THEORY: rules, levels and representations [was: OT Syntax]

On Sat, 18 Nov 2000, And Rosta wrote:

> Dirk:
> > On Wed, 15 Nov 2000, jesse stephen bangs wrote:
> >
> > > Huh.  One of the first advantages of OT that was touted to me was the lack
> > > of ordered rules, rule ordering deemed to be unnatural.  You obviously do
> > > have ordered (or ranked) rules, though.
> >
> > These aren't rules; these are constraints. Constraints are statements
> > about allowed or disallowed structures, not instructions for changing
> > one structure to another.
>
> To the best of my knowledge, all linguists who speak of 'rules' as something
> not to be rejected in favour of constraints use 'rules' as a synonym for
> 'constraints'. 'Thou shalt not kill' is a rule, but it is a constraint,
> not an instruction for changing one thing to another. More to the
> point, whenever I've had the opportunity to interrogate a linguist who
> likes to talk in terms of one structure changing into another, they (rightly,
> in my view) claim this to simply be a convenient metaphor for describing
> constraints pertaining to the relationship between two structures.
This has not been my experience, nor is this the situation suggested
by a perusal of the literature. There seems to be a deep-seated
commitment among generative grammarians (phonologists especially) to
an overtly derivational view of linguistic structure--to the idea that
a linguistic analysis starts from an underlying structure which is
then changed incrementally to an surface form. Originally, a
derivation was a way to relate different levels of structure. Since
generative phonology argued away the phonemic level of representation,
something had to fill the gap; enter the derivation. Since that point,
theoretical discussions in generative grammar have been concerned with
understanding the  relations holding between rules, and not between
levels or representations. Hence the talk of feeding, bleeding,
counterfeeding, counterbleeding orders; extrinsic ordering, the
Elsewhere Condition, cyclic vs non-cyclic rule application etc. These
are about the rules which are used to get underlying forms to look
like surface forms.

I don't believe that this is the position that Chomsky intended to
take in the late '50s-early '60s, but he had become committed to it by
the time he and Halle published _The Sound Pattern of English_.
Indeed, in _Aspects of the Theory of Syntax_ he already defines Deep
Structure derivationally; a DS is only a DS if it is part of a
derivation which includes a well-formed Surface Structure (p138).

So I do not assume that linguists who talk about rules are merely
indulging in an extended metaphor; I take them at their word.

> Hence I take the claim that such and such a model does away with rules to
> be vacuous. AFAICS, the only special thing about OT is the way it resolves
> competing constraints by means of ranking, and resolution by means of
> ranking is different from (or at least, simpler than) resolution by means
> of subconstraint (i.e. an extra constraint that stipulates what happens
> in case of a clash of 2 other constraints) only when the competition
> between constraints arises in lots of different circumstances that can't
> all be accounted for by a single subconstraint.
>
> Not that I've anything against the idea of ranked constraints; I'm just
> mystified at how this simple idea burgeoned into the huge industry that
> is OT (in the USA). The obvious answer is sociopolitical, career-savvyness,
> bandwagon joining, and then the natural tendency of graduate students to
> continue doing what their teachers teach.
Absolutely. I am primarily interested in descriptive work. I also
enjoy thinking about theoretical issues, but I know well enough that
my talents and aptitudes don't lie in that arena. When I went to the
University of Arizona, it was with the idea that my dissertation was
going to be on Shoshoni phonology. The theoretical framework was
relevant only insofar as it would allow me to express the descriptive
generalizations in a concise and insightful way. OT worked out for
me, but other frameworks may have worked out equally well. In any
case, it was implicitly understood that my dissertation would be an OT
analysis. If I had been more theory-minded, I may have resented that;
as it was, my primary concern was to describe the sound patterns of
Shoshoni. So OT served my purposes, and my dissertation doesn't
advance the theory in any significant way.

> But can such a huge academic
> juggernaut have such a flimsy intellectual basis, in a discipline that is
> fundamentally rational and quasiempirical?
I'm not sure how to answer that one. Geoff Pullum argues in one of his
Topic ... Comment columns that linguistics is *still* trying to find
its rational foundations. Of course, that was in the 80's; perhaps
we've done it by now, though I doubt it.

> It has been suggested that OT took off so because phonology was in the
> doldrums, in comparison to syntax, but this seems wrong to me: the years
> before Smolensky and Prince had seen a great flowering of superb theoretical
> work in nonlinear phonology and the beginnings of attempts to explore
> whether the mechanisms of theoretical syntax could be seen also to underlie
> phonology.
Constraints were in the air in the early 90's. This was the period of
time which saw the rise of Paradis' theory of Constraints and Repair
Strategies, Declarative Phonology (with antecedents in Unification
Grammar and Categorial Grammar), and Harmonic Phonology. So the notion
of constraint ranking came at a time when phonologists were trying to
understand the place of constraints (as opposed to rules) in a
phonological derivation. Prince and Smolensky 1993 touched a nerve;
not only by the masterful exposition of a particular theory of
constraints (which is what OT was intended to be), but also by good PR
and Prince's rhetoric. I can imagine that if Declarative Phonology
(another constraint-based theory) had had the PR engine that OT did,
there would be a real lively debate going on about now regarding the
place of constraints in phonology, the notion of representations, the
use{ful/less}ness of levels, and what rules *really* do. As it is, DP
gets great respect and remarkably little attention (like classic
novels that no one wants to admit to never having read), and OT has
become the de facto theory of constraints.

Dirk

--
Dirk Elzinga
dirk.elzinga@m.cc.utah.edu