Re: THEORY: OT Syntax (Was: Re: THEORY: phonemes and Optimality Theory tutorial)
|From:||And Rosta <a.rosta@...>|
|Date:||Saturday, November 18, 2000, 4:53|
> Jesse S. Bangs wrote:
> > > As a syntactician I take offense to you claiming it is a "phonological
> > > theory". :) It works quite well for syntax too.
> >I don't know any OT related to syntax. Please share.
> I'll explain by illustration of deriving Wh-questions in English.
If this is to be a distinctively OT analysis then you need to show that
the constraints conflict and so must be ranked.
For the particular example you give, it's hard to see whether conflicting
constraints are called for. I have a competing analysis without conflicts,
but conceivably you could come up with an analysis with fewer but
conflicting constraints and so arguably superior. But the three constraints
you give are problematic. "Stay" seems pointless, because 'Gen' should
generate all forms willynilly, and random or any other rearrangements
should simply not be part of the model. It's not clear from your exposition
why Q-head requires inversion, so I can't judge Q-head. Finally, Wh-initial
is all well and good, but it wrongly excludes quizshow questions ("Mary
will see who?").
I must in fairness concede that this example was chosen as an exposition
of what OT syntax is like rather than as an exemplar of analysis where OT
does better than other theories, but on the other hand the same could be
said of 98% of the OT analyses I've seen.
Could you give an example of an analysis does better than other theories?
Just point me to someone's article if that's easier.
BTW, I'm assuming that you're keen to engage in 'lively debate'. But if I'm
coming across as bullying or obstreporous, just let me know & the thread
will immediately be aborted.
> basic problem. You have a pair of sentences like:
> Mary saw a man.
> Who did Mary see?
> In order to derive the question, the first thing you do is add in "do" and
> put it in the past tense form "did". Tensed verbs normally appear after the
> subject, but in the question, "did" gets moved in front of the subject (so
> called Subject-Aux Inversion). Then you substitute "who" for "a man", and
> move "who" to the beginning of the sentence. So, how does OT make this work?
> First, we have a constraint called Stay which says "Don't move anything."
> This is important because we don't want to be able to move the words around
> randomly; that is, we want to rule out *"Who Mary see did?" and things like
> that. This of course means that without any other constraints, we would be
> saying "Mary did see who?".
> Another constraint, we'll call Q-Head, which says "Have a question element
> as the head of the sentence." For those of you who don't know much theory,
> this forces the existance of "ka" in Japanese questions, "ma" in Chinese,
> and forces "did" to appear sentence initially in "Did Mary see a man?".
> This constraint says that we want a sentence like "Did Mary see who?"
> Q-head and Stay are in direct competition, since it is impossible to
> satisfy both in English. We rank Q-head as more important to satisfy than
> Stay, that way we can move "did" to the head of the sentence.
> Our last constraint is WH-intial, which says "Put the wh-pronoun at the
> beginning of the questioned clause." (Or to put it more technically, "put
> the operator in the specifier of the Q-head"). This constraint tries to
> create a sentence like, "Who Mary did see?"
> This constraint is also in competition with Stay. Once again, we rank
> WH-initial above Stay, so "who" moves to the beginning of the sentence.
> To put this in a chart (* means the "candidate" violates the constraint, *!
> means the violation causes the candidate to be disprefered to the others,
> -> means "THE WINNER"):
> | WH-initial | Q-head | Stay
> ->Who did Mary see? | | | **
> Who Mary did see? | | *! | *
> Did Mary see who? | *! | | *
> Mary did see who? | *! | * |
> We can simply rearrange the order of these constraints to get other
> language types. Say, Japanese, Korean, and Chinese have question markers
> but no wh-movement. So we put WH-initial below Stay, and there you go,
> languages that have question markers at the head of the sentence, but the
> Wh-pronoun stays in place.
> For all of you who may have been brainwashed into believing that Minimalism
> is the only way to go,
= 0% of the list membership, as far as I can see. Matt is the only high-
profile Chomskyan here, and he's not a heavyduty Minimalist.
> I'll just mention that OT and Minimalism have very
> similar conceptual bases, they just differ in their formalism.
> And to be fair, for all of you who have been brainwashed into believing
> that Minimalism is the Anti-Christ, Minimalism is one of three theories of
> syntax that have been shown to be learnable, the other two being Tree
> Ajoining Grammar and Categorial Grammar. The jury is still out on OT, and
> all the rest (Transformational Grammar, Lexical Function Grammar, HPSG,
> Role and Reference Grammar, etc) have been proved to be unlearnable.
> Chomsky was not happy to hear that about his theory. :-)
Since so little is known about learnability from a psychological perspective,
I assume this argument must be about learnability from a logicomathematical
I'd like -- nay love -- to see the evidence for the claims you give. Can
you give the reference?
BTW, I remember reading a paper by Shalom Lappin & David Johnson about why
*Minimalism* (unlike certain other theories -- I forget which, but possibly
including GB and HPSG) was pyschologically unviable. I forget the details,
but the essence was that economy could be computed only over entire derivations,
so to compute the most economical derivation you have to compute every
derivation. I seem to recall that Minimalism has since taken steps to fix
that problem, though.