Science

Current Picks

One Unreported Wave Tank Salinity Gradient Bent a Tsunami Run‑Up Flume Replication

By Alice Chen / Jul 10, 2026

A 12% discrepancy in tsunami run-up heights was traced to an overlooked salinity gradient in wave tanks. The finding forces a reexamination of decades of flume data and a change in experimental protocol.

One Unpaid Experimenter Overtime Hour Bent a Primate Social Learning Paper

By Renu Shah / Jul 10, 2026

A single miscoded hour of unpaid graduate labor inflated a macaque social learning finding by 12%, revealing how funding pressures and publication incentives amplify small errors in primate research.

One Unreported Grant Reviewer Conflict Bent a Behavioral Economics Replication Consortium

By Renu Shah / Jul 10, 2026

An undisclosed conflict of interest involving a grant reviewer who later joined the consortium likely biased study selection and analysis, reducing the apparent replication rate in a high-profile behavioral economics replication project.

One Unreported Participant Familiarity Protocol Bent a Trust Game Replication

By Karim Osman / Jul 10, 2026

A large-scale replication of a classic trust game failed until reanalysis revealed that participant familiarity, not trust, drove the null result. One procedural oversight changed everything.

One Uncosted Mirror Age Gradient Bent a Dark Energy Survey Photometric Redshift

By Renu Shah / Jul 10, 2026

How a subtle reflectivity gradient across the CTIO 4-meter mirror introduced a 0.03 bias in photometric redshifts, and why the Dark Energy Survey missed it for years.

One Unreported Beam Polarization Offset Bent a Proton Spin Structure Measurement

By Karim Osman / Jul 10, 2026

A 0.4% offset in beam polarization at Jefferson Lab went unnoticed for 18 months, shifting a proton spin measurement by 5 sigma. The corrected result raises the quark spin contribution from 28% to 31% and matches lattice QCD predictions within 1.2 sigma.

One Unreleased Analysis Script Fork Broke a Computational Reproducibility Certification

By Karim Osman / Jul 10, 2026

How a single forked analysis script—altering a random seed and data filter—broke a computational reproducibility certification, exposing gaps in code archiving and audit.

One Unreported Ant Colony Foraging Path Width Bent a Collective Behavior Simulation

By Karim Osman / Jul 10, 2026

A 35% wider ant trail in one field study altered a collective behavior simulation. How hidden measurement artifacts shape computational biology.

One Unreported Stirring Impeller Geometry Bent a Polymerization Rate Constant Replication

By Jonas Eriksen / Jul 10, 2026

A replication study found that impeller blade shape alters the polymerization rate constant by up to 23%, revealing hidden bias in decades of kinetic data.

One Unspecified Stirrer Blade Depth Bent a Polymerization Kinetics Model

By Jonas Eriksen / Jul 10, 2026

A 0.7 mm stirrer blade depth error skewed monomer conversion curves for eight years, leading to retracted papers and a new IUPAC reporting standard for batch reactors.

One Left‑Hemisphere Language‑Selective Voxel Set Predicted Two Completely Separate Reading Tasks

By Renu Shah / Jul 10, 2026

A single set of left-hemisphere language-selective voxels predicts both word recognition and sentence comprehension, challenging modular accounts of reading.

One Unreported Code Dependency Version Bent a Reproducible Neuroimaging Pipeline

By Karim Osman / Jul 10, 2026

A single Python library version change broke a reproducible neuroimaging pipeline, dropping effect sizes from 0.8 to 0.3. How floating-point order in NumPy 1.24 altered cluster detection and what it means for computational reproducibility.

One Unsubmitted Ethics Amendment Refiled a Social Conformity Replication

By Renu Shah / Jul 10, 2026

A $2.6 million replication of the Asch conformity experiment nearly collapsed after a missing ethics checkbox triggered a 3-month review, revealing hidden costs in research infrastructure.

One Unreported Mouse Strain Substrain Shift Bent a Stress Hormone Circuit Paper

By Alice Chen / Jul 10, 2026

A hidden shift from C57BL/6J to C57BL/6N substrain caused conflicting results in stress hormone studies, leading to a retraction and new reporting standards.

A Single Grant Rescore Bent Seven Primate Circuit Replications

By Renu Shah / Jul 10, 2026

How a single peer-review score drop triggered the dismantling of seven primate labs, halting replication of key circuit studies and widening the gap between animal models and human inference.

A Single Unrecorded Mouse Cage Light Cycle Bent Four Fear Conditioning Replications

By Renu Shah / Jul 10, 2026

An unrecorded 12-hour light cycle shift in a mouse facility halved effect sizes across four fear conditioning replications. Circadian disruption altered basolateral amygdala plasticity, raising questions about hidden cage effects in published studies.

One Unreported Survey Question Order Shift Bent a Charitable Giving Field Experiment

By Alice Chen / Jul 10, 2026

A reanalysis of a famous charitable giving field experiment found that an unreported shift in survey question order explained most of the reported 48% donation boost. The case illustrates how subtle procedural details can produce large spurious effects in behavioral science.

A Marmoset Vocalization Protocol Moved From Primate Ethology Into Human fMRI Speech Studies

By Renu Shah / Jul 10, 2026

A marmoset phee call protocol moved from primate ethology into human fMRI speech studies, revealing conserved auditory processing across species.

One Unreported Peat Core Radiocarbon Offset Bent a Holocene Methane Budget

By Jonas Eriksen / Jul 10, 2026

A single peat core from Sweden with a 1,200-year radiocarbon offset has skewed global Holocene methane budgets by up to 15 teragrams per century. New research shows how dating errors propagate through carbon models.

One Unreported Flat Calibration Card Luminance Bent a Dark Energy Survey Supernova Flux

By Alice Chen / Jul 10, 2026

A 2% luminance gradient in the calibration card used for Dark Energy Survey flat-fields introduced a systematic error in supernova fluxes, inflating dark energy parameter uncertainty by ~15%.