I start by reading- preferably about something related to what I want to write about. And I research, which means downloading academic, peer reviewed articles- I actually read some of them. Then I scribble little ideas on scraps of paper and increasingly on my iPod touch notepad. Then I clean the bathroom. I also try to read things for fun that aren't as stiff or clunky as a lot of academic writing. Adventure novels, feature journalism from Rolling Stone or The New Yorker or the English version of Der Spiegel, popular writing about your topic all help you to pick up good habits of writing. Good sports journalism is often both dynamic and clear, so it helps set you a good example of how to present your ideas.
Two days later, having thought about my idea, I go to lunch and chat about what I am thinking about to someone who is interested in the subject. After lunch, and a nap, I try to reduce my idea to a single sentence based on my conversation. Talking to people about your idea is a great way to develop it. Try talking to a professor or helpful grad student about it. Try explaining it to a friend who has no idea about it- that will help your clarity. That single sentence is what academics refer to as a thesis statement.
That sentence expands out into an introduction, with the idea in mind that each sentence represents a section in the final paper. So the first paragraph of your paper should have an opening sentence that sums up your idea in one shot, and about 8 to 10 sentences that support your main sentence.
Then I write a paragraph using each sentence as the introduction. This should take no more than half an hour per section, even if you write very slowly. Here is a rule of thumb calculation:
10 page essay =
1page introduction this should encapsulate your whole paper(Tell'em what yer gonna tell'em)Think of each section as a question and you have to answer it in a page or so.
8 pages- 1 page per section(Tell'em)
1page - Conclusion(Tell'em what you told'em)
I keep feeding the beast until it is as bloated as a goose's liver. I then take it for a brisk walk, and in 20 minutes for a 9 pager, I pretend I am trying to explain the paper to friend who is less than entranced with the idea. I record this onto my iPod, and play it back, transcribing and editing as necessary. Usually, this is pretty easy.
At this point I have a pretty coherent, well thought-out paper of about 8-10 pages.
Some of these stages I may repeat, but the idea is to break up the process into small manageable chunks. Each section, for example is only about 1 to 2 pages long, so it is easier to write section by section rather than to try and write the whole thing in one shot.
You can also try composing your paper orally into a recording device immediately after or during your research. Then you have a first draft to work from. But using the method I outlined works better if you are dealing with really new, unfamiliar information.