"syllabicity"

[Rich Alderson]

"Patrick C. Ryan" <proto-language at email.msn.com> wrote, on Tue, 27 Apr 1999:

>First, let me tell you that I appreciate your taking the time to write a
>magisterial summary of the questions involved.

Hardly magisterial, indeed barely satisfactory, but thank you.  John Lawler at
Michigan used to have a copy of a post of mine on Natural Phonology on his web
site at UMich; I tried to find it before nattering on.

>Rich writes:

>> In structuralist terms, two phones in complementary distribution *must* be,
>> cannot *not* be, allophones of a single phoneme.  (Although a lemma
>> requiring something called "phonetic similarity" was inserted into the
>> theory when it was pointed out that in a pure framework, the English phones
>> [h] and [N], as in _hang_ [h&N], must be allophones of a single phoneme...)
>> Therefore, in the prevailing structuralist framework of the 1940s, Lehmann
>> *had* to define *i and *u as allophones respectively of *y and *w.

>Pat writes:

>Lehmann was under no obligation to be consistently structuralist, and your
>assumption that he was is pure conjecture. By "syllabicity", Lehmann
>indicated that he was quite willing to strike out on uncharted paths. If the
>evidence had indicated anything different, I am positive Lehmann would have
>embraced the position it made mandatory.

One of us has obviously missed something here.  Lehmann's work is *very much*
"consistently structuralist"--that was the point I was making.  Were he not so
consistent, he would not have been forced to make the incorrect claim about
*i *u vs. *y *w.

>Leo writes:

>>>> just as PIE syllabic [M N L R] were allophones of /m n l r/.

>Pat writes:

>>>The syllabic status of [M/N/L/R] is a totally unrelated matter. These become
>>>syllabic when deprived of the stress-accent.

>Rich writes:

>> So the fact that all *six* resonants pattern the same is irrelevant?

>Pat writes:

>In my opinion, it is a mistake to include [Y/W] among the resonants.
>Phonologically, [j] is the voiced palato-dorsal fricative; [w] is the voiced
>bilabial fricative. *And they do not pattern the same*.

If you wish to maintain that stance, we have nothing further to discuss.  The
patterning of all six resonants as a class is a well-established fact of Indo-
European linguistics.

Further, [j] is phonologically *nothing*, although *phonetically* it is (or may
be) a "voiced palato-dorsal fricative".  Its exact phonetic characteristics are
irrelevant to the placement of *y within the phonological system of PIE (under-
stood as a set of oppositions), just as the exact character of *d (plain voiced
stop or glottalic egressive stop) is irrelevant.  (The same thing is true,
_mutatis mutandis_, of [w] vs. *w.)  Thus, to argue against the patterning of
the resonants on the basis of its possible phonetic interpretation is to miss a
very big point.

>Rich writes:

>> That *ey/oy/i parallels *en/on/.n by accident?  Then you disagree with
>> Lehmann?  What of his god-like status?  Never mind, rhetorical questions.

>Pat writes:

>I do not consider Lehmann god-like although I do believe that most people on
>their best days will not equal what has has written on his worst. I also do
>not shrink from disagreeing with his written opinions but, in view of his
>sagacity, I do so with great caution.

>And I reject the idea totally that *ey/oy/i and *ew/ow/u parallel *en/on/n{.}.

And as I noted above, we then have absolutely nothing to say to one another.

>> ... the only way to demonstrate that an analysis is valid is for it to
>> explain not only historical but synchronic phenomena in more than one
>> language).

>Pat writes:

>You surely would include diachronic phenomena, would you not?

For "historical" in my original wording, I expected everyone to understand
"diachronic"; I'm sorry that that was not clear.

>Rich continues:

>> Natural Phonology is, as well as process-oriented, constraint-oriented and
>> hierarchical:  The presence of certain phonological entities entails the
>> presence of others.  Thus, vowel systems are constrained:  Certain kinds of
>> vowel system are more stable than others, and unstable vowel systems rapidly
>> turn into stable systems by either eliminating contrast or by adding
>> contrast.  In addition, processes which are not repressed may increase
>> distinctions between vowels in the system (long vowels may become tense, for
>> example, or a distinction in palatality vs. labiality may arise as in Arabic
>> short /a/ vs. long /a:/ = [&] vs. [O:]).

>Pat writes:

>Although this is not really an argument against the point you are making,
>Arabic long /a:/ does not become [o:]; this is reserved for reductions of
>/aw/.

What happened was this:

1.  An original three-vowel system /i u a/, with length, developed allophonic
variants [& O:] of /a/ under lengthening processes.  Cowgill argued this as the
source of Brugmann's Law in Sanskrit in a paper presented at the LSA in the
early 70's; it solved the *e/*o problem for me, so I have adopted it.

2.  New compensatory length oppositions arose with the loss of various obstru-
ents in different environments.

3.  Next to certain laryngeals (*H_2 and *H_4), /a/ developed a third allophone
[a]; next to *H_3, /a/ also became [O].  Thus, a phonemic opposition arose.

4.  A system /i & a O u/ is unstable, so the process of raising evened out the
oppositions to /i e a o u/.

>Pat writes:

>IMHO, this is incorrect. If we accept Trask's definition of a phoneme as "the
>smallest unit which can make a difference in meaning" and restrict "meaning"
>to "semantic difference" vs. grammatical difference, then a language in which
>CaC, CeC, CiC, CoC, CuC, etc. represent different grammatical stems of a root
>CVC, which has *one*, meaning, then the "syllabicity" in the root makes no
>difference, and hence cannot be considered "phonemic".

But this is the very point I was making:  The definition you cite from Trask is
structuralist, rather than psychological, and not the definition of the phoneme
used by Natural Phonology.  Further, even in a structuralist definition, one is
not allowed to restrict the word "meaning" as you wish to do, and so your argu-
ment for a "non-phonemic vowel" falls apart.

>But, why all the fuss about monosyllabicity when Sanskrit provides us with the
>next logical outcome of a language that, at an earlier stage, was monovocalic
>(at least, phonemically).

>Anything other than <Ca> in Sanskrit is a result of <Ca> + <H>, <w>, or <y>,
>or a combination thereof. That is why Sanskrit does not bother to indicate an
><Ca> in its writing system (only <C>). Only combinations of <Ca> + <?> *need*
>to be indicated.

Sanskrit was never monovocalic, phonologically speaking.  There is more than
one source, for example, of [e:]--see, for example, _dive dive_ "from day to
day", where the first _dive_ is the expected sandhi variant of the ablative
_divas_ "from (a) day".  Thus, again, your analysis fails to explain the facts.

								Rich Alderson