Where one begins depends to a considerable degree upon how broad or narrow the original conception was.

Consider "ethical issues of genetic engineering" --a topic so broad that the problem is 'too much' information. Think first about the label: "genetic engineering" is a loaded phrase, but that doesn't mean it's not a worthy subject. An Annie search for the keywords "genetic engineering" gets 169 items; limiting to after 1994 reduces that to 20 (suggesting that the bound phrase is somewhat out of fashion), and an examination of that set (eliminating novels and stuff obviously about plants) produces an interesting bunch of books across a pretty broad spectrum --some more science-y, some more ethics-y, some more journalism-y. But a good start toward refining the topic to something that it makes sense to try searching in periodical databases.

Now let's take a much narrower example: head trauma and brain physiology. Annie isn't much help, even for the term "trauma" (which tends to be heavily loaded in the psychological-trauma domain, though there's a LAW journal with the title Trauma --personal injury law stuff), but that's not really surprising. In fact the general literature isn't likely to be all that useful, since the subject is really primarily of interest to medical practitioners. So how about Lexis-Nexis? Let's try "General Medical & Health Topics", go with the default "medical & Health News", set it for "previous 2 years".... gets 19 documents. #2 is an immediate BINGO: "Improving outcome after brain trauma", from Patient Care, 1997 (5731 words, an Emergency Handbook with references). We're off to a flying start.
And now something intermediate: how about hypoglycemia? Let's stay with Lexis-Nexis, stay with "Medical & Health News" and stay with last 2 years: 136 documents. We note that a lot of them are from publications with Diabetes in the title, and that might get us going in a productive direction. But let's try another tack, with AltaVista. A simple search for 'hypoglycemia' returns 12,303 documents, but we can refine that set (say, REQUIRE 'metabolism...' and REQUIRE 'adrenal...') and cut it down to 1597, then satrt looking through those.

Or go broader: use YAHOO and find 16 sites that focus the examination for us --support groups, information pages, symptoms, etc.

Or try childhood leukemia: let's use FirstSearch (or Cambridge Scienctific Abstracts) BioDigest, which nets us 57 items. #11 points to (and abstracts) a 12-page summary article on "Medical Progress: childhood leukemias" from NEJM (1995). And we're on our way.
How about "behavioral symptoms of schizophrenia"? First thing I looked at was Encyclopedia of Human Behavior (SCI REF BF31 .E5 1994), which has quite a few references in the index, and an 18-page article in vol. 4. Good start, but of course not the only (or necessarily the best) place to start.
And one where the most sensible thing seems to be to head for the specialized literature and then perhaps work back: chagas disease and blood transfusions. I did a quick search "chagas transfusion" in PubMed and got 158 items, then chose one of them and tried the "See related articles" feature and got a juicy 104. These have abstracts, and I'll have a lot more to say about effective use of abstracts next week.