Science

Current Picks

One Missing Radiocarbon Batch Pre-Treatment Bent a Peat Core Chronology

By Jonas Eriksen / Jul 9, 2026

A single contaminated pre-treatment batch in a radiocarbon lab shifted a peat core's ages by over 500 years, introducing a spurious climate signal. The error was traced to incomplete rinsing, highlighting the need for batch tracking.

One Unreported Corneal Topography Calibration Bent a Myopia Treatment Trial

By Jonas Eriksen / Jul 9, 2026

A 0.1-diopter calibration drift in corneal topography bent a myopia treatment trial's primary outcome. This article traces the error, its statistical signature, and lessons for device-driven research.

One Unreported Mouse Gut Microbiome Diet Shift Skewed a Obesity Drug Efficacy Trial

By Jonas Eriksen / Jul 9, 2026

A change in mouse feed mid-trial altered gut bacteria, halving the apparent effect of an obesity drug. The case highlights how unreported diet shifts can confound preclinical studies.

One Uncosted Ocean Glider Battery Swap Skewed a Decade of Carbon Flux Estimates

By Karim Osman / Jul 9, 2026

A single battery swap that never happened on a Southern Ocean glider in 2014 propagated through a decade of carbon flux estimates, inflating uptake by ~0.5 Pg C and influencing IPCC reports and carbon-removal startups.

How One Foraminifera Oxygen Isotope Curve Resolved a Plate Tectonics Controversy

By Jonas Eriksen / Jul 9, 2026

How a paleoclimate oxygen isotope curve from foraminifera shells settled a decade-long debate about symmetric versus asymmetric seafloor spreading in the South Atlantic.

One Unreported Foraminifera Dissolution Bias Bent a Paleoclimate Stack

By Renu Shah / Jul 9, 2026

A dissolution bias in foraminifera shells selectively removes thin-shelled species, skewing Mg/Ca and oxygen isotope signals. This article examines how unaccounted dissolution can distort stacked paleoclimate records by 0.3–0.5°C and explores correction methods.

One Uncosted Tape Helium Boil-Off Model Fractured a Survey Spectra Calibration

By Karim Osman / Jul 9, 2026

A missing helium boil-off model in a major survey's calibration budget caused systematic redshift errors, affecting 12–18% of target sources. The story reveals how funding incentives and publication pressure buried a correctable problem.

One Uncosted Ice Core Melt Layer Resequenced a Greenland Temperature Stack

By Renu Shah / Jul 9, 2026

A single uncosted melt layer in a Greenland ice core shifted the alignment of a widely used temperature stack, altering the apparent magnitude of early Holocene warming.

One Undocumented Electrode Polishing Grit Bent a Lithium Dendrite Suppression Claim

By Alice Chen / Jul 9, 2026

A missing polishing grit specification in battery methods sections may have skewed years of lithium dendrite suppression research, highlighting the importance of methodology minutiae.

One Unreported Pollinator Census Transect Width Bent a Mutualism Network Stability Claim

By Alice Chen / Jul 9, 2026

An unreported variation in transect width—5 meters vs. 20 meters—in a landmark pollinator census shifted a mutualism network stability metric by 30%, raising questions about methodological rigor in ecology.

One Unreported Cathode Annealing Ramp Rate Skewed a Battery Lifetime Competition

By Jonas Eriksen / Jul 9, 2026

A 2°C/min difference in cathode annealing ramp rate caused a 7% lifetime gap in a battery competition. Postdoc Inez Kowalski uncovered the hidden variable, forcing industry to rethink test protocols.

One Unreported Reagent pH Buffer Skewed a Fairness Game Replication

By Alice Chen / Jul 9, 2026

How a forgotten pH buffer in a standard protocol silently undermined a decade of fairness game replications—and what it reveals about hidden confounds in behavioral science.

One Uncosted Cryostat Helium Recovery Loop Bent a Superconducting Qubit Coherence Claim

By Renu Shah / Jul 9, 2026

A single uncosted helium leak in a cryostat recovery loop can skew qubit coherence measurements by 30%. Fixing the plumbing could save the field millions in unreproducible claims.

One Unreported Ligand Purity Lot Bent a Palladium Cross-Coupling Rate Model

By Alice Chen / Jul 9, 2026

An unreported impurity in a commercial ligand skewed a decade of palladium cross-coupling kinetics data, revealing how cheap reagents and lax purity reporting can distort catalysis models and waste research resources.

One Unversioned Sparse Solver Default Bent a Seismic Imaging Velocity Model

By Alice Chen / Jul 9, 2026

A default tolerance setting in a popular sparse solver library silently corrupted seismic velocity models for years. The fix: explicitly specifying a parameter. A cautionary tale for computational reproducibility.

One Uncosted Subsea Cable Repair Skewed a Decade of Ocean Temperature Records

By Alice Chen / Jul 9, 2026

A single subsea cable repair that wasn't budgeted caused a 0.02°C bias in global sea-surface temperature records for nearly a decade, exposing how infrastructure costs shape climate data.

One Unreported Loan Interest Rate Bent a Microcredit Poverty Reduction Trial

By Alice Chen / Jul 9, 2026

A misreported interest rate in a landmark microcredit trial halved the apparent poverty reduction effect. The error, unnoticed through peer review, reveals systemic incentives that favor speed over verification in research.

How One Undocumented Grant Overhead Skewed a Behavioral Economics Lab

By Karim Osman / Jul 9, 2026

A US federal grant with a 5% overhead rate, capped at 8% by university policy, created perverse incentives that doubled publication output but produced unreliable results. A replication audit found systematic bias.

One Unreported Polymer Batch Drying Step Inflated a CO₂ Capture Cost Claim

By Alice Chen / Jul 9, 2026

A routine drying step omitted from a published paper inflated the cost of a polymer-based CO₂ capture system by 40%. New analysis shows the real cost is double the original claim, raising questions about how lab-scale breakthroughs are evaluated.

One Uncosted Superconducting Magnet Cool-Down Protocol Fractured a Quantum Error Correction Replication

By Karim Osman / Jul 9, 2026

A $2 million replication of a quantum error correction result failed because the Delft team used a different cool-down protocol than the original lab. The hidden variable? Thermalization time for superconducting magnets.