LINGUIST List 3.494

Mon 15 Jun 1992

Disc: Natural Phonology, English Stress, Macintalk

Editor for this issue: <>

Directory

1. , The last gasp of Natural Phonology
2. , English Stress
3. , Software, Good

As one of the few surviving Natural Phonologists left on the North
American continent let me say a few words in its defence. Unfortunately
I'm off today to the Phonologie-/Morphologietagung in Krems, Austria,
and consequently have little time to discuss issues in depth.
I appreciate Joe Stemberger's comments about the child language data,
but Stampe never argued (unless he wants to argue here himself) that
WHICH processes a child would use to solve a particular articulatory
or acoustic puzzle was predictable, since the vocal tract offers
alternatives. If it didn't, all languages would end up the same,
which they clearly don't. The difference between predictable choices
and optional ones is considered at length under the heading `motivation'
in the Cognitive Grammar literature, which I have argued in a forthcoming
paper is compatible with NP.
But the fact that children (and adult second language learners, and
people talking at the same time they eat and...) come up with similar
replacements that they have never been exposed to in the environment
but are explicable articulatorily/acoustically (and that's all that
NP means by `process') is not the only valuable insight that has
been lost with the demise of NP as a viable theory.
In his classic paper `Yes, Virginia..' Stampe showed that phonemes
and morphophonemes were not just structurally defined points in an
abstract pattern but represented facts about the nature of phonological
perception and storage. It is a theory which makes specific predictions
about abnormal language behavior (especially early second language
acquisition stages), and is the only theorythat has any interest
in any level of representation below the level of the surface
phoneme (most discussions of post-lexical rules do not include
ANY of the interesting (inneressin) things that happen in real
speech production, even though they are rule governed and language
specific.
I could go on, but I realize that the interests of phonologists have
moved on to other issues, some of which, such as the geometry of
features and the parameters of stress theory are also directly
useable by we embattled few.
There will be a book coming out later this year on recent research
in Natural Phonology, edited by Bernhard Hurch and Rich Rhodes.
Contact either of them, or me for further information.
(Incidentally, if someone could store any list answers to my
ranting here over the next month, it would be greatly appreciated. Or
is there some way to snoop in the archives?)

Geoff Nathan <ga3662siucvmb.siu.edu>
 Southern Illinois University at Carbondale

Some time ago there was a discussion of English stress on LINGUIST.
All those who took part seemed to agree about the following (this
is with reference to Northern, white US usage only):

 Words with zero stress on their last three syllables are
 only possible in case we are dealing with inherently unstressed
 suffixes, e.g., admiralty.

However, it would seem that some (many?) speakers also have
such pronunciations in the case of words like in -ery, e.g.,
dysentery, stationery (as opposed to words in -ory or -ary,
which always have a secondary stress on the penult).

While it is obviously possible to come up with various excuses
for such forms (e.g., treating the final vowel as "underlying"
/y/), it still seems interesting that this apparently exceptionless
generalization may have exceptions after all.

In response to Zev Bar-Lev's request for information on speech synthesis
(LINGUIST 3.479.2), here's what I know:

 Macintalk is "not supported"; this means Apple doesn't care whether you
 can get it working or not. However, it does work with HyperCard, on
 System 6.x - but *not* with System 7, although I have heard rumors that
 a version now being supplied with Talking Moose (q.v.) has been hacked
 to work with Sys7.

 The way you get it working with HyperCard - and incidentally a nice
 tutorial on how to use it - is to use HyperMacInTalk, a stack with
 XCMDs and XFCNs by Dennis C. Demars. It's widely available on the
 network for anonymous ftp (for instance, try ftp'ing archive.umich.edu)
 and it's quite nicely done. You can install the relevant XCMDs in
 your stacks and thus run it any way you want.

 That's what I did in my "World of Words" stacks that I showed at the
 LSA Software Exhibit in January. It is possible, though not easy,
 and not altogether satisfactory, to force Macintalk to sound like it's
 speaking another language. Oddly, it's easier to make it speak Homeric
 Greek than Old English (it doesn't do front rounded vowels, though it
 does have a /x/).

 Now for the bad news. Apple has been busily developing a "text-to-speech"
 manager for Sys7. I expect to hear more about it in a little bit, but
 my initial guess is that it's what the name implies - i., useless for
 linguists because it's based on somebody's phonemic analysis of English
 (which always ends up with English having precisely 64 "phonemes" - t he
 term should be taken ironically, because they're exactly *not* what any
 linguist would mean by "phoneme"), without phonetic hooks to (e.g.)
 lengthen vowels, raise pitch, compress attack, partially devoice, etc.

 When I've heard more, I'll post it here.

I response to Barbara Abbott's inquiry about "that Adj of NP" (ibid.4),
I've been hearing this all my life. The construction derives from the
more conventional count versio:
 How good (of) a doctor is he?
 I didn't know he was that good (of) a doctor.

The problem is that with a mass noun, one can't use the indefinite
article, and the "of" becomes obligatory, though still awkward.
But the sense is clear enough, and it's useful. So it's used.

Cheers,
-j