Preprint servers, repositories, indexes and aggregators do four different jobs, and citing the wrong one is why reference lists rot.
Finding a research paper and citing a copy that will still resolve in ten years are separate problems solved by different sites. Preprint servers, repositories, indexes and aggregators each do one part of the job, and treating them as interchangeable is the main cause of broken reference lists.
This entry separates the sources that hold papers from the sources that find them, with a note on what each covers and where it will mislead.
Finding a paper and finding a stable copy of a paper are different tasks, and the sites that do them are different sites. A preprint server hosts the author's own version before or alongside peer review. A repository hosts a deposited copy under a persistent identifier. An index tells you a paper exists and where it is, without necessarily holding it. An aggregator gathers records from many repositories into one search.
Confusing the four produces two recurring problems. A citation to an index rather than to the paper breaks when the index reorganises. And a link to a publisher landing page behind a paywall is not a source most readers can check, which is why an open deposited copy, where one exists, is the more useful thing to cite alongside it.
Version is the other trap. Preprint, accepted manuscript and published version can differ substantially, and page numbers rarely survive between them. Any citation that will be checked by somebody else should say which version it refers to, and a digital object identifier resolving to the published record is worth carrying even when the accessible copy is elsewhere.
Repositories and preprint servers that hold papers themselves, under identifiers designed to keep resolving.
The preprint server for physics, mathematics, computer science, quantitative biology, statistics and economics, running since 1991. Every submission keeps a permanent identifier and every revision remains retrievable, which makes it unusually good for citing a specific version rather than a moving target.
Last checked 2026-08-26 · reachable
The United States National Library of Medicine's full-text archive for biomedical and life sciences literature. Distinct from PubMed itself, which indexes citations; PMC holds the articles. Deposits mandated by public funders land here, which is why it is often the only open copy of a paywalled paper.
Last checked 2026-08-26 · reachable
A general-purpose repository operated by CERN, accepting papers, datasets, software and supplementary material across every field. Every deposit receives a digital object identifier, which makes it the standard destination for material that has no disciplinary repository of its own.
Last checked 2026-08-26 · reachable
Indexes and aggregators. These find papers rather than hold them, and the distinction matters when deciding what to cite.
A curated index of open-access journals that meet stated editorial and transparency criteria. Its practical value is negative screening: a journal absent from DOAJ is worth a second look, and its inclusion criteria are published rather than implied.
Last checked 2026-08-26 · reachable
Aggregates open-access records from thousands of repositories into one search, with full text where the source repository allows it. The right starting point when you know a paper exists and do not know which institution deposited it.
Last checked 2026-08-26 · reachable
A search index with extracted citation contexts, so you can see how a paper is cited rather than only how often. More useful than a raw citation count for judging whether a work is being built on or merely mentioned.
Last checked 2026-08-26 · reachable
Three habits make a citation durable. Carry a digital object identifier where one exists, because it is designed to survive the publisher reorganising its site. Where the accessible copy is a repository deposit rather than the published article, cite both and say which is which. And name the version explicitly: preprint, accepted manuscript or published.
The failure mode is a citation to a search result. Search interfaces change, result identifiers are not permanent, and a link into an index frequently stops resolving within a few years while the paper itself remains perfectly available. This is the same asymmetry that runs through the rest of this catalog, between a source published under an obligation and a source published as a product.
For the mechanics of where supplementary material goes in a submitted document rather than a repository, the entry on appendix formatting covers the style rules. For why durability of a source is worth this much attention, the record of closed services is the cautionary version of the argument.
PubMed indexes citations and abstracts for biomedical literature. PubMed Central holds the full text. A record can be in PubMed with no free full text anywhere, and a paper can be in PMC because a funder mandated deposit even though the publisher's own version is paywalled. Citing one when you mean the other is a common and confusing error.
It is a research paper that has not completed peer review. That makes it citable, and it makes the version material: findings and framing can change substantially between preprint and published version. Cite the preprint explicitly as a preprint, with its identifier, and check whether a published version exists before relying on it.
Search the aggregators first, because a deposited copy in an institutional or funder repository is frequently available even when the publisher's version is not. Failing that, the author's own institutional page or preprint server often has the accepted manuscript. What is not appropriate is a site redistributing publisher versions without permission.
Because it is maintained as an indirection. A publisher reorganising its site updates where the identifier points, so the citation keeps resolving. A raw URL into that site does not, which is why link rot in reference lists is concentrated almost entirely in citations that used the address rather than the identifier.
Every entry is a curated list of sources for one term, annotated and dated. The catalog index lists all of them.
The browser era of a mobile phenomenon, and what remains playable after Flash.
Open the entryApple’s bookstore under its original name, its rename, and where its records now sit.
Open the entryWhere bulky-waste collection is genuinely free, and how to tell a real scheme from a lead form.
Open the entry