Here are the draft results of Peter Chen's podcasting and videoblogging study. Again, while they're interesting, it's hard to say how representative they are, since the participants (me included) were largely self-selected. Chen writes that "The overall margin of error is therefore in the order of 5.08% at the 95% confidence level for the primary survey."
My memory of undergraduate statistics isn't good, so I'll take his word for it. But I'd like to remind people that there is a 5% chance that these results could be way, way off (in several unknowable directions) for the real world. Which is true for many surveys, of course.
In any case, make sure you look at the numbers carefully. Chen lays them out well, and has plenty of wise caveats, but when you're dealing with percentages of very small sample numbers, it's quite possible for some of the results to be skewed. Many of them certainly feel right, for whatever that's worth.