This has already been linked six ways from Sunday, but Nature magazine published an article in which they probed 42 articles from both Wikipedia and the Encyclopedia Britannica for correctness. The results? They came out roughly even: Wikipedia contained four errors per article compared to Britannica's three.
It's also worth noting (thanks Slashdot) that "Wikipedia articles in the sample were, on average, 2.6 times longer than Britannica's — meaning Wikipedia has an error rate far less than Britannica's."
I've always found it ridiculous to assume that a traditional encyclopedia would be more accurate and less biased than Wikipedia. After all, the Encyclopedia Britannica is still written by human authors. It is written by fewer of them, in fact, so in theory there should be a greater chance for errors and bias to creep in. But, I'm very used to the whole "many eyes make all bugs shallow" doctrine. For everyone outside of the OSS world, it's still a novel concept.
I really think this is the argument that needs to be pressed: just because Britannica claims to be "authoritative," it does not follow that it their articles are more accurate. And if you understand that there you risk inaccuracy whether you go to the Cathedral or the Bazaar, then the secondary benefits of Wikipedia — timeliness, depth of information, breadth of subject, continuous evolution, availability and price — become more convincing.
I suppose one could argue (as the E.B. did) that Wikipedia is subject to vandalism in a way that a traditional encyclopedia is not. And while it clearly can happen, it doesn't seem to be a frequent problem. But I'll admit that I don't have any data to back that up.