Are there any reliable predictors of talk quality?

In grad school, some friends and I once attended a very bad talk by a prominent ecologist. This led to a conversation over beers about whether talk quality is negatively or positively correlated with prominence of the speaker. We decided that there’s no correlation. On the one hand, prominent people become prominent by doing really good science, and really good science is essential to a really good talk. On the other hand, prominent people are busy, and their positions are secure, which might mean that they tend to put less prep work into their talks. Indeed, if you’re sufficiently prominent, giving a bad talk comes to be viewed almost as a sort of amusing eccentricity. Bill Hamilton’s talks are fondly remembered for how poor they were (I’m told he would just read his papers, in a quiet mumble). But if you’re just Joe Schmoe, no one is going to fondly remember you if you give a bad talk.

Which leads to the broader question, are there any reliable predictors of talk quality? If there were, it would be useful because it would make planning your week at the Ecological Society of America Annual Meeting much easier. Obviously, one reliable predictor of talk quality is the quality of the speaker’s past talks. But what if you don’t know anything about the speaker’s past talks, for instance because you want to take in talks on some unfamiliar subject rather than go see people you’ve heard before? Personally, when faced with this problem I either go to see people whom I’ve heard of, on the ‘I should hear this person at least once in my life’ theory, or people whose titles or abstracts sound really interesting. But what do you go on?

In the comments, please don’t name anyone who you think gives a bad talk.