Re: NATLANG: Vowel harmony rules?

On 22 Jun 2004 David Peterson <ThatBlueCat@AO...> wrote:

> There most certainly *are* accidents: They're called history.   There's
> no phonological reason why the plural of "ox" is "oxen" and not "oxes".
  I did not talk about phonological reasons, I wrote about
"phonological _methodology_". From this point of view, there is
only just one underlying plural morpheme in English, and the actual
morph is selected by two rules.

  Rule 1 gets it input from the lexicon, i.e. the headword
information of the actual word. In the lexicon every headwords have
a value of the [* plural_affix] := { ['N' plural_affix],  ['S'
plural_affix], ...} attribute:

  <plural_affix> -> <n-suffix> / ['N' plural_affix]_ , or
                    <s-suffix> / ['S' plural_affix]_ , ...

  Rule 2 works on the result "stem + <n-suffix>", "stem + <s-
suffix>", etc. of Rule 1. The other input of Rule 2 can be either
the lexicon or the phonological structure, it does not matter. The
significant is the approach: a set of rules evolve the surface
forms from underlying forms based on phonological features, lecical
attributes, etc., but not by "accidents".

  History is not an accident just it made e.g. past phonological
features appear now as lexical attributes.

> Especially since it would only be for one form.
  No matter what is your approach is, the behaviour of the aorist
marker is unique. I do not know why is it better to call the unique
features as irregularities. Moreover, when you consider the
languages of the Southern Russian Steppe, a language area, you will
see that Turkish aorist marker is an immature areal feature. From
this point of view, it is not only for one form.

> There is no phonological rule at play: It's just lexical.   Looking
> at the environment, though, will certainly be a clue to a learner of
> the language.
  Phonological methodology is not for the learners of the language.
It is for the linguists describing the language. It is for desiging
a language automaton which generates possible realizations of the
language without human interactions. How could an automaton work on
"accidents"? As an analogy, it would be hard to make computers if
there would be "accidents" when you multiply 2 by 2 (sometimes
resulting 4, sometimes 5).

> So, for example, if they encounter a verb they've never heard before
> that's monosyllabic and ends in /l/ or /r/, chances are if they add
> the aorist suffix it'll be /-Ir/.
  I my posting I wrote two different things. Once I drafted a rule
that specify the actual form of the aorist suffix based on a
lexical attribute of the verbs. It was for the science and I did
not involve the phonemical structure of the verbs at all.

  The second one was a clue: What are the statistical
characteristics of the verbs having lexical attribute [+lowering]?
It was not for the science, it was for the questioner, who did not
study Turkish, to have an overall impression of the extent of the
problem.

> But they could be wrong, and since they *could* be, positing a
> phonological synchronic rule would not account for the real-life
> facts of Turkish.
  Do you have statistics about what percent of verbs break the
"brute force" prediction I cited? And on what probability level?
Errors below 5% do not disprove the theory in biological samples,
and this rate could be even higher in linguistic applications.

  Thus my citation could be valid as a _clue for non-native
speakers_ since they have not always a Turkish lexicon by
themselves which is promptly accessible. (I did not mean that it is
valid rule for the langues desciption.)

  Moreover, I wonder whether it is a useful clue even for Turks
themselves. I speak a language which has similar features. When I
have to suffix a rare lexeme as a native speaker, I myself rely non-
100% quasi-phonological clues if I have no a proper dictionary by
me.

> I don't think it's worth introducing a new vowel harmony category
> that will only exist for a certain number of words
  I think our scientifical approaches are very differing: I do not
think it is worth introducing a new irregularity into a regular
system... :)