I have found entirely the opposite; it's very strongly correlated with spelling ability - or so it seems from my necessarily few observations, of course. I know some excellent mathematicians who write very stilted prose, and a few make more grammatical errors than I'd have expected, but they can all at least spell well.