I don't think anyone knows why metathesis or any other linguistic change occurs. The best we can do is to say with some precision what has happened, not why it happened.
A phoneme is indeed a sound unit. Change the phoneme and you have the possibility of changing the word. (For example in word change the phoneme /d/ to the phoneme /k/ and you get work instead of word.) But phonemes are not exactly the same as sounds because different languages divide up the possible sounds available into different phonemes. For example, /l/ and /r/ are different phonemes in English but the same phoneme in Japanese, hence their problems with River Valley Road. If you listen carefully, you will discover that the English /l/ is pronounced differently at the beginning and end of syllables. I'm told that in Welsh each of these realisations of the English phoneme (called allophones) is a different phoneme and can occur in any position.
I hope the above isn't too addled for you to understand. (Sorry wseiber, not a word non-native speakers are likely to come across very often.)
Bingley
