Once you've come up with the perfect title for a paper -- one that will convince people to read past the title and one that does not involve a yes-or-no question -- you then need a compelling abstract and introduction. Today I feel like writing about Introductions because recently I've been involved in a paper in which a colleague and I differ greatly in our Philosophy of Introductions.
This colleague doesn't like introductions that spend much, if any, time/space talking about the larger context of the work. To him, this is fluffy decorative stuff that detracts from the paper's main purpose: to present new data/ideas. His preferred introduction goes straight to the most detailed and technical level of the paper: We did X and Y and here it all is.
I agree that introductions shouldn't go on for so long and in such a general way that the reader becomes impatient and thinks "So what did you do? What is this paper really about?". I do, however, like to start big and work my way to the more technical levels, as to me this a good way to explain why we did the work and why anyone else might be interested. It should be possible for most papers to do this is an efficient and interesting way.
I suppose I am also aware that my intended (= hoped-for) audience isn't just the small group of people in my specific field at its most narrow definition. I'd like graduate students and colleagues in related fields to understand my papers. That doesn't mean I explain every single term and concept in great detail as if writing for a non-expert, but it does mean that I don't assume that readers will immediately understand the motivation and context of the work.
When I give an invited talk at some universities, students read 1-2 or my papers in preparation for my talk and visit. In some cases this is part of an organized seminar intended to get students more involved/interested in the seminars, and in other cases the reading is part of an informal research group activity. Discussing papers with these students is actually a great way for me to figure out whether my papers are understandable to anyone but me and a couple of reviewers (albeit too late for me to fix any problems if the papers turns out to be rather cryptic to non-specialists). [note: In some cases I am asked to recommend which of my papers would be most suitable for this purpose, and in some cases I am not asked]
Based on this kind of feedback, I know that some of my papers are not very accessible to this broader audience. (note: "not very accessible" is a euphemism for a highly technical jargon-filled paper of uncertain purpose and result). That's OK, as long as some are reasonably accessible, e.g. review papers, slightly longer papers that have room for an expanded introduction, or short general-interest papers.
I was going to come up with a hypothesis about why some people hate general intro sections in papers and others like them and think they are important, but none of my hypotheses withstood even my own brief scrutiny. Example: The aforementioned intro-hating colleague is way more famous than I am and people are going to read his papers no matter what is in them; maybe he doesn't feel the need for extraneous intro material because he knows his papers will be read anyway. But then, it's not hard to think of other famous scientists who think that people will be fascinated by their every utterance and so they utter a lot, and a lot of it is not interesting.
Surely there is a happy compromise in even a fairly technical paper -- i.e. introductory text that gives the broader context but that doesn't wallow in it for pages and pages.
10 years ago