility Project: 36% of psychology studies replicated)

└

P-hacking: researchers test multiple hypotheses until one reaches p<0.05

└

Publication bias: journals only publish positive results, file drawer effect

└

Small sample sizes produce unreliable effect sizes that don't hold at scale

→

Medical treatments based on unreplicable research(Billions spent on treatments that don't work as published)

└

Drug trials that showed promise in small studies fail in large trials

└

Nutritional guidelines based on weak evidence flip-flop every decade

└

Patients receive treatments based on studies that wouldn't replicate

→

Public trust in science erodes(Trust in scientists declining, especially among certain demographics)

└

Anti-science movements cite replication failures as evidence science is unreliable

└

Legitimate scientific consensus (climate, vaccines) questioned alongside weak studies

└

Policy-makers cherry-pick studies that support their position — all have 'peer-reviewed' backing

→

Perverse incentives persist despite awareness(Academic career still depends on publication volume, not replication)

└

Replication studies are hard to publish — journals want novel findings

└

Researchers who expose non-replication face professional retaliation

Impact

Metric	Before	After	Delta
Psychology studies that replicate	Assumed most	36% (Reproducibility Project)	64% failure
Cancer biology studies that replicate	Assumed most	11-25% (Amgen, Bayer studies)	75-89% failure
Public trust in scientists	High (pre-2020)	Declining	-15-20%
Research waste (non-replicable studies)	Assumed minimal	$28B/year in US alone	Massive

Navigation

Don't If

•You're making major decisions based on a single study, no matter how prestigious the journal
•You're citing research without checking if it's been replicated

If You Must

1.Look for meta-analyses and systematic reviews rather than individual studies
2.Check if the study has been replicated — and what the replication found
3.Be skeptical of small sample sizes, surprising effect sizes, and single-study claims
4.Prefer pre-registered studies where hypotheses were declared before data collection

Alternatives

Pre-registration — Researchers declare hypotheses and methods before collecting data — prevents p-hacking
Registered reports — Journals accept papers based on methodology before results are known — eliminates publication bias
Open science practices — Open data, open code, and pre-prints enable verification and replication

Falsifiability

This analysis is wrong if:

Large-scale replication projects find that 80%+ of published studies replicate successfully
Pre-registration and open science practices become standard without changing replication rates
Public trust in science remains stable despite awareness of the replication crisis

Sources

1.
Science: Estimating the Reproducibility of Psychological Science
Landmark Reproducibility Project finding only 36% of psychology studies replicated
2.
Nature: 1,500 Scientists Lift the Lid on Reproducibility
Survey showing 70% of researchers have tried and failed to reproduce another scientist's experiments
3.
Amgen: Drug Target Validation Study
Amgen could reproduce only 6 of 53 landmark cancer biology studies (11%)
4.
PLOS Medicine: Research Waste
Estimated $28 billion per year wasted on non-reproducible preclinical research in the US alone

H016 A002 S001