On 2006-09-17 02:06, Petr Hrubis wrote:
> Hence, quite logically I think, it is the following nasal and the
> following voiceless consonant that cause SHORTENING, not the other
> way round.
It isn't as simple as that. What is the "normal" length of an English
vowel? You can't assume that the final position is neutral in English
and that what you find there is "normal". The end of a phonological
phrase (including the case when a word is pronounced in isolation) is
the locus of important intonational signals, so a vowel in this position
(especially a stressed one) gets some extra length in order to be able
to support a tonal contour. Even reduced vowels (like the final schwa
in <sofa>) are longer word-finally than they are in any other position.
Moreover, some vowels can't occur word-finally at all -- I mean the lax
(or inherently short) ones, like the vowels of <set>, <cut>, <put>, and
in many accents the vowels of <lot>, <bat> and <bit>. Even those who use
lax /I/ rather than tense /i/ in the second syllable of <city> (as in
conservative RP) make it rather long; and in many accents of English
word-final tensing is obligatory.
If you want to compare like with like, you have to look at vowel length
in preconsonatal positions (where all English vowel phonemes are free to
occur). What we can say securely is that the realisation of one and the
same vowel phoneme is LONGER before a voiced (lenis) obstruent and
SHORTER before a voiceless (fortis) one. Nobody can seriously dispute
this statement, since it's based on the findings of innumerable
experimental studies.
The question is whether we are dealing with pre-lenis lengthenig or
pre-fortis clipping? The answer may be: both. Note that vowel length
before sonorant consonants (liquids and nasals) is almost precisely
intermediate between the durations found in pre-lenis and pre-fortis
contexts. Sonorants are phonetically voiced, but their voicing is
phonologically non-distictive. RELATIVE to the pre-sonorant position,
distinctively voiced consonants lengthen vowels, whereas distinctively
voiceless ones shorten them. As a result, the vowel of <bid> is on the
average nearly twice as long as the vowel of <bit> (and may be
physically longer than the vowel of <beat>), while the vowel of <bin> is
somewhere in between.
To be sure, there are accent-specific preferences, special cases (like
/æ/, which seems to be undergoing a phonemic split into a lax phoneme
and a tense one in some varieties of English) and the complicating
influence of other factors such as vowel height (other things being
equal, low vowels are longer than high vowels, but this is more true of
some accents than of others). In Scots, both tense and lax vowels are
phonetically short except in root-final positions and before /r/ or one
of the voiced fricatives /v, ð, z/, where tense vowels get lengthened
(Aitken's Law)
Piotr