Re: PHONO: feature theory (was: vowel harmony)

Jonathan Knibb wrote:

> Question 1: in standard feature theory, is a phoneme uniquely
> associated with a particular set of features *no matter which language
> it's in*?
The short answer is "it depends", very helpful :-)  Mainly it depends on the
number of contrasts; the more contrasts, the more features are needed.  It
could also depend on which feature system you use-- acoustic
(Jakobson-Fant-Halle) or articulatory (Chomsky-Halle), or some eclectic mix
of the two, which sometimes works (e.g. Chomsky's [coronal] nicely
classifies the dentals/alveolars and alveo-palatals).  For vowels, the
minimum is 2 features, one for front/back, another for high/low; that would
work for a typical 3 or 4 vowel system:
i [-back, +hi]
e [-back, -hi] (it could be phonetic [e,E,æ])
u [+back, +hi]
a [+back, -hi] (it could be phonetic [o, O, Q]-- if it's [a,A] then you'd
need a "redundancy rule" specifying that [+back, -hi] > [-round]

Note that it could also characterize an unnatural system with e æ o O

The feature [back] is preferred over [front] because back vowels can be
defined as intrinsically rounded.

The more vowel height contrasts, the more features you need, though [high]
and [low] covers 3 heights [+hi], [-hi, -lo] = mid vowels and [+lo] -- that
describes a system with i -- e -- E or æ,  or i--e or E --æ and so on.
English needs a 3d feature [tense] to cover i/I, e/E

To some extent then one would use the same features to describe numerically
identical vowel systems (i.e. all 5 vowel systems can use the same features,
even though the phonetic realization of the phonemes may (probably do)
differ.
>
> Question 1a: is it meaningful to speak of 'the same phoneme' cross-
> linguistically anyway?
Only in the sense of comparing them rather informally:  "English /i/  is
realized differently from French /i/" or perhaps "The phonemes /i e a u o/
are practically the same in Spanish and Indonesian".  or "Both English and
Indonesian have a phoneme /@/, but it is realized differently in each"
>
> Question 2: how do you decide which value of a feature is unmarked?
Well (a) no contradictory features; if a vowel is [+hi] it cannot also be
[+lo]; then there's conflicting opinion as to whether _all_ features must be
specified, vs. only those that are contrastive, vs. those that can be
implied (i.e. [+cons, +nasal] ordinarily need not be specified [+voiced] )

Hope this helps; I'm a little rusty, and I'm not even sure this system is
much in use anymore...................(too bad, if so, since it captures a
lot of interesting generalizations, even if it is cumbersome to use)
>
 If I could make up my own features,
> it would make things a *lot* easier. :))
>
Do it!  You might revolutionize phonology yet again :-))