Initial list as proposed by Peter Walsh:
- Citations should reference prior work that the reader needs in order to understand the current work, and nothing more. Cite when you are building on the data or conclusions of a previous paper, and cite when your methods are not obvious and the cited paper explains them in detail.
- Cite when the wording or presentation of an idea might lead the reader to conclude that you had the original idea, but it's preferable to get around this by using different wording.
- Do not include a comprehensive literature review as part of your paper. (Unless, of course, you are writing a literature review, in which case the cites are considered data.)
- Don't cite anything that can be found in a textbook, being common knowledge.
- The introduction (or background) can have a cite or two that gives an overview of the problem statement, but generally speaking these should refer to textbooks and not other papers.
- There's a real problem in some fields with "citation inflation", where authors include a literature review and pad it with cites to make word count. Reviewers sometimes require cites of their own papers to raise their own impact level. (Generally, a reviewer will be someone in that field, so is likely to have closely associated papers.)
- You can break any of the rules if you feel there is a good reason to do so, and citations are no exception. If you feel that something is particularly significant and should be referenced in your paper, or that a cite is needed for some specific reason, then feel free to do so.
- In general, if you have more than a dozen cites you should review them carefully to see if all are truly needed. It is perfectly acceptable to have a paper with as few as four cites, and if you are doing something particularly creative you might have no cites at all.
Feel free to add your suggested changes either to the comments section on this page, or, if you're a project member, update this page directly.