The North Atlantic is absorbing heat faster than every model predicted.
Eleven years of autonomous float data, pooled across 4,200 instruments, show the upper ocean warming 23% faster than the ensemble mean. The result holds across three independent calibration methods, and the full dataset is public as of this morning.
Dr. Amara Nwosu, Dr. Kenji Hoshino and the Halden Ocean Systems Lab Published 18 November 2026 · Peer reviewed · 14 min read
FIG. 1Upper-ocean (0–700 m) heat content anomaly, North Atlantic basin, relative to the 2005–2014 baseline. Shaded band shows the multi-model ensemble range.
Every Halden paper is open access from the day of publication, ships with its raw data and analysis code, and is pre-registered wherever the design allows. We report effect sizes and confidence intervals alongside p-values, always.
Pooling data from 48,300 students across 19 countries, we find a moderate, durable benefit of spaced retrieval over massed study. The effect is robust to publication-bias corrections and strongest in mathematics and the physical sciences. We found no reliable advantage in studies shorter than two weeks.
Hedges' g = 0.4195% CI [0.34, 0.48]I² = 62%Open dataPre-registered
Read ResearchOkafor, Lindgren & Vasquez · 11 Nov 2026
At 2.1 Å resolution, the enzyme's active site shows a flexible "lid" that clamps the polymer chain in place. A three-residue substitution guided by the structure raised depolymerisation rates fourfold in bench trials, offering empirical evidence for a design principle that had previously only been simulated.
Resolution 2.1 Åkcat ×4.2n = 6 replicatesOpen data
An influential 2019 study linked tree cover to lower asthma admissions. Our replication, with nine times the sample, confirms the direction of the effect but finds it roughly half the original size once traffic density is controlled. Null and partial results like this one matter as much as the headlines.
Twelve research groups, 214 scientists, one shared core facility. Search by topic, method or principal investigator, or filter by discipline to find collaborators, open positions and instruments you can book.
Halden awards roughly $14 million a year to researchers inside and outside the institute. Applications are reviewed double-blind, scored on rigour rather than pedigree, and every applicant receives written reviewer feedback.
Programme
Award
Deadline
Apply
Open Replication FundDirect and conceptual replications of high-impact findings in any empirical field. Registered-report format required.
$40k–$150k
15 Dec 2026
Early-Career Discovery FellowshipThree years of salary and research costs for scientists within six years of their doctorate. Host institution of your choice.
$285k / 3 yrs
22 Jan 2027
Instrument Access AwardBeam time on our cryo-EM, mass spectrometry and high-performance computing cores for external groups, with staff support.
Up to 400 hrs
10 Feb 2027
STEM Classroom PartnershipSeed funding for teachers and scientists co-designing inquiry-based curricula for under-resourced secondary schools.
$8k–$25k
31 Mar 2027
How review works. Each proposal is read by three reviewers who see the science but not the names. Panels score the question, the design and the analysis plan, including a justified sample size and a stated threshold for statistical significance, before any track record is revealed.
What we fund that others don't. Replications, null results, methods papers, data curation and software maintenance. If the work makes the scientific record more reliable, it is in scope, and last cycle 38% of our awards went to first-time principal investigators.
Free evening talks in the Ørsted Auditorium and streamed live with captions. Doors at 6:30 pm, lecture at 7:00, questions for as long as the speaker will stand it.
03Dec 2026 Thursday · 7 pm
What the deep ocean remembers
Dr. Amara Nwosu · Ocean Systems Lab
How a fleet of drifting robots rewrote the heat budget of the Atlantic, and what the next decade of data might hold.
14Jan 2027 Thursday · 7 pm
Why most findings shrink
Prof. Tomás Vasquez · Metascience Unit
A plain-spoken tour of publication bias, the winner's curse and how a careful meta-analysis separates signal from noise.
11Feb 2027 Thursday · 7 pm
Enzymes that eat plastic
Dr. Rina Chatterjee · Structural Biology
From a compost heap to a 2.1-ångström structure: the detective story behind a promising route to true recycling.
Everything here is free, openly licensed and tested in real classrooms by our education team. Lesson plans align to NGSS and include printable handouts, datasets small enough for a spreadsheet, and teacher notes.
GRADES 9–12 · 6 LESSONS
Statistics with Real Data
Students test a hypothesis using our ocean-float dataset and learn what a p-value does and does not tell you.
"Halden funded our replication when four other agencies told us it wasn't 'novel' enough. The original effect didn't hold, and the field is better for knowing."
Dr. Leila HaddadEpidemiologist, Replication Fund awardee
"My Year 11 students analysed real float data and argued about statistical significance for forty minutes. I have never seen them so invested."
Marcus OyelaranPhysics teacher, Classroom Partnership
"The reviewer feedback on my unsuccessful fellowship application was more useful than most of my doctoral training. I reapplied and got it."
Dr. Ingrid SolbergEarly-Career Discovery Fellow, 2025
§ 7 — About the institute
Science that can be checked
The Halden Institute was founded in 2014 by a small group of researchers who were tired of results that could not be reproduced, including some of their own. They set out to build a place where the incentives pointed the other way: where sharing data was the default, where a careful null result counted as a contribution, and where the public who ultimately pay for science could read all of it.
Twelve years on, the institute houses twelve laboratories and a shared instrument core on a single campus in Cambridge, Massachusetts. We are funded by an independent endowment and philanthropic gifts, accept no industry-directed research contracts, and publish our full accounts and conflict-of-interest register every year.
Open by defaultPapers, data, code and protocols are released together under permissive licences.
Pre-register, then lookHypotheses and analysis plans are time-stamped before data collection.
Report everythingEffect sizes, intervals and failed experiments, not just what cleared a threshold.
Teach what we learnEvery lab contributes to public lectures and classroom materials.
§ 8 — Questions
Frequently asked
Who is eligible to apply for Halden grants?
Researchers at any career stage, at any accredited institution worldwide, as well as independent scientists with an institutional fiscal sponsor. You do not need a prior relationship with Halden, and we actively encourage first-time applicants. Programme-specific criteria are listed in each call.
Do I have to publish open access if you fund my work?
Yes. Funded outputs must be immediately open access under a CC-BY licence, with data and code deposited in a recognised repository. We cover reasonable article-processing charges and provide data-curation support at no cost to the grant.
Can external groups use the instrument core?
They can. Around 30% of time on our cryo-electron microscopes, mass spectrometers and computing cluster is reserved for outside researchers through the Instrument Access Award. Staff scientists help with sample preparation, acquisition and first-pass analysis.
Are public lectures really free? Do I need a science background?
Lectures are free, open to everyone aged twelve and up, and pitched for a curious general audience. Seats in the auditorium are limited, so please reserve one; the livestream has no cap and recordings are posted within a week.
Contact
Talk to a human scientist.
Questions about a paper, a dataset, a grant call or a school visit all land with a real person on the relevant team. We aim to reply within two working days.
Visit
Halden Institute for Open Science 77 Kelvin Way, Cambridge, MA 02142