optimality in synchrony and diachrony

[Joan Bresnan]

(long message, last message, please forgive)

Martin Haspelmath asks me to clarify my reply to Matthew Dryer.
Matthew argued (in essence) that OT can't be truly functionalist
because functional constraints are external to grammar, shaping
language in an evolutionary way, while OT constraints are internal to
grammar.  Matthew and Fritz Newmeyer believe that for this reason OT
is fundamentally misguided, and Martin Haspelmath is inclined to agree
with them.

My reply was that this argument sounds plausible because it is a
tautology, true only by definition.  It rests on an erroneous
definition of "grammar", which is not what an OT "grammar" is.

What is a grammar and how exactly can it be "shaped" by evolutionary
forces?  Matthew's view of grammar may be shared by many on this list,
but it one that has been completely abandoned among most phonologists
and also among a number of OT syntacticians.  It is the view expressed in the
following quotations from Matthew's original msg (emphasis added):

        "For example, if two languages have a difference that one might
        describe in terms of economic motivation competing with iconic
        motivation, where economic motivation wins out in one language and
        iconic motivation in the other, I WOULD NOT WANT TO SAY THAT THERE ARE
        GRAMMATICAL RULES IN THESE LANGUAGES THAT REFER TO ECONOMY OR TO
        ICONICITY."
and:
        "... 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."

The conception of grammar that these passages invoke is one in which a
grammar consists of RULES and/or PRINCIPLES AND PARAMETERS.  Since the
structure of a language is determined by the rules and/or parameter
settings of its grammar, on this view, it doesn't make sense for there
to be "internal" functional forces also shaping the language.  [I
leave aside the question of how exactly languages are supposed to be
shaped by EXTERNAL forces on this view.]

But an OT "grammar" is not a grammar in this sense.  Nor is it a
"grammar" in the Principles and Parameters sense.  OT rejects
Principles (because they are inviolable) and Parameters (because,
among other reasons, they are categorically either on or off and
therefore incapable of explaining the EMERGENCE of the UNMARKED) [see
also the critique of parameter learnability in Tesar and Smolensky].
So OT is a much more radical departure from the conventional thinking
about what grammars are than Matthew and Fritz seem to recognize.

An OT "grammar" is a model of how a set of possible structures (the
candidate set) is shaped by the prioritization of conflicting
universal constraints, into a particular language.  The constraints
reflect (hypothesized) properties of the human articulatory,
perceptual, and cognitive systems.  They evaluate the markedness and
the faithfulness of the structures as expressions of content.  The
faithfulness constraints require that features of the input content be
preserved in the output expression.  They thus serve the communicative
function of expressing contrasts in content, protecting content
against the eroding effects of markedness constraints on forms.
Markedness constraints penalize complex or `difficult' structures, and
so tend to erode contrasts.  A particular language harmonizes the
conflicting constraints by prioritizing (ranking) them in its own way.

Note that the constraint component that does the selection is external
to the structure-generating component.

Unlike the older, conventional conceptions of grammar which it
replaces, OT provides both an EXPLICIT way to model how languages can
be shaped by functionalist pressures and also an explanation for why
language-internal processes or "rules" (which it deconstructs
completely in terms of markedness and faithfulness) reflect
typological patterns.

My example (from Kager's textbook): Final obstruent devoicing is a
productive process in the grammar of Dutch among other languages.
Presumably it applies to borrowed words, and affects Dutch
second-language learners of English, etc. etc.  It is not a historical
relic, frozen in a few lexical forms, but a living process or "rule"
of the sort that phonologists would write in their descriptions of the
grammar of Dutch (in the broad sense of grammar used here).  The OT
analysis of this devoicing "rule" in Dutch derives its effects from
ranking three constraints: (1) faithfulness to voicing contrasts, (2)
one markedness constraint reflecting the difficulty of perceiving
voicing contrasts in certain less-salient positions (syllable codas),
and (3) another constraint reflecting the general markedness of voiced
obstruents (compared to vowels, for example)--you've got a closure and
you're vocalizing, and it's tough. [pardon my phonetics].   Now if you
take this analysis and arbitrarily rerank these constraints in all
possible ways, you create little obstruent-voicing/devoicing
"grammars" for different languages.  What you find, is that all these
rankings give you only three different possibilities: voicing
contrasts in obstruents everywhere (like English); voiceless
obstruents everywhere (like Polynesian); and voicing contrasts present
in the salient onset position of syllables and absent is the coda
position (Dutch).
 In a nutshell, the analysis of final devoicing in Dutch predicts the
existence of the well known typological asymmetry in obstruent voicing
across languages.

It is because the OT constraints are universal and not
language-particular rules or parameter settings, that this deep
connection between typology and language-internal "grammar" is
possible.  You literally cannot do the (real) OT grammar of any
particular language without doing typology.

Martin Haspelmath wrote:

> I think I didn't understand the thrust of Joan's example about the
> devoicing tendency across languages and in particular languages either.
> This is something that typologists have long been aware of (see e.g.
> Bill Croft's markedness chapter in his typology textbook). The usual
> functionalist explanation for both cross-linguistic patterns and
> language-particular regularities is that they show the effect of
> diachronic change.
>

But precisely how?


Joan Bresnan