This page describes how we publish: what a claim has to resolve to before it appears, what a version means, what happens when we get something wrong, and what we do not do. It is a disclosure, not a mission statement. Every rule below is enforced by the software that runs this site, not by good intentions.
Every quantified claim resolves to a source record
A figure in one of our papers is not free text. It is a typed field with a value, a unit, a stated basis, and a link to a source record describing how the number was arrived at and over what sample. A paper cannot be published while any quantified claim lacks a resolving source — this is a hard gate in the publishing system, not a checklist item.
If a source record is later removed, the exhibit that depended on it is suppressed on the published page and replaced with a visible notice. We would rather show you that something was withdrawn than quietly serve you a shorter paper.
Specimen figures are labelled as such
Where an exhibit is illustrative rather than measured — a worked example, a notional roadmap, an author-constructed matrix — it carries a specimen flag on its face and in its source record. A specimen figure is not a finding and must not be cited as one.
Published versions are immutable
When a version is published it is frozen and hashed. No editing path in this system can alter it afterwards. Revising a paper does not rewrite it; it opens a new version alongside the old one, and both remain permanently addressable.
This is what makes a citation to our work durable. A reference to a specific version resolves to the text that was actually read, on the day it was read, for as long as this site exists.
Corrections are visible, not silent
A correction opens a new version and publishes a record saying what was wrong, which version carried the error, and which version resolves it. That record appears in a public register, on the paper itself, and in the version history. We do not edit published text in place, and we do not remove errors from the record.
A superseded version stays reachable and says plainly that it has been superseded, with a link to the current text. It is removed from search indexes but never from the web.
What our accession numbers mean
Every paper receives a permanent identifier of the form CRX-WP-0000 when it is first published, and keeps it for life. Versions are addressed beneath it. A citation should name both — the work and the version — because only the pair identifies a specific text.
What we do not claim
- Our samples are usually small and rarely random. Where a paper reviews six programmes, it says so, and it does not generalise beyond them.
- We do not run blind peer review. Papers are reviewed internally and by named external readers where we have them; the review round is recorded, but it is not equivalent to academic peer review and we do not describe it as such.
- We are not neutral about method. Where we think a programme is measuring the wrong thing, we say so, and we show the reasoning rather than asking to be trusted.
- We do not publish anything we cannot source. A claim we believe but cannot trace does not appear, in any form.
Corrections and challenges
If you believe something we have published is wrong, tell us. Point at the accession number, the version and the section. We will look, and if you are right we will issue a correction that says so in public.