I have a confession to make: I read Retraction Watch. It's one of my pieces of mental junk food, like bad soap opera. And though I started for the soap-opera and schadenfreude, I've stayed also because of the serious questions about ethics that get raised.
First, a little background: for those who don't know, a retraction is when a scientific publication gets, well, unpublished. It can be for benign reasons, like discovering that a critical sample was mislabeled, or it can be for misconduct like plagiarism or fraud. It's pretty much always a black eye for everybody involved however: even the benign reasons for retraction are generally seen as signs of shoddy scientific behavior.
So retractions are basically tabloid fodder of the scientific world to begin with, and Retraction Watch is a project by two science journalists to keep an eye on what scientific publications gets retracted and how. It's actually a serious issue, since publishers are often just as happy to sweep retractions under the rug, and apparently a surprising number of citations accrue to papers after they have already been retracted, by people who don't realize they are building on now-invalidated science. I can understand that, really, since once I've cited a paper once, I'm not likely to go re-check on the paper each time I cite it again in the future---after all, I already know the work, right?
Reading Retraction Watch has made me think a lot more about my responsibilities as an author, especially when a) there's a big temptation to shade things in ones own favor and b) I'm getting involved in more and more interdisciplinary research where it would be easy for me to not know if theres something funky in a collaborator's contributions. Most of the time, the cases are pretty simple, and there's not much lesson to be drawn from them: "Don't pretend that you did work that you didn't do," and "Don't put your name on somebody else's paper" are pretty elementary. But then there are the people caught by dishonest co-authors, the authorship disputes, the priority fights, that all give case studies and insights into how things can go wrong for a researcher who is trying to do the right thing---important life lessons to try to learn so you can avoid falling into the same traps.
And then there are the boundary cases, like: how much text reuse counts as self-plagiarism? I certainly am not averse to reusing chunks of related work or background sections, modified of course for the audience. But having an ongoing research program means you will often be talking about the same project to the same audience multiple publications in a row, simply because you have new results in the same area. It's pointless, I think, to paraphrase only for the sake of paraphrasing in a related work section. The introductory material, however, I think should pretty much never be reused: if your understanding of what your work means and how it fits into the bigger context isn't changing, then you aren't paying enough attention, either to your own work or to the world around you.
Different people, different research communities, different organizations all have different standards: for example, in biology it seems that even just putting out an abstract about something can be considered a "self-scoop" and preclude publishing in major journals, while in computer science it is standard practice for journals to invite "extended versions" of existing papers---of course with clear marking to that effect at the top of the paper, including a pointer to the prior version. Personally, I think the "reuse" of computer science is the more ethical position, since it balances fast publication of important results with the ability to get more highly detailed and thorough explanations into the literature as well.
Whatever the issues, watching the articles and the arguments on Retraction Watch gets me thinking about all these border cases, and at least ensures that I am always consciously acting by a standard that I have good reason to think is ethical, and that I would be willing to defend.