LINGUIST List 7.626

Fri Apr 26 1996

Disc: Ungrammaticality

Editor for this issue: Ljuba Veselinova <lveselinemunix.emich.edu>

Directory

1. benji wald, Ungrammaticality
2. Karl Teeter, Native speaker judgments
3. Alexis Manaster Ramer, Ungrammaticality

I want to comment on Larry Hutchinson's last message to the List,
because I think that some of the overstatements he made there imperil
the possibility of the good points he made being taken with the
seriousness they deserve.

He opens with:

"There is another aspect to this: since 1957 or so a lot of people
seem to have been confused about what a grammar is supposed to do. Is
a grammar supposed to account for native speakers' judgments about
grammaticality (synonomy, ambiguity,...) or account for grammaticality
synonomy, ambiguity,...)? These are not the same thing!"

I don't think this is an effective way of presenting this issue. As
far as I can tell, it is almost tautological to say that a grammar is
supposed to account for grammaticality (the second alternative),
though I'm not sure in what way to understand "synonymy", in
particular, and "ambiguity", in ways to be discussed further below,
given the changes that generative theory has undergone since 1957 (=
Chomsky's Syntactic Structures).

Next he says:

"Judging a sentence to be ambiguous is an ACT. That a sentence is
ambiguous is certainly not."

The first sentence above is an important one. I would say that
judging a sentence to be ambiguous, or to mean one thing or another in
general, for that matter, or to be "good" or "bad", is behavior, and
therefore "performance". This may have always been realised by
syntacticians, but it's implications were not, and are still matters
for further development. The "blindness" I mentioned in my last
message is one problem, a performance problem.

Larry's example is relevant to the blindness problem. He says:

"Present to a class the sentence "I heard the girl playing my song",
and poll them on ambiguity. I submit that the majority will judge it
to be unambiguous (I have done this many times)."

I'll bite. I suppose the "ambiguity" is one of tense:

a) I heard the girl -- she WAS playing my song when I heard her
b) I heard the girl -- the one WHO IS playing my song NOW, not
 necessarily when I heard her (Maybe I
 don't even hear her playing my song NOW,
 maybe somebody has told me that she's
 playing it now somewhere else.)

Am I putting myself out on a limb to deduce that a) is the
interpretation picked by (most of) the class, and that b), in the
various circumstances I have indicated, does not occur to them? Why
do I pick a)? Because it's the first interpretation that occurred to
ME -- and I might not have even looked for other interpretations if
Larry hadn't indicated that it was ambiguous. He's not clear about
whether he had his classes search for other interpretations. They
might have reason to get mad if he didn't, and then says, ha! ha!
Guess what, it can also mean... Do they then deny that these
interpretations are possible? I doubt it. So I have just raised an
issue which amounts to experimental design, and the interpretation of
*performance* data based on that design. By the same token,
grammatical judgments of any kind are performance data. They're only
a problem when we disagree with them (which obviously happens with
disturbing frequency -- so they are a problem -- but not a paralysing
one).

Larry then asks:

"Do you want your grammar to account for this fact? Or do you want
your grammar to account for the fact that the sentence IS multiply
ambiguous (whether or not particular people judge it so)?"

The rhetorical purpose of forcing alternative questions here is
inscrutable. If I have to choose, I choose the second alternative --
for the example!! -- and I don't believe that the class refuses to
acknowledge the "ambiguity" when it is presented to them, unless they
have a less flexible understanding of what means "ambiguity" than
linguists do. (I doubt that.)

Nevertheless, we might argue about whether the "ambiguity" is
grammatical (= syntactic?), or semantic or pragmatic. Note that:
 I will hear the girl playing my song
is ambiguous in the same way (accepting his use of the term "ambiguity"),
but
 I hear the girl playing my song
is not. Or is it? Can it "mean" I hear the girl playing my song burping
(while she plays it), or muttering about why her promoter schedules her
to play such trash. And then it can also mean
 I (always) hear the girl playing my song (now) playing
 something or other (whenever I turn on my car radio).

Assuming only tense is at issue here, the problem remains what the
nature of the ambiguity is, syntax or semantics of -ing, or what?
I'll leave it at that (though a lot more can be said about both the
syntax and semantics of -ing), because that's a good question, but I
don't think it was clear from the way Larry presented it.

What about his first alternative above? Should a grammar account for
preferred interpretations? It's a good question. I think the answer
is, it depends. I won't pursue this because it would lead me into
issues about how grammars change. With regard to the example, I would
say NO! But I am interested in accounting for the preferred
interpretation. I think in this case, "most accessible
recontextualisation" is a better way to put it. Then we see what
(some) linguists call "pragmatics" operating. The first tense clue we
get is "past" in "heard". So why change unnecessarily as the sentence
progresses? So we stay with it, and understand "the girl (WAS)
playing". There are other features involved in recontextualising that
sentence but that's the main one, and it's enough to indicate why it's
most likely to be recontextualised in that way. "Flying planes can be
dangerous" is more of a toss-up, "Old men and women" favors "old" for
the entire construct. It's certainly interesting to contemplate
pragmatic effects on "ungrammaticality" judgments, but Larry doesn't
do that, so his contribution is appreciated but limited. We do get to
see how "blindness" can operate in failure to produce multiple
recontextualisations of decontextualised examples.

(NB we discount "context of linguistic discussion" as a contextualisation
of our examples. That context counts as "decontextual". Wanna argue?)

Larry goes on to say:

"Notions such as "performance" and "the ideal speaker-hearer" were
designed to collapse the distinction in an odd sort of way. The real
intent, though, was to get linguists out of the people business. No
actual person is this ideal speaker-hearer, so none should be polled.
And of course no such ideal can be polled about anything either. Exit
polling (from linguistics)."

I think the distinction he's referring to is between judgments as
performance and as competence, but I'm not sure. I'm more interested
in his historical comment.

I think he's wrong. Before that fatal day (Saussure's pronouncement,
or Chomsky's?) linguists were not in the "people business", no more
than they had to be if they didn't happen to know the language
themselves, and/or have written texts. Syntax in particular and
Grammar, as its etymology shows, was developed according to a
philological tradition, and developed its concepts and impetus from
written language. In fact, until Chomsky 20th century linguists
didn't have much to add to the impressive and mountainous amount of
work on syntax (or grammar, if you will) that had already been
produced by traditional grammarians using late 19th century tools and
concepts. Look at descriptions of languages which were not done in
the philological tradition (i.e. most of them) before Chomsky.
Phonology, morphology, some folktales (to record connected discourse,
not to analyse it). Does Larry mean "folktales" by the "people
business"? I don't get it. If Larry means linguists should realise
they're in the "people business" I agree. But it's not the fault of
generativists that they're not. They were just following in the
footsteps, or leaping over the shoulders, of their predecessors.

Finally, he says:

"Of course, anyone who continues using expressions like
"judgments/intuitions of the native speaker" is going to continue to
face student unrest (if not mutiny) in Syntax I, because the audience
is going to take such expressions at their face value."

I refer back to what I already said above about experimental design.
Beyond that, in context "student unrest" is an encouraging sign. --
Benji

Dear Larry: Responding to yours on the LINGUIST list 4/23 (re:7.591),
what a grammar is supposed to do is present a theory of a language.
Native speaker judgments are raw data, which is as messy for
linguistics as for any science (perhaps more so). What any scientist
has to do is process raw data to eliminate irrelevancies, and that is
what grammaticality is about.

	As for ambiguity, a GRAMMAR may show that a given sentence has
alternative interpretations and hence SHOULD be ambiguous, but has
nothing at all to say about the concept of ambiguity itself, or for
that matter about grammaticality.

	I am afraid student unrest, which I have centrally engaged in
as a student, from age 4 to 67 so far, is completely irrelevant to
worry about "judgments/intuitions of the native speaker". As a
linguist, like any other scholar, you are to be judged on what you
make of your raw data; native speaker judgments are always right, but
vary according to the questions asked. If you actually ask your class
in Syntax I the question, "is this sentence grammatical?" excuse me,
but you deserve your problems! They are only native speakers, who
always speak correctly but have no idea as to why they do: you are the
linguist.

	Now I sound to myself as if I am being distant, so let me just
point out that the problem here is that in fact most actual utterances
are ungrammatical (especially in Syntax I classes). That is why our
Master Leonard Bloomfield taught us that the fundamental assumption of
linguistics is that "some utterances are the same". Here enters
structuralism, and also grammar, which as yet has only a structural
reality -- its empirical reality is yet to be determined. Am I
responding to your query? Tell me if not! Yours, kvt

I am not sure if I am understanding Larry Hutchinson's posting
correctly, but if he is criticizing the way that (too) many linguists
belonging to the "generative" tradition and perhaps others as well
talk about judgements, intuitions, etc., in connection with
(un)grammaticality, then I have to agree. It is amazing to me how
many people for example do not seem to remember that grammaticality is
a theoretical concept and that it makes no sense whatever to try to
browbeat informants to answer queries like "I KNOW no one would ever
say this, but is it merely unacceptable or is it ungrammatical?". As
Chomsky has I think always made it rather clear (except in his very
earliest writings, when he toyed with the idea of behavioral tests for
grammaticality), it is not the informant's business to deal with this
distinction. It is a theoretical matter which is the investigator's
business and which depends BOTH on what the informant's judegements
are AND on what the grammatical theory is.
 
Although there are some arguments I can think of against the
particular way in which Chomsky conceptualized the distinction between
competence and performance, I find it difficult to believe that anyone
can really reject the idea that he achieved significant progress over
what existed before and that there has been no significant progress on
this issue since. Clearly, linguists and grammarians before Chomsky
also tacitly ignored whatever performance errors they encountered, but
were not aware of what they were doing. Chomsky in fact simply called
attention to what was being done on the sly before him and to the
enormous importance of it. As such, I like to refer to him as the
"discoverer of performance". Too few people seem to remember that he
also developed some of the earliest or maybe the earliest (Yngve being
the other contenderfor this prize) formal models of performance,
although he then abandoned that kind of research.
 
But what is really quite terrible is the way his really quite
reasonable views (although as I said I think they could stand some
improvement) have been turned into the way linguists elicit AND report
judgements, which hopelessly confused what is actually data with what
is theory and involves as Larry observes constant conflict with
students in intro classes and as I have observed (who has not?) with
informants where of course there should be none. The whole point is
that judgements are bits of performance and that intuitions themselvs
are inaccessible to observation, even to introspection, beacuse all
you get is judgements even from yourself.
 
But none of this means, or should mean, that we "eliminate" people
from linguistics. On the contrary, it is another of Chomsky's
contributions to make it clear that you cannot simply talk about
"language(s)", only about "speakers" whose linguistic behavior (I
would say) or mental representations (he would say) are more or less
alike. Again, I think one can object to certain specific points of
his conceptualization of this, but it clearly seems to me to be
superior to anything we had before and again I know of no better
formulation that is currently available.
 
Alexis MR