We are surprisingly unaware of what we're saying. The actual words, I mean - we're astonishingly good at remembering the content or the general gist of a conversation, but very bad at remembering precisely what words were used. Many an argument has been built on just such a memory failing.
Linguists are more inclined than most to notice the way people say things, as well as what they actually say (or instead of what they say, sometimes). We're familiar with the informants who, presented with a questionnaire, claim never to use construction X, and then go on to do so in the next sentence, and what's more we know that we do this ourselves. But even we are not always aware of the way we speak.
This is one of the things I teach my students, in passing. In an early dialectology seminar we discuss a list of non-standard grammatical constructions that they are supposed to have surveyed their friends' use of. One of them is the doubling of comparatives and superlatives, such as most biggest or more uglier. Every year, without fail, the students look at me in utter disbelief when I suggest that people do this a heck of a lot. I do it, all the time. I know that it's a shibboleth (a linguistic thing that marks you out as different. Or a moron, if you read the insane rantings on the Tumblr #grammar tag) but I like it. But they really don't think that anyone ever says that sort of thing. I tell them that if they listen out for it, they'll hear it, and generally I've already done it once by the time we get to that point anyway.
There was a lovely example of just this at the PG conference held at my university last week. My friend gave a talk on sentence-final like (it was canny good like). In the question period (during which I asked a question that inadvertently included a sentence-final like), she mentioned that there is also reported to be a sentence-final but. Then she said, I've never heard it but. Brilliant. Within the very sentence in which she doubted that it was common, she uttered it herself.*
*(It was an intermediate version of final but, actually, one of the ones that Jean Mulder describes as not a true sentence-final particle, but equally not simply ellipsis with a missing but-clause. But this is all for another day, another post.)