Re: NATLANG: Vowel harmony rules?

On 16 Jun 2004 David Peterson <ThatBlueCat@AO...> wrote:

  I am supposing that "Turkish" means Osmanli Turkish here.

> Turkish's underspecified vowels are capital A and capital I.
  There is a third type in Osmanli: A+I. This means that there is a
close-open harmony axis, too. The most significant representant of this
type is the marker of "genis, zaman" ("General Present" [I do not know
its ordinary English translation]), as well as the homophonic suffix of
the present participle.

  The close-open harmony is not related to the vocalic structure of the
words but it is a lexical attribute of the stems. Normal stems require
close suffix vowels, in this case an I-type one <-1r, -ir, -ur, -u:r>.
The so-called "lowering" stems govern open vowels, (=articulated by low
tongue position), that is A-type <-ar, -ir>. (N.B. vowel-final stems
bear non-vocalic <-r>).

  Despite of the fact that lowering--non-lowering harmony is a lexical
feature, we can predict it: lowering stems are one-syllable consonant-
final verbal stems, expect when final consonant is <-l> or <-r> and
there are four additional verbs ending in <-n> that are non-lowering
(<san-, yen-, den-, kon->).

  (N.B. "Lowering" stems a new Hungarian theory, I applied it to the
Turkish.)


> *However*, Turkish has one extra interesting facet.   Specifically, if
> a word ends in a palatalized consonant (I believe only k, g and l can do
> this?), then it will spread a [-back] feature, even if the vowel before
> it is [+back].
  According to my knowledge, <k>, <g> and <l> is palatalized before
circumflexed <a^> and <u^>. I was unable to find words with ending
<-ka^, -ga^, -la^, -ku^, -gu^, -lu^> in my dictionary, though,
therefore I could not test this type of vowel harmony.

  However in Chuvash back stems may end in palatalized <-n'>, <-l'> and
they behave as you described.

  In Turkic language, the harmony valence of the anti-harmonic words
(containing both back and front vowels) are determined by their last
stem vowel, e.g. Osmanli Turkish <esas ~ esas|lar> 'base ~ bases'.

  The Chuvash vowel harmony rules in brief:
- Vowels can be back: <a> /a/, <u> /u/, <i:> /1/, <a(> /@\/;
  or front: <e> /e/, <u:> /y/, <i> /i/, <e(> /@/.
  (In Viryal dialect there are additional vowels: back <o> /o/, <o(>
  /8/ and  front <o:(> /3\/)
  (N.B. Letters with breve accent <(> are reduced vowels, they could be
  also written in X-SAMPA as /a_X/, /e_X/, /o_X/ and /2_X/.)
- Vowel harmony extends to consonants: consonants near front vowels
  are palatalized.
- Palatalized /l'/ and /n'/ can occur in back-vowel stems. In word-
  final position they evoke front harmony. Therefore we can presume
  that word-final /l'/ and and /n'/ is followed by a "zero-vowel"
  which is front one. (The Cyrillic script of Chuvash reflects this
  situation: /l'/ and /n'/ is written by <l> and <n> resp. plus a soft
  yer.)
- Chuvash suffixes can be homomorphic and dimorphic.
- Homomorphic suffixes are (of course) anti-harmonic and they are
  contain always front vowels: a full <e> (e.g. nominal plural marker
  <-sem>) or reduced <e(> (e.g. Simple Future 3sg marker <-e(>). A
  third type could be the front "zero-vowel", because in Continuous
  Present, the third person singular is marked by the palatalization of
  the stem-final consonant, that is adding a soft yer in writing.
- Dimorphic suffixes may contain alternating full vowels <a> and <e>
  (e.g. accusative-dative <-(n)a> ~ <-(n)e>) or reduced <a(> and <e(>
  (e.g. 2pl possessive marker <-a(r>, <-e(r>). That is, there is no
  labial harmony in Chuvash (cf. Osmanly Turkish I-type suffixes)
  [N.B. Probably Viryal has, but I do not know]
  The 2sg possessive marker is special because it has the back form
  <-u> and the from form <-u:> (AFAIK this is the only suffix of this
  type).
- The 3sg possessive marker is originally homomorphic <-e(>, but
  assimilation takes place here which result in a trimorphic <-i> ~
  <e(> ~ <-u> ending. (<-i> is "V" + <-e(> where "V" is vowel of the
  vowel-final stems, <-u> is a contraction of stem ending <-a(v> +
  <-e>).
- After vowel-final stems some vowel-initial suffixes loose their
  initial vowel, while the others got a linking consonant between the
  two morphemes.

---------------------------------

  In terms of underspecified vowels, Hungarian is more complex than
Turkic languages. To complete my synopsis on 2th May, we have the
following types (feature mark [+lwr] denotes lowering stems):

1. [+back] <a> [O] ~ [-back] <e> [E]
2. [+back] <o> [o] ~ [-back +labial] <o:> [2] ~ [-back -labial] <e> [e]
3. [-lwr] type #1 ~ [+lwr] type #2
4. [+back] <u> [u] ~ [-back] <u:> [y]
5. [+back] <a'> [a:] ~ [-back] <e'> [e:]
6. [+back] <o'> [o:] ~ [-back] <o"> [2:]
7. [+back] <u'> [u:] ~ [-back] <u"> [y:]
and an odd type
8. [+back] <ja> [jO] ~ [-back] <i> [i]

Comments:
- There is an old pronunciation difference between lexeme <e> in
  type #1 and type #2 (open [E] vs. close [e]) but this is neutralized
  in [E] in the modern standard (however, my dialect still keeps the
  difference).
- Lowering stems are nearly unpredictable in Hungarian, however nouns
  are usually non-lowering, while adjectives are lowering: this is true
  also when an adjective is used as a noun, e.g. <A vo:ro:so:k
  vo:ro:sek> 'The Reds [=Communists] are red'. Verbs do not know type
  #3. Note that not just stems but non-terminal suffixes can be
  lowering or non-lowering in Hungarian, e.g. plural marker <-k> is
  lowering, cf. <vo:ro:so:k> 'reds (noun)' ~ <vo:ro:so:ket> 'reds
  (noun) + ACC' and not *<vo:ro:so:ko:t>.
- The labial harmony is determined by the labiality of the last vowel
  of the stem (or of the last item of the suffix chain) but -- on the
  the contrary of Turkic languages -- Hungarian has no simple rules for
  back-front harmony valency. The basic information was drafted in my
  posting on 2th May.
- There are a number of homophonic, anti-harmonic suffixes, when used
  in suffix chain, they do not change the harmonic valency of the word,
  e.g. <falu|si|ak> 'cuntry-folk' and not *<falu[si|ek>.
- Some suffixes loose their (first) harmonic vowel before vowel-final
  stems (suffixes).

-----

I used X-SAMPA notation in //'s and []'s, written data in <>'s contain
flying accents: <(> for bereve, <:> for diaeresis (umlaut) and <'> for
acute or apostrophe, and <"> for double acute.