Citation labels

There are various ways to label citations (references to the bibliography of a paper). This page looks at the difference between the following classes of citation styles:
numeric
e.g., [25], or 25. The entries in the bibliography are either sorted by author (LaTeX's plain bibliographystyle), or in the order of first citation (LaTeX's unsrt bibliographystyle).
name-based
e.g., [BCKT89] (LaTeX's alpha bibliographystyle) or (Briggs et al., 1989), for the paper
Preston Briggs, Keith D. Cooper, Ken Kennedy, and Linda Torczon. Coloring heuristics for register allocation. In SIGPLAN '89 Conference on Programming Language Design and Implementation, pages 275--284, 1989.
Which one should you use? A name based label style!

Why use a name-based label style?

Because it makes reading easier, for the following reasons:

But style guides recommend numeric labels

The reason these guides give is that numeric labels supposedly discourage the mistake of treating the citation as part of the sentence (it should be treated purely parenthetical, i.e., if you delete it, the sentence should still be complete). However, I have seen enough citations like:
When a backward computation algorithm constructs the entire dynamic dependence graph prior to slicing (e.g., Algorithm III in [1]), as our prior experience in [26] shows, for [...]
(wrong) in spite of using numeric labels, and I have also seen correct citations used with name-based labels. The sentence above should be something like:
When a backward computation algorithm constructs the entire dynamic dependence graph prior to slicing (e.g., Algorithm III of Agarwal and Horgan [AH90]), as our prior experience shows [ZGZ04], for [...]
or something like
When a backward computation algorithm constructs the entire dynamic dependence graph prior to slicing (e.g., Algorithm III of Agarwal and Horgan (1990)), as our prior experience shows (Zhang et al., 2004), for [...]
So, to get people to use the correct writing style, numeric labels are insufficient. We have to teach writing style to them.

But even if the claimed connection between label style and writing style existed, the readability advantages of name-based labels would trump such a minor writing style issue.

But doesn't a name-based style consume more space?

They may consume a little more space, but the enhanced readability is worth this cost many times over. There are better ways to save space.

Also, authors may work around the unreadability of numeric labels, resulting in much more space consumption than a label style like LaTeX's alpha style. E.g., instead of

We compared a conventional colouring register allocator [BCKT89] to [...]
an author might write:
We compared the conventional colouring register allocator of Briggs et al. [25] to [...]

But the style imposed by the publisher requires numeric labels

Ignore that requirement, if possible (and it usually is). It's the reviewers who decide whether your paper is accepted, and they probably prefer the name-based style (if they have thought about the issue). It's the readers whom you want to read (and cite) your paper, and making the paper more readable is certainly good for them.

I have ignored such a requirement a few times until now, and I have not received any comment from reviewers or publishers about that.

If you are preparing a paper for ACM proceedings, don't use the official sig_alternate.cls etc. style (which is broken in many ways in addition to not allowing the BibTeX alpha style); if you want you paper to look in the ACM style (except for citation labels), use Mike Sperber's sigplan-proc.cls style.

If you organize an ACM event, note that ACM actually does not require you to use their style (although that's not clear at first); at least that's what I found out pretty late in the IVME '03 game. Similar things probably apply to other publishers.


Anton Ertl