Science

Current Picks

One Lab's Uncalibrated Optogenetic Laser Swapped 12 of 20 Place Cell Maps

By Jonas Eriksen / Jun 8, 2026

A routine calibration check caught a laser drift that shifted 12 of 20 place cell maps, raising questions about reproducibility in optogenetics.

One Lab’s Flaking Test‑Tube Liner Swapped 12 of 20 Rodent Anxiety Assays

By Alice Chen / Jun 9, 2026

A degraded polyethylene liner in drinking tubes invalidated 12 of 20 rodent anxiety tests. The case reveals how cheap plastics, blind spots in infrastructure, and publication pressure can undermine years of behavioral neuroscience.

One Unindexed Vial Warped 14 of 20 Antibiotic Susceptibility Curves

By Alice Chen / Jun 8, 2026

A single unindexed vial in a Porto lab caused 14 of 20 antibiotic susceptibility curves to shift, making bacteria appear resistant. The error, traced to a forgotten calibration strain, highlights systemic failures in sample provenance and funding incentives.

One 3‑D Printed Maze’s Floor Texture Swapped 12 of 20 Rodent Learning Curves

By Alice Chen / Jun 9, 2026

A 3D-printed maze with swappable floor textures reveals that surface roughness alters rodent learning curves, challenging assumptions in behavioral experiments and calling for new standards.

One Survey’s Single Wide‑Filter Choice Blinded 14 Galaxy Morphology Studies

By Jonas Eriksen / Jun 9, 2026

A reanalysis revealed that a single wide broadband filter used by a major survey systematically biased 14 galaxy morphology studies, misclassifying mergers as spirals and obscuring tidal features.

One Social Priming Protocol’s 12-Word Instruction Wording Caused a Replication Failure

By Renu Shah / Jun 9, 2026

A subtle 12-word instruction change in a social priming experiment turned a 20% effect into a null result. This article examines how procedural wording choices shape evidence in behavioral science.

One Single‑Sediment‑Core Gamma Ray Scanner Warped 13 Paleoclimate Chronologies

By Alice Chen / Jun 10, 2026

A gamma ray scanner from an Australian lab revealed annual layers in a Lake Chala sediment core, exposing errors in 13 published age models with offsets up to 300 years.

One Unindexed Solvent Batch Re‑ranked 14 of 20 Catalytic Selectivity Curves

By Karim Osman / Jun 9, 2026

A single unrecorded solvent lot shifted 14 of 20 catalytic selectivity curves, exposing how hidden batch effects undermine high-throughput screening and computational workflows. Lessons for reproducibility infrastructure.

One Grant Agency’s Preprint Requirement Made Peer Review a Public Audit Log

By Jonas Eriksen / Jun 10, 2026

A major European grant agency now requires preprints before review, turning peer review into a public audit log. This article examines the policy's effects on review quality, early-career researchers, and what other funders can learn.

One 1970s Plasma Instability Model Now Powers Weather Radar Arrays

By Alice Chen / Jun 9, 2026

A 1978 plasma instability model from fusion research now underpins weather radar algorithms, cutting false alarms by half and improving storm detection by 30%. The 25-year journey reveals how ideas cross scientific fields.

One Grant Agency’s Reusability Rule Made Code Archiving Routine, Not Optional

By Renu Shah / Jun 9, 2026

A 2018 NIH policy made code archiving a grant deliverable, sparking debate. Two years later, compliance rose ~40%, but quality and infrastructure lag.

A Single 3-D Printed Part Warped 13 of 20 Climate Model Pressure Fields

By Alice Chen / Jun 9, 2026

A 3-D printed part deformed under temperature cycles, causing systematic biases in climate model pressure fields. How a tiny manufacturing detail snowballed into a reproducibility crisis.

One Single Low-Oxygen Protocol Warped 13 of 20 Marine Sediment Respiration Rates

By Jonas Eriksen / Jun 8, 2026

A low-oxygen handling protocol skewed respiration measurements in 13 of 20 marine sediment studies, potentially biasing global carbon flux estimates.

One Archive’s Data‑Retrieval Fee Excluded 14 of 20 Global Temperature Reconstructions

By Jonas Eriksen / Jun 9, 2026

A paleoclimate archive’s per-dataset fee blocked 14 of 20 global temperature reconstructions, skewing multi-proxy averages and raising questions about open science and infrastructure funding.

One Funder’s Access-Equipment Rule Moved Genomics Into Every Ecology Lab

By Jonas Eriksen / Jun 9, 2026

A single grant clause from the National Human Genome Research Institute—requiring shared sequencing equipment—unintentionally transformed ecology, making genomic tools routine in labs that once relied on morphology.

One Social Priming Study’s Dishwasher-Detergent Brand Shift Produced a Failed Replication

By Karim Osman / Jun 9, 2026

A landmark 2008 social priming study appeared to show that subtle cleanliness words could shift detergent preferences. When a large-scale replication attempt found no effect, it became a cautionary tale about small samples, context, and the fragility of priming.

One 17‑Second Neural‑Network Reset Warped 14 of 20 Galaxy Morphology Classifiers

By Jonas Eriksen / Jun 9, 2026

A single neural-network reset during a survey pipeline run altered morphology labels for 14 of 20 classifiers, exposing fragility in automated galaxy taxonomy and triggering a re-evaluation of training-data validation.

Two Psych Labs Report Identical Effect Sizes From Opposite Statistical Pipelines

By Renu Shah / Jun 9, 2026

Two labs analyzed the same social-priming data using frequentist and Bayesian pipelines. Both found d ≈ 0.20, but their interpretations diverged sharply. A case study in how statistical choices shape scientific conclusions.

One Ziegler‑Natta Catalyst Batch Contaminated 14 of 20 Polymer Stereoregularity Curves

By Renu Shah / Jun 9, 2026

How a single contaminated batch of Ziegler-Natta catalyst distorted 14 of 20 published stereoregularity curves, leading to a decade of unreliable polymer data and a pivotal reproducibility lesson.

One MRI Scanner’s 3‑Tesla Upgrade Swapped 13 of 20 Resting‑State Maps

By Jonas Eriksen / Jun 9, 2026

When a university imaging center upgraded its MRI from 1.5T to 3T, 13 of 20 resting-state networks vanished. This article examines how hardware changes undermine reproducibility in neuroscience.