21

A paper on arXiv can have multiple versions. It is rather frequent (at least in math) that the first version and the latest version are even from different years. When citing such a paper, is it better to put in the year in which the paper was first published on arXiv, or the year of the latest version (and probably explicitly saying which version is cited then)?

Assume that neither the journal nor the publisher have any written guidelines about this particular scenario; for example, when one is themselves putting a paper on the arXiv.

Improve this question
2
  • 7
    Since you can always access past versions of a paper on the arxiv, why not cite a specific version? thus, write "arxiv:1010.1212v3". Then there is no question about the year of publication.
    – the L
    Commented May 19, 2016 at 10:46
  • That shouldn't be a big problem. Usually by the time your paper is accepted, the paper you cited are very likely published in another journal. Cite the journal then.
    – Arctic Char
    Commented May 20, 2016 at 2:53

2 Answers 2

Reset to default
28

I would cite the exact version you are referring to in your work. Therefore the year should be the year of that version. You should also use the actual arxiv identifier in the citation. This is to minimise the effect of potential changes in the cited work on your paper. Moreover, your goal is to provide the reader with a precise pointer to the cited work so that they can easily find it.

Improve this answer
1
  • 3
    I can't +1 this often enough...
    – darij grinberg
    Commented May 19, 2016 at 13:54
2

In this case I would likely go with the date of the original posting, unless there are fairly significant differences between versions (a paper might have its title changed, or results noticeably broadened or restricted from the original posting, for any number of reasons). One way or another, it is highly recommended to have an "accessed on" date in the bibliography entry for any citation of online materials, such as arxiv preprints. It would be naturally assumed that you were consulting the most recent version of the posting available at that time, and not later or earlier ones, which is easily determined and accessed (at least in the case of the arxiv). Without this the reference can otherwise be problematically ambiguous, as it is entirely possible that the cited preprint will get a new version at several stages during the process of getting your own published, and the cited article may get published and get new arxiv versions after your own has been published and given a "final" arxiv update.

Improve this answer
2
  • 1
    Tiny comment: For arxiv, the 'accessed on' date is superfluous if you state the version number. And if you don't state the version number, it's potentially not enough to say accessed on since you could access any existing version on that day.
    – Joel
    Commented May 19, 2016 at 17:48
  • "[I]t's potentially not enough to say accessed on since you could access any existing version on that day." Strictly speaking, yes. But there's a very strong implication that an access date with no specified version number means the most recent version available on that date.
    – David Richerby
    Commented May 19, 2016 at 18:41

You must log in to answer this question.