If there is a tree of papers on or involving X, all of which give a brief introduction to X and I need a citation for my use of X, do I need to give the root of the tree as a citation of X or can I cite any of the nodes in the tree that explain X for my use of it?

Concretely I need to cite why using point wise mutual information can be used to improve co-occurence matrix based word embeddings. I do not really know where PPMI was first used for this purpose, but I have stepped down towards the root for a few papers now. Do I need to find the paper that first used PPMI for this or can I cite any of the ones I already found?

  • 1
    Even if you don't need to, it might be to your benefit to cite the "root", since people searching for work on the topic might search for papers that cite the root paper. Being one of those hits would give your paper higher visibility. Commented Jul 10, 2019 at 1:48

2 Answers 2


I (as a mathematician) would use one of three options:

  • cite the most recent paper. This is what I do when I only need a reference to the topic and no particular result in detail; ideally, the most recent paper is a survey (or sometimes I pick the most recent survey), this way I can skip a lot of references in my own article saying, "see [xyz], and the references therein." This somehow implies that the contribution of my paper is related, but it is nowhere necessary to know the exact details of the results of that source.

  • cite the original source. This is rare in my case, because over the years people have added so much knowledge that citing the original works seems to be a bit out of date. I only do this when I really want to stress that this origin is important for my work in particular. As I said, this rarely happens. I am pretty sure that many people who cite the original source have never read the original source and just cite it "because you cite it."

  • cite some other paper(s) "from the tree" (as you call it). This is what I usually need and do, mainly because I need to refer to several different details from different sources. Then I have a reason for referring to an "inner node in the reference tree" (neither root nor leave), and I would only do that if I can give a reason, either only to myself, or even write in my own paper why I pick this particular reference.

  • 2
    This assumes, of course, that the chain of citations from current papers is unbroken back to the first mention. If it is broken, then you need to cite additional papers, perhaps, so that you capture everything relevant to your work. There are a number of cases, such as parallel work, in which the chain is incomplete.
    – Buffy
    Commented Jul 8, 2019 at 12:15
  • @Buffy thanks, yes, I would cover this in my third case. Commented Jul 8, 2019 at 22:25

Try to choose the most precise reference. If there is truly no difference between sources and you could cite any number of references, try to find a good review paper to cite.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .