Re: THEORY: two questions

Dirk Elzinga wrote:

>> >Indeed. While there are examples of each, Nichols claims to show
>> >that there is in fact a statistical correlation between verb-
>> >initial languages and head marking. Though some have questioned
>> >the reliability of her conclusions ...
>>
>> Your response raises a deeper question, which is:  What is the
>> significance [no pun intended] of statistical tendencies in linguistics?
>> Are typological correlations like the one Nichols observes merely
>> historical accidents, or do they reveal something fundamental
>> about the structure of language which linguistic theory needs to
>> account for?
>
>What I gathered to be the point of Nichols' book (or at least
>one point) is to emphasize the typological perspective on the
>similarity between languages. It seems to have been written as
>an indirect response to the "super-groupers"--those who would
>like to create language families like "Amerind" or "Nostratic".
>Since much of that work rests on typology as well as hard-core
>reconstruction (and, as some critics would have it, wishful
>thinking), Nichols' purpose seems to be to provide another
>way to think about the typological facts. That is, the
>predominance of head-marking in the languages of the Northwest
>Coast isn't necessarily the result of genetic relatedness, &c.
Ah, I hadn't thought of that as a possible motivation for
typological work--as ammunition against the macro-groupers.
From that perspective, Nichols's goal is admirable.  The ability
of neighbouring languages to influence each other's structure
(and hence, typological classification) tends to be woefully
underestimated by the macro-groupers.  Dixon makes a
(justifiably, IMO) big stink about this in "The Rise and Fall of
Languages".  He goes so far as to conclude that there is no
real evidence for genetic relatedness even for some of the
more established macro-families, such as Altaic, Niger-
Kordofanian, and Pama-Nyungan.

>> My personal bias is to ignore such tendencies, even
>> statistically significant ones.  Since the goal of generative linguistics
>> is supposedly to answer the question "What is a possible human
>> language?", I'm inclined to focus on features of language which
>> *always* occur, or *never* occur, and to disregard those features
>> which *usually* occur, or *almost never* occur.  (Of course,
>> since our data sample is pitifully small, we can never be completely
>> sure that a given feature will always occur, or never occur, but
>> we do the best we can.)
>
>I see Nichols' research program as being orthogonal to the
>generative enterprise. Speaking personally, it's nice to step
>back once in a while from examining the bark of my tree (Gosiute
>Shoshone phonology) to see the typological forest out there, and
>to read about how forests grow. As for what it says about UG,
>who knows? I don't.
Well, *every* feature of language says *something* about UG,
even if it's something negative.  But I take your point.  And I'm
not saying that typological generalizations are useless, either.
Such generalizations illuminate an important and non-obvious fact about
human language, which as that, out of a large number of logically
possible solutions to the design problems of grammar, languages tend
to go for the same solutions over and over, while ignoring others.
If that doesn't say something about UG, then nothing does.

>> And yet, it is intriguing that certain typological features appear
>> to be highly common (but not universal), while other features appear
>> to be extremely rare (but not unattested).  What, if anything, are
>> we to make of this?  Why, for example, is SOV order commonplace
>> while VOS order is relatively rare?  Is this merely an unexplainable
>> historical fact, something which could easily have turned out
>> otherwise?  Or is it inevitable?  Is Universal Grammar set up
>> in such a way that SOV order is somehow 'easier to get' than VOS
>> order?  And if SOV order is easier to get than VOS order, then
>> why does VOS order occur at all?  As someone with an interest
>> in typology, I struggle with these questions all the time...
>
>Yes, I know what you mean. Many of these kinds of questions seem
>to be addressed in functionalist literature. The functionalist
>explanation for these kinds of facts don't preclude UG, though
>(or do they?), and I find many functionalist arguments extremely
>compelling in phonology. I don't know enough about syntax to
>have a clear opinion, but I suspect that of the range of
>possible human languages (WRT word order anyway), there is some
>winnowing done by functional principles which might lead to this
>kind of skewing.
Speaking as a devout-but-moderate generativist, I have no problem
with functional explanations in principle.  After all, language is
used to communicate, and it's reasonable to assume that that function
shapes the design of language to some degree.  What I object to is
the sweeping claim that functionalist arguments can explain *all*
every aspect of language design.  That seems to me to be patently
false.

It's a tricky question, though--one which has become even tricker
in recent years with the advent of Minimalism:  I think you're right
when you say that functionalist explanations for certain phenomena
do not preclude UG.  In fact, as I interpret Chomsky's Minimalist
Program, determining the proper scope of functionalist explanations
is vital for determining what UG is.  In Minimalism, Chomsky claims
that certain properties of language design are imposed 'from the
outside' by the fact that language has to interface with other cognitive
and physiological systems (e.g. speech production, processing, semantics).
For example, the fact that words have to be spoken in a particular linear
order follows not from any inherent requirements of the language faculty,
but from the fact that we use speech (sound transmission) as our
medium for communicating.  (This can be seen by examining signed
languages, which have non-linear properties that spoken languages
lack.)

From this perspective, we can think of UG as consisting
of those properties of language which cannot be explained as
necessary consequences of how the language faculty interfaces
with other systems.  In other words, the fact that we process
sounds in a certain way, the fact that we divide the world into
semantic categories in a certain way, the fact that our speech
organs move in a particular way, etc., all of these factors impose
'external' restrictions on the design of language (in the same way
that gravity imposes certain restrictions on the design of tables--
restrictions which have nothing to do with how tables are supposed
to be used).  The set of properties of language which *cannot* be
explained away as consequences of these external factors, says
Chomsky, *that's* what UG is.

I guess a strict functionalist would say that once you eliminate all
of the externally-imposed design features of language, you're left
with nothing.  I.e., there is no UG.  A generativist would say that when
you eliminate all those features, there's still something left over.
What exactly it is, that's what we don't know.

Matt.