OT and functionalism

[Frederick Newmeyer]

I feel that Matthew Dryer makes an excellent point when he writes:

> functional principles and motivations apply
> primarily at the level of the evolution of particular languages.  If
> economic motivation has won out over iconic motivation with respect to
> some aspect of the grammar of a particular language, that means that
over
> the past few thousand years, the particular grammatical changes that
have
> led to the current state of the language reflect the influence of that
> functional principle.  But once that has happened, the grammatical rules
> have an existence that is independent of the explanatory principles that
> have influenced them.

        As I will try to argue now as concisely as possible, the kind of
'functionalist OT model' that Joan Bresnan has been advocating in her
postings (if I understand her correctly), in which the constraints
themselves are directly functionally motivated, falters when one tries to
apply it to a wide variety of disparate phenomena within a particular
language.
        OT analyses, particular in the realm of syntax, have, in general,
been rather circumscribed in their domain of application. Typically, they
focus on some little corner of the syntax, such as clitic order, auxiliary
inversion, and so on. The machinery intrinsic to the OT approach works
beautifully in such situations. But grammars are tightly integrated
wholes, so decisions about how one process should be handled are likely to
have repercussions for another, seemingly unrelated, process. This fact
leads ultimately to a much greater disparity between grammatical
principles and their functional roots than is posited by 'functionalist OT
models'.
        Let's posit two functionally-motivated OT constraints for English,
ICONICITY and HEAVY-LAST:

ICONICITY: Syntactic constituents reflect semantic units.
HEAVY-LAST: Heavy constituents follow light constituents.

ICONICITY can be illustrated by the fact that, in every formal account,
adjectives are generated under the same phrasal node as the noun that they
modify. HEAVY-LAST can be illustrated by the fact that within the verb
phrase, sentential complements are positioned after phrasal complements.
The functional roots of these constraints are so obvious that no
discussion is necessary.
        Now, then, which constraint is ranked higher for English,
ICONICITY or HEAVY-LAST? Well, it depends. In some cases, we have
identical grammatical elements in variant orders with no meaning
difference, each option corresponding to a different ranking of the two
constraints. For example, both of the following are grammatical English
sentences:

(1)     a. A man who was wearing a silly-looking red hat dropped by today.
        b. A man dropped by today who was wearing a silly-looking red hat.

Sentence (1a) reflects a ranking of ICONICITY over HEAVY-LAST; sentence
(1b) the reverse ranking. In some cases, however, only a higher ranking of
ICONICITY is possible. Simple adjective phrases cannot be extraposed from
the nouns that they modify, no matter how heavy they are:

(2)     a. An extremely peculiar-looking man dropped by today.
        b. *A man dropped by today extremely peculiar-looking.

And in other cases, HEAVY-LAST seems to be ranked over ICONICITY. When
comparatives are used attributively, the adjective is separated from its
complement by the head noun, despite the fact that together they serve to
modify semantically that head noun (3a-b). Their structural unity can be
obtained only in a manner that is consistent with HEAVY-LAST (3c):

(3)     a. That's a more boring book than any I have ever read.
        b. *That's a more boring than I have ever read book.
        c. That's a book more boring than any I have ever read.

        What is the solution to this ranking paradox? Only, I would say,
to abandon ICONICITY and HEAVY-LAST as constraints internal to English
grammar. Rather, what we need are something much more like the 'parochial'
rules, principles, and constraints of standard models of grammar that
interact to yield the grammatical sentences of the language. That move,
however, is incompatible with 'functionalist OT models'. The relationship
between detailed grammatical statements --- such as those that position
the constituents of NP --- and their functional motivations can be
extremely indirect. In other words, we are left with the sort of view
advocated in Newmeyer (1998), in which, in a global / historical sense,
grammars reflect external forces, but without each language-internal
grammatical statement being tied to a particular functional motivation.

Newmeyer, Frederick J. (1998). Language form and language function.
Cambridge, MA: MIT Press.

--fritz newmeyer