On 2008-02-15 00:57, Roger Mills wrote:
> I'm sure you're aware of this, but others may not be: PAN *e represents
> central schwa, not a mid/low front vowel. (At least, last I heard...) Its
> use dates from days of the typewriter, when inserting a real schwa into your
> ms. was laborious and prone to omission, no longer a problem thanks to
> computers, except in email. But the system was indeed asymmetric, and *e was
> somewhat restricted as to occurrence in the proto-language, probably was
> always unstressed, and shows many peculiarites in later development.
In the Eskimo there are similar restrictions on the use of the central
vowel (it may be true of Shalishan as well, but I know too little about
them to be sure). I suppose the problem with the "/a, i, u/ + schwa"
system is that it's highly symmetrical in articulatory terms (and so
speaker-friendly), but suboptimal from the auditory point of view
(there's a cross-linguistic preference for peripheral vowels, so central
vowels tend to be avoided in small vowel inventories).
/e, o, i, u/ violates another important preference -- there should be an
/a/ in the inventory. All the three extreme vowel qualities are utilised
in most systems, and in particular it's very unusual for a language to
miss /a/ or /i/. A high /u/ is less essential and may be replaced by
another rounded, perhaps because the three basic vowels are primarily
characterised as [+low] (A), [+high, (-back)] (I) and [(-low), +back]
(U). Among small-to-medium-sized vowel systems there are two "islands of
stability": /a, i, u/ and /a, e, o, i, u/. Four-vowel inventories are
inherently "defective" and tend to increase the number of phonemes or to
assign some kind of special status to their schwa-like member, making it
little more than an allophone of zero.
Piotr