I do find I reason differently when Ithink in French, because the sentence sturcture leads me to things a different way round.

Yeah, this and things like the experimentally verified necessity of high level programming languages and mathematical notation (ever tried doing arithmetic with roman numerals?) convince me that there's something sapir-whorfish going on somewhere. I'm presuming therefore that the argument against sapir-whorf takes a form similar to "we create language, and are not created by it" - asserting that where we want to reason in a way that our current language doesn't support, we'll just extend our language or invent a new one. So the correlation is there, but the causal relationship is reversed...