--- In
cybalist@yahoogroups.com, "Glen Gordon" <glengordon01@...> wrote:
>
> David:
> >What about it Piotr? Can "yet", "it", "pie" and "me"
> >be analyzed as /jEt/, /?jt/, /paj/ and /mjj/?
>
> Yes it can! And thanks to the budding new-age study
> of "shadow consonants", everything can be linked to
> everything! All is One. Oooomm....
True, Glen. Here's the issue of abstractness rearing its head again.
I've kept out of this discussion so far, but since I am first and
foremost a phonologist, perhaps I should throw in my tuppennyworth.
Once we start reducing the vowel inventory in this way, what can
prevent us from reducing the whole thing ad absurdum? Why assume that
/a/ _must_ be a vowel in Sanskrit? Why not introduce a mystery glide
(belonging in the same class with <y> and <v>) which is realised as
[a] when syllabic and which can also be employed to do other useful
work (like making odd syllabifications predictable) in its capacity as
a shadow consonantal segment? It's possible to get rid of _all_ vowels
in this way.
To note that [j] and [i] are regularly related via phonological rules
is one thing. To claim that _every_ [i], also in non-alternating
morphemes, is therefore derived from underlying /j/ is to employ the
"free ride" principle: what works when motivated must also work when
unmotivated. Chomsky and Halle, who pioneered that approach in _The
Sound Pattern of English_, analysed _all_ instances of English [aI] as
being derived via a _synchchronic_ rule of Vowel Shift from underlying
/i:/. The failure of the expected operation of Trisyllabic Laxing in
words like <nightingale> (not "nittingale", unlike <divinity> with a
lax vowel) had then to be explained by introducing a shadow underlying
fricative /x/ and ordering the rules so that the loss of /x/ with
compensatory lengtghening was ordered after the Trisyllabic Laxing
rule but before the Vowel Shift. And so on, and so forth, making
English phonology unspeakably complex but fearfully symmetrical.
This is how phonology was done in 1968 and for a few years afterwards.
At present, free rides and shadow segments are banned from most
models, and modern phonologies tend to disfavour excessive
abstractness. The segmental feature [syllabic] is no longer in vogue
either. "To be syllabic" is shorthand for "to occupy a nuclear
position in the structure of the syllable". If we have a language in
which [i] and [j] are positional realisations of the same underlier,
there's little point in marking that underlying phoneme as [+ cons],
thus enforcing its membership in a selected major class by mere
stipulation (would anyone do that for French, for example?).
Assuming for the sake of the argument that [i] and [j] (<y>) never
contrast in Sanskrit, how does one prove that an analysis which makes
[j] a non-syllabic allophone of /i/ is less highly valued? At the
level of systematic phonetics, Sanskrit has quite a few vowels
(definitely including /i/ and /u/) -- _real_ vowels, no doubt about
that. That is what determines its place in the typology of vowel
inventories, whatever one's favourite abstract analysis of Sanskrit
morphophonology.
Piotr