LINGUIST List 11.146

Mon Jan 24 2000

Disc: Newmeyer: Language Form & Language Function

Editor for this issue: Anthony Rodrigues Aristar <aristarlinguistlist.org>

Directory

1. John Graham, Re: 11.129, Disc: Newmeyer: Language Form & Language Function
2. Mailbox, Disc: Newmeyer: Language Form and Language Function

Message 1: Re: 11.129, Disc: Newmeyer: Language Form & Language Function

I would like to offer some observations regarding comments made by
Ingo Plag (LINGUIST 11.129) and the general discussion thread about
Newmeyer: Language Form & Language Function. As a person who has
studied with functionalists and now with formalists, I am well aware
of the charges and counter-charges that have been leveled by the
discussion participants. First I would like to comment on the status
of grammaticality judgments in formalist theories. Ingo Plag (11-129)
relates an experience of confronting questionable sentence judgments,
the methodology of collecting judgments, and the implications of
such. Let's take Ingo's situation: suppose I hear a conference talk
about my native language, and I doubt some of the judgments that the
presenter gives. I then confront the presenter with my doubts. What
are the expected reactions? In my experience the two most popular are:
1. Tell me I'm wrong. Yes, unbelievably I have often heard this. (For
instance, my English does not have strong That-trace effects. Now, try
to guess how many times I've been told that I am "wrong".) Since I am
a native speaker who is honestly reporting judgments, and, assuming
the presenter is also a native speaker (I leave aside the issue of
non-native speakers here), then what do we have? A disagreement. Two
native speakers with a honest disagreement. What to do? Well, the
presenter could go on insisting I'm wrong, or proceed to one of the
following options. 2. Ignore the point. Or, more subtly, seem to
acknowledge the point, but then hand-wave it away with a comment about
"inessential details," "minor problems," or the like.

I know the above two options seem cynical, but then there are other
possibilities: 3. Acknowledge genuine native speaker variation,
collect a number of judgments, and filter out the "statistical
noise". This is perhaps seems better than the first two options, but,
like option 1, it is actually a form of prescriptive grammar (and
linguists do not prescribe, right?). By acknowledging the simple
correctness (as in option 1) or the statistical normalcy (as in this
option) of a judgment, one is simply stipulating what is correct, in
the face of contrary evidence. 4. Illegal dumping. Say that the
difference is one of "semantics" or, if there are semanticists nearby,
"pragmatics". (There are other places for illegal dumping also. Say it
is due to "intonation" or is a "PF operation".) 5. Acknowledge
genuine native speaker variation, and make a note of it. Move on.

I advocate option 5 (not to say that 3 cannot be used as well, just
without subsequent acceptance of only one valid judgment). I
particular, I believe that much of the functionalists' frustration
with and criticism for formalists on this issue come from the
formalist tendency to adopt options 1 -- 4. Why should we not expect
genuine native speaker variation? I am not advocating a "relativism of
judgments" here. Rather, I am advocating more investigation: Wouldn't
the more interesting line of research be to see if variation forms
groups, perhaps delineated by parameters? (For example, if a person
does not have that-trace effects, what else is predicted about that
person's grammar and why?)

Finally, and again simply in my own experience, I feel that formalists
tend to see grammaticality judgments as central, while functionalists
opt for usage. I believe that both in isolation are epiphenomenal. 
Whether or not a sentence (or utterance, proposition -- take your
favorite term here) is "in" (grammatical, acceptable -- ditto) a
language or not is a function of (at least) both the formalists' and
functionalists' criterion taken together. If this in any form is true,
the Newmeyer's book is indeed very important; members of both camps
should read the each other's work, not to construct straw men, but
rather to construct bridges.

 -- John Graham
 Ph.D. Program in Linguistics
 University of Connecticut

This is really a (partial) answer to some of the points raised by Andrew
Carnie's (AC) last contribution (LINGUIST 11.110) to this discussion.

Simple matters first: As far as I can recall these many years later,
the early work on trying to prove the 'psychological reality of
transformations' was done by George A. Miller. He tested (again, if
memory serves me right) memory loads of some simple (brief) sentences,
given the Ss in the form of statements, negated statements, question,
negative question, passive, negative passive, and questioned, negated
passive. His results were what you'd expect: The more operations the
Ss had to 'undo', the greater was the memory load. (This work was done
still within the "Syntactic Structures" model.)

Later work, e.g., by Jerry Fodor (and by then already within the
"Aspects" model), showed that matters were a lot more complex than
that. In fact, by the later sixties, I knew no-one who any longer
believed in a simple relationship between transformations and what one
might want to call their 'psychological reality'. Most of us thought,
though, that transformations represent certain relationships between
sentences; that, in fact, transformations were some sort of formal
representation of these relationships.

Turning now to AC's summary presentation of why "the descendants of
Generative Semantics ... abandoned [the transformational enterprise]
wholesale" gives, to my mind, a false picture of what happened.
First, the real issue was *not* transformations and/or their
psychological reality; but rather, the question of semantics and how
it relates to Deep Structure (which we later called "Underlying
Structure --US -- to differentiate it from DS), and the psychological
reality of *this* relationship. Arguments also turned on questions of
synonymy and how to represent it, and on the question of prediction,
since e.g., semantic class membership often allows to predict certain
aspect of syntactic behavior, and these generalizations could not be
captured in a model using interpretative semantics. There were many
other issues; e.g., the nature of the lexicon; where, again, the
models proposed by Chomsky and his followers did not allow us to
capture many generalizations we thought worth capturing.

I refer interested readers to Randy Harris' excellent book, " The
Linguistic Wars", for a detailed and insightful presentation and
discussion of this part of the history of linguistics.

We did not, BTW, "abandon the entire transformatioinal enterprise":
Generative Semantics" was a *revision* of the "Aspects model".

I can't speak for other cognitive linguists, but for me Chomskyan
circles aren't really the only "mainstream" ones. Seriously, I can
understand Lakoff (I unfortunately don't know the '91 paper AC refers
to) when he "claims that Generative Grammar (Chomskyan circle type)
ignores results from psychology", because the often used term
"psychological reality" is simply too vague, and anyone can use it to
mean whatever he or she wants. Certainly, what cognitive linguists
mean by the term, and what Chomskyan formalists do, is *not* the same.

In addition to the differences in the goals of linguistics between the
two schools I presented above, there is the question of the
relationship of language to the rest of the mind/brain's higher
functions that separates them. (And this, too, is an aspect of
psychological reality.) That is, Chomskyans generally accept a much
stronger version of modularity (the separations between linguistic and
other higher functions) than do cognitivists. One fairly important
argument here is the matter of constructivism. That is, cognitivists
adduce evidence that, from the time we are infants, we actively
*construct* models of our environment, which we use in our efforts to
understand the world. Needless to say, such models are both
linguistic and non-linguistic; by which I mean that individual models
are often a mixture of both. This fact is then seen as an argument
against the strong modularity position. Cognitivists cite much other
evidence against this position. Unfortunately, we still know too
little of the mind/brain to evaluate much of the evidence properly,
whichever sides cites it. This reminds me of the saying that, "The
devil himself can cite the Bible to his advantage."

Hope I managed to clarify some points.

Peter Menzel