What difference will voice recognition software make to writing fiction?

I've just had a PhD student successfully defend her thesis. It was nearly all written using voice recognition software. It actually made very little different to the style of the text, but then there are pretty rigid conventions about how a thesis should be written. It's got me thinking though as to whether 'writing' a novel by talking it out aloud, rather than typing it in, or writing it down, would make a difference to the prose, narrative, style, etc? Would it push the style to tell rather than show? Would it increase the amount of dialogue, or even the quality of the dialogue? Would it lead to narratives that are more like verbal storytelling? Sometimes when I read a novel, the quality of the voice is so good, I think that the writer must have read it into a dictaphone and then transcribed it. I've no idea whether they did or not, or whether they were just very good at capturing the cadence, pauses, half-sentences, etc, of verbal storytelling. Would it lead to shorter, tighter stories? I'm not sure how I'd get on composing a story entirely through the spoken word. I reckon I need to think and type. I think I'd also feel pretty odd sitting in a room talking to myself! Anybody got any views or experience? How would talking to compose a story affect the story?