Theiling Online    Sitemap    Conlang Mailing List HQ   

Re: THEORY: [CONLANG] OT Syntax

From:And Rosta <a.rosta@...>
Date:Saturday, November 25, 2000, 1:58
Marcus:
> And Rosta wrote: > > > > That depends on your theory of "Gen". Prince and Smolensky > actually provide > > > two different models. One - the one nearly everybody uses - generates all > > > possible forms, then selects the most "optimal" form based on the > > > constraints. The other only generates the forms relevant for the next > > > constraint on the list. After that constraint has made a decision, the > > > optimal candidate is used as the base form for generating the next set to > > > be judged from. McCarthy is the only person I know of to use such a model > > > in Phonology (but I'm not a phonologist, so I don't know the literature). > > > Under this second theory of Gen, you do not get the forms generated > > > "willynilly". Indeed, McCarthy has suggested that Phonology should follow > > > Syntax in using such a Gen. (That was at the West Coast Conference of > > > Formal Linguistics, 2000). > > > >I have a gut preference for the former, more declarative Gen. But I'd > >be interested to see the arguments pro & con. > > I also prefer the former. The advantage for the second version is that it > makes the theory even more restrictive, and it requires that each > representation that a given constraint judges will be less marked than the > preceding constraint's. I am not sure where the empirical differences > between the two will be, because McCarthy's presentation at WCCFL only > discussed the topic in abstract terms without much in the way of application.
I'd like to see concrete examples of the latter approach in order to get a sense of how it works. Ideally, I'd like to see Gen generate everything and anything in the universe; the things that fail to satisfy the constraints of the grammar just end up not counting as sentences of the language. In other words, Gen is a total irrelevance to grammar; the entire language is stated in the constraints.
> > > Not at all. We can introduce another constraint that requires arguments to > > > occur in their cannonical position in certain contexts like echo questions > > > and such. You rank that constraint above Wh-initial, and the facts fall > > > out. But in neutral contexts, this new constraint is dormant and fronting > > > applies. > > > >It seems to me to be an inescapable weakness of both OT and Minimalism that > >they have such difficulties dealing with 'free variant' structures -- > >structures that are different but are the same at interpretation at the > >conceptual interface. > > This is certainly true of Minimalism. Chomsky explicitly states that there > is no optionality in syntax. All different structures have a slightly > different meaning to them, even if you cannot pin down the exact difference.
I find the opposition between optional/obligatory nonsensical in a declarative constraint-based grammar. All constraints are obligatory. If two sentences look like optional variants of each other, it is simply the case that each sentence happens satisfy all constraints of the grammar.
> However, apparently you are unfamiliar with the stochastic OT model.
I am.
> I'm probably not the best person to be explaining this, but here it goes: > > Under Stochastic OT, constraints are not in a simple "linear". Each > constraint is a "wave": imagine the one positive peak of a sine wave, and > you get the idea. Each constraint is a separate wave, and they overlap. So, > say 10% of the time, Constraint A precedes Constraint B, but the remaining > 90% Constraint B precedes Constraint A. This means that in 10% of the > utterances, the optimal candidate will be determine by A instead of B, but > in the other 90% the candidate faithful to B will win out over the one > faithful to A. > > There is a paper relating this to Syntax by Ash Asudeh at Stanford. > Unfortunately, I have not been able to read it yet, because the copies he > uploaded to Rutgers are corrupt.
For a model of competence, this seems wrong to me. All that matters is that Sentence X and Sentence Y are both grammatical; their frequency in usage is an irrelevance.
> > Wh-questions with and without fronting would be such > >an example, for although echo questions are not interpretationally equivalent > >to wh-fronted questions, quizshow questions are. > > But you must admit that quizshow questions are stylistically marked. > Therefore it is not unreasonable to claim that this sylistic aspect has > prevented movement, but in the interpretational portion of the derivation > (LF in a GB grammar, post-Spell-Out in Minimalism), the wh-element does in > fact raise. The typical argument that this is indeed the case is that > quizshow questions are ungrammatical in exactly the same contexts as > wh-fronted questions. In fact, wh-in-situ is what motivated the existance > of LF movement in the first place.
I'll admit that quizshow questions are stylistically marked, but I'd be unhappy to admit stylistic markedness as a syntactic feature. At any rate, the argument I'd prefer to make is that what is really misguided about Minimalism is that there is some kind of default form that is altered only when it violates some constraint or other in the course of a derivation -- that visible movement occurs only when there's something wrong with the version without visible movement.
> >Ah -- I'd been wondering who at UCLA is an OT syntactician. I know Jane > >Grimshaw has been doing OT syntax, and Joan Bresnan has been using it in > >her work. > > I understand there is a fair amount of OT syntax being done at UC Santa > Cruz. And I think somebody said something about Amherst once, but I'm not > sure at all. > > >OK, I get it. But in fact the implication that "OT syntax" is an OT-ization > >of some other theory, so you get "OT Minimalism" and "OT LFG", etc. etc., > >seems > >much more on the mark to me, and makes more sense. > > I agree. The paper by Asudeh I mentioned above is based on LFG. > > > > Not at all. I would just like to point out that I am not an OT > > > syntactician. I am an amateur dabbler, who spends most of his time split > > > between Minimalism (which I am very dissatisfied with) and Field Work > > > (which I love). > > > >Interesting. It seems to be true of all the artlangers here that at heart > >they are fieldworkers, even when, like Matt & Dirk, they're also > >theoreticians. > > Half the point to being a linguist is too learn "exotic" languages. The > more different from English the better. I actually like to make my theories > account for the obscure languages, then cram English into those theories -- > the opposite of the standard approach linguists take. I've had the thrill > of watching professors squirm at my proposals, but having to hedge on their > criticisms since they believe in Universal Grammar and Chickasaw plainly > does what I claimed. > > >I've never had any desire to do fieldwork, but then I'm more loglanger than > >artlanger. > > Fieldwork is addictive. Pam Munro told me that once, but I didn't believe > her. Now I do.
I hate the idea of relying on informants. If you work on English, as I do, you know that the judgements of most people, linguists and nonlinguists alike, are abominably unreliable. It takes the Dwight Bolingers of this world to have the acuity of perception to produce really accurate judgements. (And you don't have to be a native-speaker, either; expert non-native speakers (Ivan Derzhanski-type speakers) tend to make better judgements than natives.) I suppose if you're interested in Bernard Comrie-level description, it doesn't matter. But if you wanted to get embroiled in really really tiny but really really crucial details of binding constraints, say, then it'd be a nightmare.
> > > But I have professors who hate OT, and fail to see that it is not > > > fundamentally different from their own pet theories. That irritates > > > me sometimes. I have a friend who wanted to write her thesis in OT, > > > and all her professors refused to advise her until she switched back to > > > Minimalism/Antisymmetry. > > > >Bloody hell. That's outrageous. Was she set on not switching institutions? > > She wrote her thesis, got her MA, then dropped out to become a professional > singer in a Bulgarian Choir. She seems more happy now. > > > > All I can say is that the work has been done by people such as Ed Stabler > > > of UCLA and several of his students, such as Henk Harkeman. People are > > > doing similar work at Rutgers, IIRC. I should be able to be more specific > > > sometime during the next quarter, when I plan to take Ed's class on the > > topic. > > > >I'll remain interested if you share it with us then. > > While do.
Does that mean "will do"?
> > > For instance, the production of most biclausal structures takes longer to > > > pronounce than the "language buffer" in our brains can handle. > That is, our > > > minds can only see so far ahead of what we are actually saying. The > > > distance has been measured (according to Bruce Hayes, but I don't remember > > > the exact length or who did the work). Some agglutinative/polysynthetic > > > languages can form words that take longer to pronounce than the buffer can > > > handle. (The evidence comes from agglutinative languages that determine > > > stress from the end of the word, and checking to see how long a > word can be > > > formed before stress cannot be accurately computed anymore.) Minimalism > > > requires that the entire sentence be constructed before pronunciation > > > occurs. But how is that possible, if the sentence is longer than > the buffer > > > has space for? I can't believe that a multi-clausal structure can be > > > constructed, stored, then pronounced, when it is not even possible to > > > compute the stress pattern of a single super-long word. > > > >The most charitable answer I can give here is that Minimalism is not to the > >slightest degree a model of processing, except in the metaphors it uses to > >describe itself; it is a model only of competence. The obvious riposte to > >this is, How can an ostentatiously mentalistic model of competence ignore > >processing, given that a halfway decent model of processing could quite > >possibily handle the rest of competence? > > I wanted say that! :-) > > One of my syntax prof's seems to be getting tired of reminding me that > Minimalism doesn't model production. But I find it harder and harder to > believe that given the claims that are made.
I too have the impression that there's lots of prestidigitative eating-cake- and-having-it in TG. When it accords with results from psychology then these results are counted, when it does not accord with results from psychology, this is because the results from psychology pertain to performance. Personally I quite like TG, but only because I reject mentalism. If I embraced mentalism then I would be aghast at the paucity of decent mentalist syntactic theories around.
> > > We have to have a theory that allows the formation of the > sentence as it is > > > being pronounced. No theory of syntax or grammar that I know of is capable > > > of this yet. But we will by the time I finish my dissertation in > five years > > > or so. :-) (That's only a half joke. I am working in that direction during > > > my spare time. I occasionally discuss portions of my developing > theory with > > > various people -- professors and fellow grad students--, but nobody has > > > heard anything close to the whole story. The most damning criticism I've > > > been given so far is that my ideas are symmetrical to Minimalism.) > > > >I'm surprised at your claim. LFG was founded as a psychologically more viable > >alternative to TG. I don't see the processing problems in HPSG or Categorial > >or Word Grammar (the theory my PhD grew out of). I know of other, > >processing-driven formalist models of syntax, too. > > I have never heard of Word Grammar. Where can I read up on it?
It's declarative, putatively monostratal, surfacey, rather like HPSG but with dependency-based rather than constituency-based syntax (except for coordination). It is also very committed to Cognitive Linguistics's claim that grammatical knowledge is integrated with general knowledge.
> As for HPSG and Categorial Grammar, they may be less problematic for > processing, but I don't like them as a form of Syntax. I've had very > limited exposure to HPSG, but I've done a bit of work in CG, and don't > believe for a second that it is the right way to go.
I tend to agree, but you should make sure you've seen Mark Steedman's _The syntactic process_. As for HPSG, it is, in my view, the very embodiment of Common Sense. It never does anything crazy or stupid or wacko, but it suffers from the same lapses, failures, errors, myopia, etc. as Common Sense does. A theory for conservatives.
> As for LFG, I'm more > willing to accept it than the others. Maybe my problem with it is that I've > only read analyses done in the framework, I've never actually read the > exposition of what it should be. That goes on my list of things to do over > the Winter Break.
Are you thinking of Bresnan's new book, or what? Finding a decent exposition of LFG was a perennial problem. --And.