Re: NATLANG: Vowel harmony rules?

On 18 Jun 2004 David Peterson <ThatBlueCat@A...> wrote:

> What I mean is for "bil", for example, you get "bilir", but not *bilmik.
  That is why I introduced a new type of uderspecified vowel. For
details, see below.

> There doesn't seem to be any obvious reason why this should be.   For
> that reason, I think it would be simpler to have two stipulations (well,
> three):
>
> (1) If a verb stem ends in a vowel, it will take /-r/ in the Aorist.
> (2) If a verb stem ends in a consonant:
>      (a) It will take /-Ar/ if it's of Class A, and is not specified further
>      (b) Else, it will take /-Ir/
  What is the reason that you suppose a common underlying morpheme
<-Ar> for case 2/a? You could write instead:

 (2) If a verb stem ends in a consonant:
      (a) if it's of Class A:
         (i)  it will take <-ar> if the stem is back and
         (ii) it will take <-er> if the stem is front.

  I think the only reason for defining "underspecified" vowels as "A"
or "I" that we are about to reduce these long definitions. We "save" 2
subrules with "A" and 4 subrules with "I". In line with this the rules
are transferred to the phototactics and lexical attributes of the
morpheme (note in Hungarian vowel harmony is driven both the
phonotactics and the lexicon).

  But why should we stop here? Why cound not we combine underspecified
vowels "A" and "I" into a new group "5"?

  You say that it is the only suffix of this type in Osmali. It is true
[or not, after writing this paragraph I recalled the negative infix],
but your above analysis is unique, too. How many verbal markers are in
Osmanli which have three different morphs for the same function?

  I think that <-r>, <-Ar> and <-Ir> are not different morphs but they
are the surface representation of the single morpheme <-5r> combined by
various types of stems. The underspecified vowel "5" has the following
values:
(1) Zero vowel after another vowel.
(2) After a consonant
 (a)if the previous stem (suffix) is lowering, treat "5" as it were "A"
 (b)if it the stem (suffix) is non-lowering, treat "5" as it were "I".

  The suffix of the infinitive is different, because it has the
underlying form <-mAk> not *<-m6k>. Thus we have "bilmek" because "A"
is not affected by "lowering harmony".

> Now the form of the non-lowering verb stems becomes an accident (which
> doesn't seem so bad, since there are exceptions), and not the form of
> the Aorist (which would be unfortunate, IMO).
  I do not see any unfortune here. I see this question in a complex
Ural-Altaic framework, and from this perspective, this phenomenon is
more frequent. However, it bothers me that aorist suffix is seems to be
the only Osmanli verbal suffix that has such an accident. In all the
other cases I know verbal suffixes have just one underlying form and
the actual allophones are determined by the phonological and lexical
features of the _stem_ (and not of the suffix).

  Meanwhile I was thinking it over and I found a possible test of my
theory: the negative infix shows a similar dimorphism <-mA> ~ <-mI>. If
we could create a common underlying form <-m5> for it (with the
assumption that the preceeding stems and suffixes[!] are lexically
lowering or non-lowering), it would validate my opinion. I have to
study the question of the negative infix in detail (but it takes some
time), and possible you also could survey it with a reasonable
benevolence against my theory.

> I'm very interested in the Hungarian phenomenon, though.   That might
> explain a lot.
  In Hungarian this is connected with nominal stems. A number of
suffixes have a triple undetermined vowel "3" = [+back] /o/ ~ [-back
-labial] /e/ ~ [-back +labial] /2/. Some of these have two additional
allophones before certain stems (or suffixes) as it were of type "A" =
/O/ and /E/. Thus we have suffixes showing vowel harmony type "3" only,
and others showing harmony type "A" and a third group that belong to
type "5" (="3"+"A").
  Hungarian plural marker <-k> is a complete parallelism to Turkish
aorist marker, that is it has a asyllabic variant, a variant
<-3k> and a variant <-Ak>, the distinction between <-3k> and <-Ak> is
determined by the stem, e.g. /Sva:b/ 'Swabian' is a regular non-
lowering stem, therefore its plural is /Sva:b|ok/. The word /la:b/
'foot' is lowering, its plural is /la:b|Ok/. The same is the situation
in case of the accusative marker <-5t>: /Sva:b|ot/ 'Swabian+ACC',
/la:b|Ot/ 'foot+ACC'.
  But the plural marker is lowering itself, therefore we have
/Svab|ok|Ot/ as the accusative of /Sva:b|ok/ and not */Sva:b|ok|ot/.
  Another interessant feature of lowering harmony is that it is somehow
connected with the parts of speech: if a word changes it parts of
speech even without derivational affixes, it can change its valence of
lowering harmony, though. E.g. we have an adjective /v2r2S/ 'red' and
it is regular non-lowering, its plural (when used predicatively) is
/v2r2S|2k/. When this word is used as a noun meaning 'man/woman with
red hair, communist, flower with red petals, etc', the stem became
lowering, its plural will be /v2r2S|Ek/ (this change is not systemic:
it is connected to the actual stem).