Re: Questions about generative phonology

Hey.

On 1/13/06, Peter Bleackley <Peter.Bleackley@...> wrote:
> With reference to the chart at
>
> http://www.ctlwmp.cityu.edu.hk/lingintro/english/sounds/phonology/features/feature-interact.htm
>
> 1) How would [4] be represented?
It depends on its phonological behavior. With respect to the feature
[±continuant], in my dissertation on Shoshoni phonology, I gave this
sound the feature [+cont], since it is the realization of /t/ between
vowels, where the preceding vowel is [+back]. (When the preceding
vowel is [-back], /t/ is realized as [ð].) This parallels the other
voiceless non-continuants, each of which have continuant realizations
between vowels: /p/ -> [β], /ts/ -> [z] (Goshute has /tθ/ -> [ð]
here), /k/ -> [ɣ], /kʷ/ -> [ɣʷ]. So the complete feature specification
for [ɾ] is something like

[+cons]
[-son]
[-syll]
[+voice]
[+cont]
[cor]
[+ant]
[-dist]
[-str]

(BTW, I don't agree with the chart WRT the feature [distributed];
dental fricatives are [+dist] and alveolar fricatives are [-dist]. The
website you linked to has the right definition of [distributed], but
unless the pronunciation of [θ, ð] and [s, z] is very different in
Cantonese than in English, he seems to be wrong about the feature
assignment.)

> 2) How does generative phonology deal with diphongs?
In classical generative phonology, diphthongs were usually represented
as sequences of two adjacent vowels, each with a distinct feature
matrix. In contemporary versions of generative phonology, feature
matrices are linked to prosodic positions; when a feature matrix is
linked to more than one prosodic position, the segment is realized
phonetically with additional length (i.e., a geminate). When more than
feature matrix is linked to a single prosodic position, the result is
a so-called "contour segment". Affricates are a good example of
contour segments, being both [-cont] and [+cont] (this is explicitly
acknowledged in the IPA representation of affricates as digraphs). A
potential contrast of long and short diphthongs can then be realized
as two feature matrices linked to either two prosodic positions (long)
or one (short).

> Would either of these require additional features not on the chart?
No.

> 3) Am I correct in thinking that [x] is [+str][-sib]?
Generally, only one of these features is used. Ladefoged's Course in
Phonetics defines  sibilant as "A speech sound in which there is
high-pitched, turbulent noise, as in English [s] and [ʃ] in sip and
ship."

Chomsky and Halle (1968) define strident sounds and those sounds which
are "marked acoustically by greater noisiness than their nonstrident
counterparts."

Both definitions cover about the same ground, but Ladefoged's
definition seems to be absolute, but C & H's relative (they give [ɸ,
β] as examples of nonstridents compared with [f, v]).

I would not define [x] as [+str], but again, it will depend on the
language in question. In Shoshoni, there are two segments which are
underlyingly strident: /ts/ and /s/. In the Goshute dialect, /ts/ is
replaced by /tθ/, leaving only one strident. So in Shoshoni, I don't
need both [strident] and [sibilant] and chose [strident] arbitrarily
as the name of the feature expressing this acoustic property.

Dirk
--
Gmail Warning: Watch the reply-to!