Which edition to reference?

[quasihumanist]

Mathematicians don't usually deal with this kind of problem because they usually just pay lip service to the full history of what they study, so I'm hoping folks who do can just tell me what the right thing to do is and why.

I'm writing a paper on a concept that (as far as I can tell) was first introduced in an exercise(!) in the first edition of a (well-known) book back in the early 1970s.  The first paper to seriously study the concept was in the late 1970s.  The book came out in a second edition in the 1990s and contains changes to those exercises which reflect that paper and and some of the subsequent developments.  (This isn't something heavily studied.)

Which edition of the book should I cite?

(To complicate this, I doubt that I could get access to the 1st edition, especially given the pandemic, though I own the 2nd edition.)

[Ruralguy]

You cite the edition that actually has the exercise you are citing.

[dr_codex]

You cite the edition that actually has the exercise you are citing.

^ This.

Also, you add a footnote pointing out that there is another, significantly different, edition.

[ab_grp]

Does the paper cite the first edition and/or provide the exercise text? It seems as though a combination of the above could be useful.  As a reader, I would be interested in the history of the problem.  However, it sounds as though you are uncomfortable with citing the original if you don't have it in your possession and can't refer to it for accuracy (which I can understand)? Can you post the name of the book in case someone might have it? I have a couple hundred math books, and I would guess others might, too (unless you think it would out you, of course).  Does it exist online in part or whole somewhere?

[quasihumanist]

The book is Knuth's The Art of Computer Programming.  It's probably possible to find the current edition online (if only in pirated form), but historical editions require a library that didn't throw them away when the new edition came out.

Reading the paper that first studied this concept, my impression is that it was written in ignorance of the earlier appearance in the book, but the authors of the paper were made aware of the appearance in the book when the paper was first circulated as a preprint.  (Yes, this was the 1970s, but the authors of the paper (also relatively famous - but this could specifically out me) did know Knuth personally.)

Both editions have versions of this exercise(*), which defines the concept I'm studying and asks some elementary (i.e. reasonable HW for an undergrad class) questions about it.  The precise version of the exercise is irrelevant because I am citing to give credit for the definition, which appears in both versions.

(*) For the first edition, I only know this from the citation in the paper, which only refers to the book and does not mention the exercise specifically.  I know the exercise in the second edition is different because it has a citation to another paper written in the early 1980s.

[ab_grp]

Well, if you did want to get your hands on it at some point, it appears to be available on eBay in first edition.  My volumes are 3ed, unfortunately.  I could check to see which version my spouse has, if it helps, but it sounds as though you may not need the actual text.  Sorry I couldn't help, but it does sound like an interesting problem.

[Parasaurolophus]

If it's suitable, I can point you to an electronic copy of the second printing of the first edition, from 1969. (Or I could send you the pdf, if you prefer.)

[Hibush]

If you are writing about the  advances in programming that caused the exercises to change between the two editions, I would expect you to cite both editions. You are treating them as distinct works.

[Vkw10]

If Parasaurolophus' offer doesn't work out, then Kurth's webpage at https://cs.stanford.edu/~knuth/taocp.html has some publishing history that might help you figure out the volume you need so you can make an ILL request if your university or public library is doing ILL. It's also rather fun to read.

My library has Vol. 2 and has recently started doing ILL again, with the caveat that requests may be impossible since many libraries still aren't shipping books. Not having ILL for four months has reminded me just how important libraries are to my work.

[quasihumanist]

It looks like my library might be able to find me a copy of the 1st edition.  It'll take them a few days (if they don't mess up).

Let me try to explain the situation a little more.

In the first edition of the book, there is an exercise saying something like:

For this exercise, define a pidget of a goob to be blah blah blah.  Prove that the number of pidgets of a goob is at least the number of quidgets of the same goob.

The paper a few years later, says:

In this paper, we define the doodad of a goob to be blah blah blah.  We show that the number of doodads of a goob is at least the number of quidgets plus the number of tidgets.  Oh by the way, doodads were first introduced as pidgets in the book (citation).

In the second edition, there is an exercise, which cites a paper citing the above paper, saying:

For this exercise, define a pidget of a goob to be blah blah blah.  Prove that the number of pidgets of a goob is at least the number of quidgets of the same goob and at most twice the number of quidgets.

Now, I am writing a paper that proves that the goobs for which the number of pidgets is exactly equal to the number of quidgets plus the number of tidgets are exactly the goobs that are frumient.  In the introduction I have a sentence like "Pidgets were first introduced in book (citation) and first studied in paper (citation)."  I don't really want to add much more detail since this is not a history paper and I'm not trying to be comprehensive in the introduction.

Which edition of the book do I list in the bibliography?

[Parasaurolophus]

Now, I am writing a paper that proves that the goobs for which the number of pidgets is exactly equal to the number of quidgets plus the number of tidgets are exactly the goobs that are frumient.  In the introduction I have a sentence like "Pidgets were first introduced in book (citation) and first studied in paper (citation)."  I don't really want to add much more detail since this is not a history paper and I'm not trying to be comprehensive in the introduction.

Which edition of the book do I list in the bibliography?

Sounds to me like you do indeed need to cite the first edition (but not necessarily the first printing, since printings don't change content), since you're talking about the introduction of the exercise. So, if your library doesn't come through, my second printing of the first edition ought to do the trick.