Innovation in scientific publishing and its implications for Crossref DOI registration practices - Request for input - Crossref

Content Registration
, , , ,
1

Lots of exciting innovations are being made in scientific publishing, often raising fundamental questions about established publishing practices. In this guest post, Ludo Waltman and André Brasil discuss the recently launched MetaROR publish-review-curate platform and the questions it raises about good practices for Crossref DOI registration in this emerging landscape.

This is a companion discussion topic for the original entry at https://www.crossref.org/blog/innovation-in-scientific-publishing-and-its-implications-for-crossref-doi-registration-practices-request-for-input
3

Hi Ludo, André, and Ginny,

Thanks for the post and the interesting questions posed. Having come across some of these issues at eLife I thought I’d share some relevant thoughts/questions that came up during discussions with Crossref when determining our DOI convention for eLife’s PRC model:

  • Does the content you publish have a different citation to the preprint (on the preprint server)? If so, it’s a good candidate for a new DOI (for various reasons)

  • Do you consider your organisation as a (joint) custodian of that content? For example, if there were concerns with its validity would you issue your own withdrawal, retraction, expression of concern or update notice? If so, it’s a good candidate for a new DOI.

  • You cannot register DOIs for reviews without specifying what it’s a review for (which is logical and fundamental). Registering your own DOIs gives you greater control over what to attach the reviews to. We register a distinct DOI for each version, and relate the accompanying reviews to that specific version. Preprint servers follow different conventions, with some registering a DOI for each version and some opting to register one DOI per preprint. For the latter, how can you distinguish which version it relates to? This is currently only possible by linking to a specific DOI (or a versioned URL, which is not persistent).

  • Crossref’s relations model is really powerful (and IMO under-utilised). We make quite extensive use of it, to always link back to the preprint server, and between the versions we publish (as well as datasets, archived code and so on). In my view, increased use of the relations model enables greater transparency and is fundamentally more representative of what the work is.

For your second question on types, we determined that the only way to authentically register our Reviewed preprints was to register preprint DOIs with accompanying review DOIs. There are some features of our model that MetaROR shares and some that it does not (or it’s not clear to me if it does). Here are the features that helped us determine our path forward with types:

  • Under our model, following publication of an eLife Reviewed preprint authors are free to submit to another journal. Registering Reviewed preprints as journal articles would make the identification of truly ‘duplicate’ publications more difficult, and could cause further complications for authors in the existing ecosystem. Registering these as preprints still emphasises the role of preprints while not precluding us from linking to the ‘final’ version published at a different journal.

  • A (deliberate) consequence of our model is to undermine the importance of a single final, purportedly ‘validated’ version of record. Registering these items as journal articles serves to perpetuate this reliance.

  • In addition to Reviewed preprints, eLife also lets authors publish a finalised eLife VOR (because, despite my comment above, this is still meaningful for some authors, allows them to fully engage with the existing ecosystem, and represents one part of the ‘curation‘ aspect of PRC). There is therefore an obvious separation built into our model between preprint and journal article.

Finally, this is probably not going to be the last ‘new type’ of scholarly communication - we’d favour an approach that relies less on (the proliferation of new/different) types, and more on relations between objects.

All the best,

Fred

1 Like
4

Hi
Thanks Ludo and André. Interesting topic!

At Peer Community In, we believe that articles submitted to peer-review platforms such as MetaROR should ideally retain the same DOI as the version on the preprint server, rather than being assigned a new one. Our reasoning is as follows:
-Submitting a preprint to a peer-review platform does not, by itself, involve modifications by the authors.
-While the process may generate reviews and assessments, these evaluations can be either positive or negative.
-Recognition of a “peer-reviewed” status seems most appropriate when an article has been meaningfully revised in response to reviews and when the editorial team considers it deserving of a particular mark of quality or validation (a positive curation).
-Conversely, if no substantial revisions are made, or if the editorial team issues a negative assessment, assigning a new DOI could be misleading.
-A new DOI in such cases might inadvertently suggest to readers that the article has undergone a validation step comparable to journal publication, which may not be the case.

For a more elaborated description of the ambiguities in Peer-Reviewed Preprints and the PRC model, see https://www.coalition-s.org/blog/peer-reviewed-preprints-and-the-publish-review-curate-model/

Denis Bourguet, Thomas Guillemaud (Peer Community In)

5

As far as I understand, from a metadata perspective, relationships such as isReviewOf, reviews, isRelatedTo, and in some cases hasPreprint / isPreprintOf, already allow links to be expressed between preprints and their reviews. What I am still trying to better understand is how Crossref envisions representing these relationships over time, particularly when a reviewed preprint later evolves into a substantially revised version or a formally published journal article.

I am not an expert in Crossref’s underlying infrastructure, so this is very much a question rather than a proposal: could Crossmark ever play a role in signaling version awareness or continuity across related DOIs—linking preprints, reviews, and published articles—even when those DOIs are registered by different entities? Or is relying on relationship metadata considered the intended and sufficient approach within the current Crossref model? I would appreciate any clarification on how Crossref sees this evolving.

Rosario

1 Like
7

Thanks for the comment. Our current thinking is that capturing relationships between items is sufficient to represent different versions and how they have been reviewed, even when they have been published by different organisations. We have also recently added a version field, which indicates when something is likely to have different versions.

As this discussion highlights, though, it’s an evolving area and likely to become more complex in the future. We want to keep the metadata accurate and simple to interpret while making space for innovations in publication, and the current approach seems flexible enough to do that for now. We’ll keep an eye on trends in the community and see whether there are other complementary approaches to add down the line.

2 Likes
8

Let me begin by outlining how we at JMIR Publications handle DOI registration for our publish-review-curate (PRC) journals, before turning to what I believe is the deeper conceptual issue underlying your questions.

JMIRx is an overlay journal series published by JMIR Publications that has operated a publish-review-curate model since 2019. In JMIRx (e.g., JMIRx Med), peer reviews and authors’ responses to reviews receive their own DOIs because they are original scholarly objects. We agree that reviews should be treated as first-class research outputs.

However, it is important to remain transparent about the current ecosystem: assigning a DOI does not automatically result in bibliographic indexing. The National Library of Medicine (PubMed) has explicitly informed us that peer-review reports are not indexed, even if they carry DOIs. The aspiration that reviews will be treated identically to traditional journal articles in discovery systems remains, for now, aspirational.

For the preprint itself, we do not mint new DOIs. The preprint DOI (assigned by medRxiv or bioRxiv in our case) remains untouched. We do not republish identical documents, and we do not create duplicate DOIs for the same object. We assign a new DOI only to the final, accepted version of record — the version that we (or Plan P partner journals) formally curate after peer review.

That distinction is crucial.

The binary editorial decision to accept or reject is not an incidental feature of publishing. It is the institutional act that transforms posted content into a curated scholarly record. Once an article is accepted, it becomes associated with our brand. We assume responsibility for permanent archiving (e.g., PMC), submission to PubMed, post-publication commentary, corrections, and retractions. That is what curation means in practice: stewardship and accountability.

In your post, you suggest that for some new platforms “the traditional distinction between peer-reviewed articles published in scientific journals and non-peer-reviewed articles posted on preprint servers is no longer applicable.” I would argue that the traditional distinction has evolved — but it has not disappeared.

The distinction was never primarily about whether a manuscript has been reviewed. Peer review is a process. Journal publication is an institutional commitment. What distinguishes a journal article from a preprint is not the presence of comments or reports, but formal certification and long-term stewardship.

The distinction is therefore better framed as:

  • posted content vs certified and stewarded content

  • evaluation vs institutional commitment

  • review vs curation

Publish-review-curate models have modularized these functions, which is an important innovation. Peer review can now occur independently of certification. But unless an editorial entity assumes irreversible responsibility for a version of record, the object remains posted content — even if richly evaluated.

This connects directly to the argument articulated by eLife that a deliberate consequence of their model is to “undermine the importance of a single final, purportedly ‘validated’ version of record,” and that registering reviewed preprints as journal articles would perpetuate reliance on that concept.

It is entirely legitimate to weaken the aura of validation attached to a version of record. Science should be iterative and open to revision. But the version of record is not primarily about validation; it is about fixity, accountability, and governance.

The scholarly infrastructure depends on canonical reference points:

  • Citations must resolve to stable objects.

  • Retractions must attach to a specific version.

  • Legal and ethical responsibility must attach somewhere.

  • Indexing systems require a canonical referent.

  • Tenure and funding evaluations still operate on finality.

One can decouple peer review from certification. That is what PRC models do. But eliminating the primacy of a stewarded version requires redesigning indexing, accountability, and archival systems. At present, those systems remain binary. An item is either indexed in PubMed or it is not. It is either curated under institutional responsibility or it is not.

It is also worth noting that even eLife’s own architecture retains this separation. They explicitly offer a final eLife version of record because it “is still meaningful for some authors” and allows engagement with the existing ecosystem. That implicitly acknowledges that the VOR remains operationally necessary.

The deeper issue, therefore, is not metadata classification. It is governance.

If a platform:

  • does not make a final acceptance decision,

  • does not assume exclusive archival responsibility,

  • does not handle retractions as steward of record,

then registering the object as a preprint is coherent. But then it should also accept that it is not performing full curation.

One cannot simultaneously:

  • reject the institutional importance of a version of record,

  • avoid assuming long-term stewardship,

  • and claim equivalence with journal publication.

The real innovation of PRC is not abolishing the version of record; it is modularizing evaluation and certification. That is a powerful shift. But certification and stewardship remain binary institutional commitments.

Against that backdrop, my answers to your two questions are straightforward:

Q1: Is it acceptable to have two Crossref DOIs for essentially the same article?

If the document is identical to the preprint, a second DOI for that same object should not be minted. Preprint DOIs should remain “posted content.” What can and should receive DOIs are the review reports and editorial assessments, properly linked to the preprint via Crossref relationship metadata (e.g., isReviewOf).

Q2: Should items on the MetaROR platform be assigned type “article” or “preprint”?

Absent a binary decision that confers permanent curatorial responsibility, the object remains a preprint — albeit a reviewed one. Elevation to “article” status signals that an institution has assumed stewardship of the version of record.

In summary, the question that should guide whether an article-type DOI should be assigned or not (which is a binary question) is not whether peer review has occurred. The question is the publisher or or some other entity willing to assume irreversible and exclusive responsibility for the scholarly object. Exclusivity is required because it violates scientific norms to publish the VOR of an article in multiple journals/venues. That institutional commitment — not the presence of reviews — is what continues to distinguish posted content from a curated journal article.

Gunther Eysenbach MD, MPH, FACMI
Adjunct Professor, University of Victoria

*Laboratory for AI and Innovation in Scholarly Publishing
*

Chairman & Publisher,

JMIR Publications