I can add a couple of thoughts from the academic perspective (I'm a social/behavioral scientist by training and a professor of public health):

1) Storing articles as reference or as project support: my advice would be to invest in a computerized reference manager (I personally use Endnote, but there are lots of them out there) and make that your reference material filing system for journal articles. I would either store the articles electronically as PDFs or in a paper filing system, but I'd make them reference rather than project materials. The reason for that is that as you develop a program of research you'll find that you'll refer to and use the same articles over and over again across specific projects, so having them centralized in a way that isn't tied to a particular project makes sense. When I need to read (or reread) an article for a project I'm working on, I make a next action for it (e.g., "read Smith 2012 and incorporate ideas about XXX into introduction to YYY manuscript"). I usually do my reading on screen with PDF, so I just open endnote and access the PDF from there when I'm ready to do the action; if you prefer paper copies you could move the article from your reference file to your action support file when you create the next action, then re-file it in reference when you're done. The computerized reference managers also have some nice advantages for writing papers -- many have the ability to automatically insert and format citations for you in word processors. Over time, this is an incredible time savings for you (especially for those of us who work in interdisciplinary areas -- depending on which journal I'm submitting to, I may need to use one of three different citation styles).

2) Project defining and support materials for research studies: there's a lot about how you handle project support, etc. for research studies that is just like how you'd handle it for any other project. A couple of specific things you might want to think about -- I find it useful to conceptualize study design, data collection, initial statistical analysis, and writing as four separate and distinct projects. There are a few reasons for this. First, when I'm designing a study I often don't know what the ultimate outcome will be -- it might end up in a manuscript on its own as a single study paper, it might get packaged with other studies in a multistudy paper, if nothing pans out it might end up in a file drawer, etc., so I can't really define the publication/presentation endpoint at the beginning. Second, at least for the kind of work I do there might be a lag between when we finalize the study design and when we start data collection (given personnel power, lab space, and financial constraints), so I find it cleaner to define them as separate projects. Finally, there's a nice sense of accomplishment in finishing off projects; if everything from initial design to publication is a single project, that sense of accomplishment can sometimes be years away (at least in my field -- in others the time lag is shorter).

Hopefully this is helpful -- good luck setting up and using your system.

Best,
Marc