Our one and only, not just President of TYCA National, but TYCA-PNW member, Sharon Mitchler(see her blog! has a nice overview of some research on responding to student writing. In short, it's often more art than science, but it's always nice to have the discipline's scholarship bear that out. I wish I brought Teaching English in the Two-Year College (May 2006 edition) with me so I could accurately quote what she said, but here's the gist of things that resonated with me. If you are a TETYC subscriber, you can get it online by following the preceeding link.
Deductive, open-ended comments and questions as well as inductive, more directed questions, tend to work well alone and in tandem. The deductive (and I hope I'm using these terms as she did) led to slower progress initially, but solid long term student progress. The more directed comments led to quicker progress, but it too was sustained over the long haul, but maybe not so well as the inductive comments.
I tend to use both sorts of comments, scribbling notes in the margins, sometimes directions as well, along with some mechanical and grammatical markings, but I don't even try to find all the "errors" since most student work needs a focus on big picture rhetorical concerns. I also write an end note, type actually, so they can read it (unlike my marginal comments, by which I mean in the marginal, but we might also speculate to their marginal value--har, har). I limit my end note, most of the time, to the three most pertinent global concenrs, such as the need to make a worthwhile point, provide specific evidence, and to tie it all together. Sound familiar?
In a recent forum with a candidate to be our new dean of Arts and Letters or some such to be renamed division, he pointed out that at the University of Virginia, responses to students were to include the marginal comments, a typed end-note and then a workshopping of at least one paper per student, by which I take to mean the whole classes spends a period discussing a student work or two. I don't do that but it sounded like a good idea and this combination is, he said, supported by research as being effective. Well, I'm hitting on two out of three, at least in what I do, but how well, that's another thing altogether.