Hypothesis

Replications fail because the replicating researchers lack the expertise and diligence to recreate the original conditions, so replicators with higher h-indices should succeed more often.

What failed

Across 100 scientists who attempted to replicate five studies, replication success and investigators' h-indices both varied widely, but there was no relationship between them.

Broken assumption

Failed replications reflect the replicators' lack of skill rather than problems with the original findings.

Why

The abstract does not offer an alternative explanation; it concludes only that low replicator expertise does not account for the failures.

From the published paper: Protzko et al. PeerJ, 2020. ↗CC BY

Hypothesis

Published computational methods for linking enhancers to their target genes would clearly beat a simple genomic-distance baseline when tested on a curated benchmark.

What failed

On the new BENGI benchmark, which combines the cCRE registry with experimentally derived interactions, the best method (TargetFinder) was only modestly better than the distance baseline within a cell type and often no better than it across cell types.

Broken assumption

Supervised enhancer-gene predictors learn regulatory signal that generalises beyond simple proximity.

Why

The abstract gives no mechanism; it concludes current methods need improvement and offers BENGI as a framework for developing them.

From the published paper: Moore et al. Genome biology, 2020. ↗CC BY

Hypothesis

Orienting native English listeners positively or negatively would shift how they rate non-native English speech, as Reid et al. (2019) found.

What failed

In a preregistered online conceptual replication (n = 189), listeners did not change their ratings after the bias orientations, and performed similarly across all five speech measures and across age and race groups.

Broken assumption

'Native speaker' is a stable enough category for bias-priming effects on speech ratings to transfer to a broader online sample.

Why

The authors attribute the difference to method (in-person versus online) and/or participants: about one-third of 'native' participants who passed standard screening reported other languages and residence in 12 English-speaking countries, which they argue makes 'nativeness' a culture-specific and limiting category.

From the published paper: Brown et al. Applied Psycholinguistics, 2023. ↗CC BY

Hypothesis

Ataluren would slow lung disease in people with nonsense-mutation CF; a reanalysis with more sensitive CT scoring would reveal a treatment effect on structural lung disease that the original trial's FEV1 endpoint and Brody-II scores missed.

What failed

The 48-week phase 3 trial in 238 patients had already missed its FEV1 primary endpoint. Re-scoring 391 CT scans from 210 patients with CF-CT and PRAGMA-CF showed PRAGMA-CF disease and mucus plugging progressed over 48 weeks, but progression did not differ between ataluren and placebo, regardless of tobramycin use.

Broken assumption

Ataluren's read-through of nonsense mutations restores enough CFTR function to slow structural lung damage.

Why

The abstract does not explain the lack of effect; it concludes PRAGMA-CF is sensitive enough to serve as a trial outcome.

From the published paper: Tiddens et al. PLoS ONE, 2020. ↗CC BY

Hypothesis

A probe-detection attention bias modification program that trains attention away from threatening faces, which had shown large effects in clinic-based trials, would work as well when delivered over the internet at home.

What failed

In a double-blind placebo-controlled trial of 79 participants analysed by intention to treat, both groups improved over time on social anxiety, general anxiety, depression and quality of life, but there were no time-by-group interactions and the between-group effect size was close to zero at post-test and at 4-month follow-up.

Broken assumption

Attention bias modification that works in a supervised clinic setting transfers unchanged to unsupervised home use over the internet.

Why

The authors do not pin down a cause; they conclude the program may need to be changed before it can be disseminated online.

From the published paper: Carlbring et al. BMC Psychiatry, 2012. ↗CC BY

Hypothesis

A single session of anodal tDCS over the left dlPFC would improve working-memory performance in older adults.

What failed

In a single-blind crossover with 30 older adults comparing 1 mA, 2 mA and sham stimulation, N-back accuracy and response times did not change during or after stimulation at either current strength.

Broken assumption

One session of prefrontal anodal tDCS is enough to boost working memory in older adults.

Why

The abstract gives no mechanism; it notes that optimal stimulation parameters for cognition are poorly understood and concludes one session is unlikely to help working memory in old age.

From the published paper: Nilsson et al. Frontiers in Aging Neuroscience, 2015. ↗CC BY

Hypothesis

Implicit Puritanism: Americans, because of their Puritan heritage, intuitively moralize work (and link it to sex morality) in a way other cultures do not.

What failed

In a preregistered multi-culture replication, several of the original effects failed to replicate in every sample and were judged likely false positives, while the effects that did replicate also appeared in the comparison cultures, contradicting the claim of a distinctly American work morality. Proposed regional, religious and education moderators found no support.

Broken assumption

Intuitive moralization of work is a distinctively American, Puritan-derived trait.

Why

The authors conclude that work is intuitively moralized across cultures rather than uniquely in the United States.

From the published paper: Tierney et al. Journal of Experimental Social Psychology, 2020. ↗CC BY

Hypothesis

Extracellular vesicles shed by brain endothelial cells could be isolated from peripheral blood as a 'liquid biopsy' of the blood-brain barrier, giving a biomarker for psychosis.

What failed

Lacking a single specific brain-endothelial protein, the team immunolabelled three proteins and confirmed by proteomics and flow cytometry that vesicles with a CD45−/CD31+/GLUT1+ signature exist — but generating enough material to confirm the proteomic analysis demanded too much time and resource, and the work was stopped.

Broken assumption

Brain-endothelial extracellular vesicles are abundant and specifically identifiable enough in blood to serve as a blood-brain-barrier biopsy.

Why

The authors conclude the approach was not feasible because vesicles with the target signature are too scarce, and because the relative abundance of putative blood-derived versus brain-endothelial vesicles contradicted published data, raising doubts about the markers' specificity.

From the published paper: Tunset et al. Journal of Trial and Error, 2025. ↗CC BY

Hypothesis

People would overweight good news relative to bad news when updating beliefs about purely financial outcomes, as found for ego-relevant beliefs.

What failed

In a laboratory experiment on financial beliefs, there was no evidence of asymmetric updating: on average, updating was close to the Bayesian benchmark, and no sizeable subgroup of asymmetric updaters was found despite substantial individual heterogeneity.

Broken assumption

The good-news/bad-news updating asymmetry seen for ego-relevant beliefs extends to purely financial beliefs.

Why

The paper reports that the asymmetry did not appear in this financial domain; it does not claim a mechanism for why it is absent.

From the published paper: Barron et al. Experimental Economics, 2020. ↗CC BY