I agree that if there is a reason to rewrite – if the field has changed, if there is data that better demonstrates the point, or if the author has changed their views, even slightly – then rewriting is called for. My claim is that if it is simply rewriting to satisfy a non-self-plagiarism condition, it is a waste of time; its a standard that doesn’t earn its keep.

I guess I have no real problem with the copying of other people’s introductions either, with proper citation. So far as I can see, though it is not widespread, I doubt that most people in the social sciences would have a problem with this.

I agree that there are some advantages to reusing text. On the other hand, there may be advantages to a policy against it as well. One observation is that rewriting some background sections is a very small task. (For me, at least, proving the theorems takes perhaps twenty times as long as writing them up. Rewriting 5 or 10% of a research paper will make only a tiny difference in overall productivity, especially since the background and introduction are by far the fastest parts to write.) So a correspondingly small benefit could make it worthwhile.

One potential benefit is that a newly rewritten introduction may better reflect the very latest views of the author and may be more closely tailored to the rest of the paper, compared to a single piece of boilerplate written years ago and then repeatedly reused. Another is that it encourages authors to use different examples and explanations. There’s no requirement to do so, but if you are writing several brief introductions, you’re likely to include a little more material in total than you would ever have put into a single introduction.

It doesn’t seem completely clear which effects are most important in practice.

From my perspective, the only real ethical issue is in compliance with community standards. If your colleagues expect each paper to be written from scratch, then secretly reusing text is a mild form of misrepresentation. On the other hand, if everyone agrees that reusing text is fine, then there’s of course no ethical issue.

Incidentally, what do you think of copying other people’s backgrounds or introductions? In the last year or two, there was a big scandal about this in physics. It particularly comes up for non-native speakers of English, who sometimes argue that writing background material is a real burden for them, and that if someone has already done a good job of it, what’s the harm in adapting it rather than starting from scratch? Of course, I think most people would consider it unacceptable plagiarism if it is done without explicit notice. However, should people be allowed to write something like “Most of the following section is taken verbatim from the excellent introduction to [1].”? Do you know of any fields where this is explicitly allowed? That would be a really interesting data point.

I’m the guy you thought had low ethical standards. To be fair, you changed that view when you realized that different fields have different standards. But now I want to defend those standards as ethical. Suppose you are a scientist and you are developing a hypothesis. To be concrete, let’s talk about psychology. Suppose your hypothesis is that brain modules got more highly interconneted over evolutionary time. You write a paper defending this view, citing neuroanatomical and functional evidence. Then it occurs to you that you can use this thesis to develop a view with regard to human creativity. You write that paper. Now do you really think it is – at best – permissible for you to reuse the conclusions and quick summary of your view in paper 1 in paper 2? Any field that requires you to rewrite those conclusions in new words is a field that requires you to waste your time. Any field that requires you to redefend the view (of course in new words) is not only wasting time, it is taxing readers’ patience beyond endurance. Since both of these practices impede the growth of knowledge, they are unethical.

So how do you deal with (previously-published) axioms, definitions and so on? Omit them, reformulate them each time, or cite them to an earlier paper?

If they’re simple, one just restates them on the fly. (They probably come out pretty similarly worded, and perhaps identically. In any case, the important thing is the meaning rather than the wording.) If they’re complicated, I don’t think people would object to copying a definition or theorem statement from another paper, assuming the original source is of course clearly cited.

Theorems and definitions are usually pretty short, though, typically at most a few sentences. My feeling is that copying multiple paragraphs would be really frowned upon (although you could probably get away with it), and copying multiple pages would get you in serious trouble with the journal editor.

So how do you deal with (previously-published) axioms, definitions and so on? Omit them, reformulate them each time, or cite them to an earlier paper?

One of the main issues seems to be how much the standards vary from field to field. I work in pure mathematics, where the standard appears to be 0% reuse. I’ve never discussed it with an authority figure, but I never imagined any substantial degree of reuse was acceptable (and I’ve confirmed this feeling with several colleagues). It’s probably fine to copy a few sentences from a previous paper, but copying numerous paragraphs may open you to charges of unethical behavior, even if you cite the paper they appeared in. If you tried to publish a paper consisting of 50% previously published text, it would actually derail your career. (The one exception I know of is for material not previously published in a journal but rather in conferences in certain fields, like theoretical computer science. There, it can be reused in journal papers, but only if they are clearly labeled as revised versions of the conference papers.)

So when I saw comments like #1, I assumed the author just had really low ethical standards. It wasn’t until comment #8 that I realized there might actually be differences in the publicly articulated standards in different fields. It’s intriguing.

Do people in other fields get guidance like this in graduate school? If so, what’s considered acceptable?

(They may have been sometimes cold, smug, and harsh, but that was one program that was very thorough about induction into the profession.)

Conference papers are a whole other story – an article could be nothing but a reworked conference paper. That’s exactly what conferences are for.

So I don’t know what these authors mean by clearer rules, because I got very clear rules; they were just rules these particular people don’t happen to like. If journal editors don’t like them, perhaps they should establish specific policies in their submission guidelines. But as Adam said, who appointed Collberg and Kobourov high poobahs of scientific merit?

Simultaneous publication of similar articles tailored to different outlets is, meanwhile, quite different. If one is, say, presenting three case-studies built around the same conceptual frame, one is necessarily going to have to explain that frame, in basically similar ways, in each of the three pieces. Unless there are those who would argue that one should deliberately refrain from seeking to publish more than one such study?

yes. If this is self-plagiarism I guess most professional writers of non-fiction do the same thing. It is behavior that is economically motivated in that case, but probably the same argument can be made for academic research with its publishing requirements leading to a similar motivation.

When I was in college, my intermediate macroeconomics professor was up for tenure. Not that great of a professor, but not too many of my econ classes could tout that. At any rate, he was a world class publisher. He had four articles out in leading economic journals. He didn’t get tenure because the department wanted more articles than he had published. Where’s the incentive to write high-quality comprehensive examinations of a subject when the reward schedule favors quantity over quality?

So the solution should be thus: departments should favor not necessarily fewer, but higher-quality (ie non-self plagiarizing or repetitive) works by professors. There’s obviously more to the subject that I’m overlooking for the sake of simplicity, but I think this is comprehensive to what is at hand.

The following stinks of sour grapes:
“It also rewards those authors who are able to break down their results into overlapping least-publishable-units over those who publish each result only once.”

If the journal reviewers feel the contribution was original and sufficient then the work deserves to be published. Nobody appointed Collberg and Kobourov high poobahs of scientific merit. Also—wanna bet there is overlap in the technical report and paper listed on the website?