In conventional academic publishing, an article usually goes through submission, peer review, revision and acceptance before its final published version appears. My own experience with a manuscript submitted to the Journal of the Association of Information Science and Technology (JASIST) illustrates this process. My manuscript went through two rounds of peer review, and I revised it in response to the reviewers’ comments. These earlier versions were part of the process of improving the article, but they were not published as separate articles with their own DOIs. The process ultimately resulted in one published article.

F1000Research follows a different approach of post-publication peer-review. It allows an article to be published before peer review is complete. Authors can then revise it in response to reviewers’ comments and publish later versions. Earlier and later versions remain available and each can be cited separately.

In other words, what is normally a series of steps leading to one published article can become a series of published versions of the same scholarly work.

I began to think of such a work as an evolving scholarly object: a contribution that changes over time while its earlier and later published versions remain part of the scholarly record.

But what happens when databases try to describe these changing works? A title may change between versions. Different databases may classify the same work differently or fail to show clearly how its versions are related. The information used to identify and describe the work may therefore be inconsistent. I call this metadata instability.

Another problem concern counting. If one scholarly contribution appears in several records, a database may show several publications even though those records relate to the same underlying work. This does not necessarily mean that the records are incorrect; the different versions have indeed been published. However, this creates publication divergence. The difficulty is understanding how the records relate to one another and what they represent.

Citations introduce a further complication. Readers may cite different versions of the same work, so citations can be spread across several records instead of appearing under a single record. I call this citation divergence. Figures 2 and 3 illustrate examples of different published versions being cited in a subsequent article. Such divergence may affect citation-based indicators.

Figure 2 – Different published versions cited in a subsequent article

Figure 3 – Different published versions cited in a subsequent article

These distinctions matter because researchers use bibliographic databases to find articles, count publications, follow citations, and assess research. If the relationships between published versions are not clear, publication and citation counts may not tell the whole story.

To investigate this problem, I examined how F1000Research was represented in Scopus and OpenAlex. I was not trying simply to decide which database was better. I wanted to understand how they represented a publishing model in which one scholarly contribution could have several published versions.

Looking back, what first appeared to be a problem of duplicate records led me to a broader question about the changing nature of scholarly publishing. The issue was not simply whether records should be removed. It was how databases could represent different published versions while making clear that they belong to the same underlying scholarly work.

If one scholarly work can have several published versions, each with its own record and citation history, how should databases represent those versions while preserving their relationships to one another?

That is the question at the heat of my study.