Larry over at Sandwalk recently addressed complaints that Richard Dawkins' new anthology of great science writing only includes writing by scientists. Larry has some good comments, and suggests it's appropriate to have an anthology of writing exclusively by scientists. He also lays out what the top three criteria of good science writing should be:

I maintain that the top three criteria for good science writing are: 1) accuracy, 2) accuracy, and 3) accuracy. Everything else is much less important.

I agree, maybe phrasing it a little differently: accuracy is the first rule, but if you can't meet the first rule, there is absolutely no point in getting to any other rules. If accuracy is a critical foundation, other elements are just as critical when writing for a general audience, once you've got accuracy down. When writing for a general audience, scientists fail if they don't write accessibly and engagingly. You need to be accurate when conveying science to the public, but you also have to be sure you actually convey your accurate information. A creatively written science piece that's completely inaccurate insidiously undermines science literacy, but a perfectly accurate but impenetrable piece can turn off people's interest in science. (I'm not making any claims about my own writing, but hey, I try.) The physicist Richard Feynman described science as creativity in a straightjacket, and science writing should be the same. It's tough to write accessibly, engagingly, and precisely. Precision is hard to achieve, and lack of it is probably the most common problem in mediocre science journalism - not outright errors, but a sloppiness with terms and concepts that creates a false understanding of the science. This is what makes science writing so hard, but it's also what makes reading good science writing so rewarding.