First of all, it's too scary to present without anything written at all. Mark, good for you, but this is going to be beyond most of us. There's always the possibility that you may dry up or space out in the middle of your talk, and you've got to have something in front of you to help out. HOWEVER: Don't bring an entire paper that you're planning to have published somewhere. Prepare a reading script instead. I always do this now; a reading script is different than a full-blown scholarly treatment, in that it's shorter, hopefully clearer, and leaves out subsidiary and supporting material that is inessential for oral presentation. I've learned that for a 20-25 minute presentation, a script of 10-15 pages is ample.I agree with this absolutely. I should add here that I am not personally in favour of presenting "without anything written at all". I think it is worth taking a script, as a kind of security blanket, or to take a card with some headings and some key-words on it, or whatever prompt might be useful. For myself, I take written bits and bobs and have them ready in case needed, but aim to be familiar enough with the material and the structure of what I want to say that I will only refer to those if necessary.
The reason for my bringing this up in the first place is the idea that there is something necessarily superior, something more academically appropriate about reading a paper. This is the thing that I am struggling with. What I remain bothered about is the disjunction between the way we all work in our day-to-day lecturing and the way we behave at conferences. What I suspect is that one has simply not caught up with the other. We are in the habit of reading out papers not because we have thought it through and have decided that it is the best way to communicate with other scholars but because it is an academic convention, something we have all inherited, that we assume is the way to do things without question. A hundred years ago, I'd bet you'd hear most of your undergraduate lectures read-out by your lecturers. Under such circumstances, it was natural to read-out conference papers too. But now none of us read out lectures to undergraduates, do we? So this is how I am beginning to see it, with apologies if I am overstating the case. If we are comfortable lecturing ex tempore on a day-by-day basis, over a much longer time period (hour long lectures), on topics that are not always intimately related to our research (we all lecture on stuff we have never written on), to people less patient than scholars (undergraduate students), then surely it is a more straightforward thing to talk for 20-25 minutes to friendly faces on material we know intimately?