I'd also like to note that presentations are very dissimilar to conversations, and some interviews. Conversations are at least a two-directional flow of information (more directions with more people), whereas the bulk of presentations are a one-way flow of information.
When I'm in a natural conversation, the list of points I want to make is made up on the fly and is usually quite fluid (changes back and forth as I or the other people talk as a function of what is being said).
That is, talks, even if not already "written", are not the same as natural conversation at all.
I also find that really great impromptu speakers actually do speak more coherently in their natural conversations too. I assume because they're used to pre-editing what they're saying in their heads as they talk.
If natural speech were a modern programming language, prewriting would be analogous to compiling to bytecode. Then you JIT compile it again when you speak. You rarely go directly from source code to machine language, perhaps because there's so much diversity among the "machines" (humans).
Heck, I even "transpile" my talks from time to time. The last time I gave a presentation, my paper was written in English but I had to present it in Korean! Needless to say, that wasted a lot of brain cycles.
Eventually you get to the point that you know that when subject X comes up, Bob is going to deploy his well-worn chestnut that you've heard 147 times before, and you can damn-near repeat it word-for-word.
You're hanging out with the wrong people. Or something.
> when subject X comes up, Bob is going to deploy his well-worn chestnut that you've heard 147 times before,
You should stop hanging with Bob. There are other people out there who have real conversations about what's going on with them, their view of current events, their new ideas for what ought to be done, their new inspirations. . .
I think the more likely and plausible explanation for why some people sound more prepared when speaking is because... they simply are better at it. They've practiced thinking through things, and are adept at putting thought into words. They've developed the think-and-communicate muscle, through lots and lots of practice, and now they can use that muscle with greater effectiveness than your average bear.
https://www.ted.com/talks/pico_iyer_where_is_home?language=e...
And some experienced speakers will have their speaches written out and timed.
When I go to conferences and sit in the audience, I type transcripts in real-time. I somewhat agree with you, but I have some other evidence that I don't know what to do with:
http://diyhpl.us/wiki/transcripts/
I have found that some people are amazingly well spoken. Absurdly well spoken, as if they specifically chose words so that I could transcribe and type them more easily. Matt Corallo has been really easy to type:
http://diyhpl.us/wiki/transcripts/mit-bitcoin-expo-2015/matt...
For comparison, gmaxwell is more difficult to transcribe because of compressed deeply parenthetical spike bursts:
http://diyhpl.us/wiki/transcripts/gmaxwell-bitcoin-selection...
and I remember jgarzik gave me lots of trouble recently during his talk:
http://diyhpl.us/wiki/transcripts/scalingbitcoin/issues-impa...
For more comparison, here was an impromptu sit-down session with a few (high-energy) friends:
http://diyhpl.us/wiki/transcripts/scalingbitcoin/snarks/
Great results abound when making authoritative statements. I find that I sound the most elegant for hours after I am done with some 10 hour high-concentration programming session. I attribute this to paying more attention to the purpose of communication, whether to transfer knowledge or achieve some other goal, and that good writing/speech is very similar to good programming.
And I have low tolerance for communication when the epistemology is all wrong....