Your Lab Notebook Is a Liability: A Working Checklist for Reproducible Science

Ask a room of American researchers about the reproducibility crisis and you will get a lecture on publishing incentives, p-hacking, and the hypercompetitive grant cycle. All legitimate. But ask a lab manager the same question and you'll likely hear something quieter: the freezer nobody can inventory, the notebook that left with a departing postdoc, the analysis script that ran on a laptop now sitting in a desk drawer.

Reproducibility does not fail at the level of science policy. It fails at the level of the bench.

Where Trust Actually Breaks Down

Large replication efforts — including the landmark psychology replication project and subsequent attempts in cancer biology — have found that a substantial share of published findings fail to reproduce at original effect sizes. The causes are debated, but a recurring culprit is prosaic: insufficient methodological detail, reagents that no longer exist, and raw data that was never retained in a usable form.

Consider the typical sequence. An experiment is run over several months by two different students. Conditions drift slightly. The notebook entries live in three formats — a paper log, a shared drive folder, and someone's personal cloud account. When a reviewer requests the raw data eighteen months later, reconstruction becomes archaeology.

Funding timelines make this worse. A typical NIH R01 runs in five-year cycles, and the pressure to publish within that window rewards speed over documentation. Data management plans, now required by most US federal funders, are too often written for the grant proposal rather than for the lab bench.

The KSMV Lab Checklist

None of the following requires new funding, new equipment, or institutional buy-in beyond a department meeting. Each item can be adopted this week.

1. One notebook, one rule. Every experiment, by every member, is recorded in a single designated system — electronic or paper — with dated, signed (or timestamped) entries. No side notebooks. No exceptions.

2. Version-control everything that isn't data. Analysis scripts, protocols, and figure-generating code belong in a versioned repository (Git or an institutional equivalent) from day one, not after publication.

3. The reagent registry. Maintain a shared inventory of antibodies, cell lines, and critical reagents, including lot numbers and source. A result that depends on a lot-specific antibody isn't reproducible without that number.

4. Raw data is sacred. Processed files can be regenerated; raw instrument output cannot. Archive it in an immutable, backed-up location with a documented folder structure.

5. The departure protocol. No one leaves the lab — graduation, job change, grant end — without a documented handoff: data locations, credentials transferred, protocols annotated. Make it a checklist signed by both parties.

6. Negative results get filed, not buried. A shared internal log of failed attempts saves future lab members months of duplicate work and counters the file-drawer effect.

7. Pilot the reproduction. Before submitting a major paper, have a lab member who didn't run the study re-execute the analysis from the archived materials alone. If they can't, the paper isn't ready.

An Argument for Institutional Memory

Some will object that this checklist adds overhead to labs already stretched thin. The counterargument is that it is not overhead — it is insurance. The cost of a reproducibility failure is not borne by the field in the abstract; it is borne by the lab that cannot rebuild its own experiment, cannot answer a reviewer's query, or cannot support a grant renewal with its own records.

There is also a defensive dimension. Federal funders increasingly scrutinize data management practices, and journals now routinely request raw data and code at submission. Labs with functioning internal systems meet these demands in days. Labs without them scramble for weeks.

The reproducibility crisis will not be solved by a manifesto. It will be solved — or not — one notebook, one freezer inventory, and one departure checklist at a time. KSMV encourages member laboratories to adopt the checklist above, adapt it to their discipline, and report back on what works. Trust in science is built the same way a good experiment is: deliberately, locally, and with records to prove it.