|
My scant second-language acquisition training (and I do mean scant) says that intonation and rhythm in a sentence matters a lot. Especially if it's a native or native-like pronunciation. This jives with my psycholinguistic training and a view of phonology that I was never taught in school (and which I don't actually think is a complete answer).
We tend to start building meaning and forming hypotheses about what words are coming as we listen to a word and the sentence it's in. To a large extent, language may be redundant, but we also often don't really pay attention, we just listen just for confirmation as to which posited choice is correct. If we hear 'kopee', by the time we get to the 'p' we've already ruled out all the words that don't begin with 'ko-', 'sandwich', 'popcorn', 'guinea pig'; given the context we're also ruling out a lot of irrelevant words like 'kopeck' and 'copayment'. If there's a native-like rhythm to the words and the sentence, then by the time we get to the 'p' we even know that the word's going to stop after the second syllable. In any event, if the grammar's reliable we know what grammatical category (noun, verb, etc.) to expect, and that limits the word choice. (Pitch in knowledge about word frequencies, and you have a native-speaker's ability to deal with a Cloze test.)
Joan Bybee also has a schtick about how we understand words, and it's not by listening to the sounds and mapping them onto some sort of abstract representation like Chomsky would have said. Instead we remember what a word sounds like, the whole thing, and sort of attach each token that we hear to the right meaning, so as to produce a cluster of overlapping tokens. Then we generalize over all the tokens to come up with an average. I think that's a bit oversimplified, there are too many times when we 'see' inside a token, but then again she may think so, too--I haven't read her book, just a couple of articles and some summaries. 'p' may not be a lot like 'f', but it's closer than 't'--since the word 'copy' is ruled out by context we'd be casting about for something that's 'copy'-like, not 'kotex'-like, and about the nearest thing is 'coffee' ... satisfying our real-world expectations.
Neat linguistic exercise.
|