It occurs sometimes that I am in possession of a factoid for which I know all of the standard terminology and can express the factoid in that terminology, and also know that the factoid is prior knowledge, but do not have a reference for the factoid. This is especially problematic in my field (machine learning) where the online resources for learning things are very good, and quite extensive. However, citations are sometimes sparse in these blog-style articles, and I find myself knowing the subject matter, but not knowing the appropriate citation for it. Note that the blog is not the appropriate citation, because they're just paraphrasing it from elsewhere. It would be nice to lean on the experience of the community, but "Does anybody know a citation for fact X?" is generally an inappropriate question on stack exchange sites.

Surely this is a problem in other disciplines as well. As such, I am mostly interested in general "citation reverse-search" skills, tips and tricks. However, field-specific answers are also welcome.

NOTE: This is specifically for situations where the problem is known to be a prior one, and where the terminology is known. Linking disparate fields with differing terminology of basically the same concept is a known Hard Problem™, and I am not trying to solve it here, except insomuch as those same skills relate to "reverse-search" of problems with known terminology.

To reiterate for clarity: What are handy tools, tricks tips etc for finding references for known facts?

  • 1
    Leaving a comment on the blog post works sometimes, as do search engines.
    – Anyon
    Dec 26, 2018 at 15:35
  • 7
    On Mathoverflow, "does anybody know a citation for fact X" is a perfectly acceptable question (provided it is a research-level one); we even have a tag for it. Dec 26, 2018 at 17:02
  • 3
    Yeah, can you explain why you think such questions are generally inappropriate on the discipline-specific Stack Exchange sites? My experience is that they are appropriate and generally received well. Note, however, that the right answer to such a question is sometimes "this is too well-known to need a citation". Dec 27, 2018 at 7:54
  • 2
    Note that one (of many) reason that "does anyone have a citation for fact X" is an acceptable question, is that just because "everyone knows it" and "all the blogs assume it" and "all of our research is based on top of it" and "it always seems to be true in every case I've seen" (you can probably see where this is going ... ;) ) doesn't necessarily mean it's true, and certainly doesn't mean that there was at some point research supporting it.
    – Brondahl
    Mar 1, 2019 at 11:47
  • 2
    Discovering that a widely held belief is actually not based on any research is a valuable realistion and something that should be called out in whatever academic paper you are writing that would otherwise be using a citation
    – Brondahl
    Mar 1, 2019 at 11:48

4 Answers 4


Actually, there are often many "facts" that really aren't. It is a fascinating research question to dive into the internet and libraries to verify the origin of something taken as a fact. This happens in every field and especially with pithy "quotations" that appear on calendars. Just for fun I like to look for the true source, in one instance I came up with five different supposed authors! This is also why you should reference your facts: So that others can verify that you are stating it correctly.

In the area of machine learning you can use the ACM or the IEEE digital library to dig around and look for a source. There are no shortcuts, this is research :)


Wikipedia. Go to the bottom of an article to find references, open the first paper cited and then find the references on that paper.


I usually have this problem when the fact is not so obvious as to not need a citation (i.e. standard undergraduate coursework material) but not so difficult as to have been addressed in a seminal paper. Textbooks are a good way to fill in the gap, and I think this is how seminal textbooks work their way to so many citations. In ML, ESL is a good book to check against. BDA3 crops up in Bayesian stats papers a lot; a "fact" I might cite this book for would be the asymptotic normality of a posterior distribution under certain regularity conditions, which is about at the level we're talking about.


Since you indicate that the relevant terminology is known, you can search for this terminology online to find other academic sources that discuss the fact (and hopefully reference its origin). Of course, doing so using google would probably just lead you back to the educational blogs that you picked up the fact from in the first place. Instead use (something like) google scholar, or potentially better a subject specific abstract indexing service (I don't know which one are common in machine learning and computer science).

You must log in to answer this question.

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