“Most teaching on methods assumes that the student will start a research project with a general question, then narrow that to a focussed question, which will dictate the kind of data needed, which will in turn support an analysis designed to answer the focussed question. Nothing could be further from reality. Most research projects – from first-year undergraduate papers to midcareer multiyear, multi-investigator projects – start out as general interests in an area tied up with hazy notions about some possible data, a preference for this or that kind of method, and as often as not a preference for certain kinds of results. Most research projects advance on all of these fronts at once, the data getting better as the question gets more focused, the methods more firmly decided, and the results more precise. At some point – the dissertation-proposal hearing for graduate students, the grant-proposal stage for faculty, the office hour with the supervising faculty member for any serious undergraduate paper – an attempt is made to develop a soup-to-nuts account of the research in the traditional order. Now emerges the familiar format of puzzle leading to literature review leading to formal question, data, and methods. Even then, the soup-to-nuts menu is likely to be for a different meal than the one that ends up in the final paper.”
Andrew Abbott, 2004, Methods of Discovery: Heuristics for the Social Sciences. W.W.Norton.