LINGUIST List 4.126

Tue 23 Feb 1993

Disc: Subject/Object Asymmetry

Editor for this issue: <>

Directory

1. , Re: 4.122 Subject/Object Asymmetry
2. Joe Stemberger, Re: 4.122 Subject/Object Asymmetry
3. , 4.122 Subject/Object Asymmetry

Message 1: Re: 4.122 Subject/Object Asymmetry

Alexis Manaster-Ramer writes:

>David Pesetsky suggests that it does not matter if long-held views
>are shown to be wrong unless this "entailed the incorrectness of some
>more general claim". Now, I have heard this kind of talk so often
>it does not surprise me to hear it again, but I wonder exactly what
>that is supposed to mean. For EVERY claim that has ever been made,
>and then refuted, you can find a "more general" claim which we may
>safely proclaim is what we really care about. If, for instance,
>we discover that no feature of language is innate, that there is no
>such thing as syntax, that there are no transformations or theta-roles,
>etc., etc., I am sure someone will point out that these really were not
>the interesting claims of Chomskyan linguistics after all.

This in turn is a "kind of talk" that *I* have heard so often that it
does not surprise *me* to hear it again, though it saddens me. The
script seems to go as follows: If X sticks with an idea demonstrated to
be ill-conceived by Y, X is taken to task for ignoring Y's work. If X
does pay attention to Y, offers a new idea that takes Y's work into
account, but doesn't throw all the babies out with the bathwater, X is
taken to task -- but this time for responding to Y's work by means other
than leaving the field. I find it remarkable that the same people that
get mad because X allegedly adheres to some putative dogma called
"Chomskyan linguistics" get even madder when X *doesn't* adhere to the
dogma called "Chomskyan linguistics". The logic seems to be "won't you
please stand still so we can do you in!"

Now to the specific point. What you mistake for bad behavior is merely
the rational step of pointing out what remains of a previously held
network of ideas and claims when one of them is excised or replaced by
another. Whether replacing one or more planks in a theoretical ship
entitles you to keep calling the ship by the same name is indeed Plato's
problem, but it is not Plato's Problem -- so I don't see why we should
worry about it.

>
>In addition, as has often been pointed out, it is by no means clear
>whether many of the more general claims that Pesetsky is referring to
>CAN in principle be refuted. For example, the general claim that the
>whole subject-idiom issue started with was twofold: (a) it was claimed
>that idioms had to have the same constituent structure as the rest of
>the language and (b) that there is such a thing as a VP.
>Now, once it was pointed out that there are plenty of idioms that do
>not fit this picture, suddenly it turned out that this was not the issue
>at all, and that only the asymmetry between the allegedly impossible
>S-V idioms and the possible V-O ones that was relevant. [...]

I don't follow. Suppose we have indeed discovered that there is no
subject-object asymmetry in idioms. Then either there is something
wrong with (a) [once (a) is reformulated to make sense; see Everaert's
note for this] or there is something wrong with (b). There is extensive
evidence for (b). Therefore I would choose to maintain (b) and abandon
(a). If I read you correctly, you would abandon both. This doesn't
seem sensible. Idioms were supposed to be an argument for the relevance
of VP to theta-assignment, given some particular assumptions about
idioms and their relation to theta-assignment. If the empirical side of
the argument fails, the blame can be laid at the feet of too many
premises of the argument for anyone to make facile judgments about the
status of the conclusion. That's just a fact about the logical
structure of the Marantz/Chomsky idiom argument, not an example of
theoretical squirming. I don't think anyone has changed their judgment
about what is really "the issue". The question is whether the empirical
claims can be sustained, whether anything interesting about idioms has
been learned, and what the general theory should look like now.

Enough.

-David Pesetsky

Message 2: Re: 4.122 Subject/Object Asymmetry

I think that our failure to come back and explicitly reject the theories
that made a wrong prediction about idioms merely highlights a general
concern that I have with how we gather support for a hypothesis.

When we try to explore the ramifications of the hypothesis, there are at
least three possibilities:

1) the hypothesis makes a necessary prediction about some data; if the
 prediction is false, so is the hypothesis. It must be rejected.

2) the hypothesis doesn't by itself make a prediction about the data, but,
 in combination with another assumption, it might. In fact, if the data
 are in accord with this extended prediction, it would constitute
 impressive evidence for the hypothesis. BUT, if the data are NOT in
 accord with the extended prediction, it's no big deal. It's not the
 hypothesis that was at fault, but the additional assumption that we
 made.

3) There are some data that we COULD try to explain using the hypothesis.
 There are also 15 other explanations for the data. But what the heck,
 the hypothesis is just sitting there, so let's use it to explain the
 data. If that attempt fails, no big deal. We'll use one of the other
 15 ways to explain it.

The first possibility is the one that we have to be really concerned about,
because it really does constitute a test of the hypothesis. And if that
test fails, we have to reject the hypothesis.

But the second two possibilities really aren't a problem for the
hypothesis. If we fail here, it's just that we were being over-zealous in
trying to find ways to use the hypothesis.

My concern is that we usually don't differentiate these three
possibilities. We don't make clear in our papers whether it really makes
much difference whether we're right or wrong about how to interpret the
data that we're presenting. Readers have to look really closely at the
paper and think it through carefully to see whether counterevidence means a
real problem for a theory, or a trivial change. I think it would be real
nice if we were clearer about the stakes involved in the papers.

I also think that there is a reason that we don't address this very often.
I don't think that the first possibility arises very often. We don't often
make claims that really WILL lead to rejecting the theory (or major
parts thereof) if we're wrong. Most of our work tends to involve the second
and third possibilities. And even when it's the second possibility, the
claim is usually qualified. In phonology, most of our hypotheses predict
that something is POSSIBLE but not REQUIRED; a given data set may not show
the predicted possibilities because other possibilities also exist. (For
example, labial consonants CAN block rounding harmony in vowels, if [Labial]
is spreading, but don't HAVE TO block rounding harmony, since it may be
that the feature [round] is spreading. But what's important is that labials
CAN block the harmony, even if they don't have to.)

There's a recent example from language acquisition. There was a claim that
there was u-shaped learning of irregular forms: children initially say
FELL, then regularize to FALLED, then ultimately say FELL again. This was
viewed as impressive evidence for rules. Now, there's a lot of debate over
whether learning really is u-shaped here. If it IS, it's usually
quite subtle. And it may depend on our exact definitions of what we mean by
"u-shaped" learning. We don't have the answer yet, but suppose that this
claim about development turns out to be COMPLETELY false. That doesn't mean
that we'll abandon rules. It wasn't a NECESSARY prediction. The prediction
came from OTHER assumptions, which we'll now change. No big deal. We just
have to look elsewhere for our smoking-gun proof for rules. It just
would've been nice if everybody realized that it wasn't a necessary
prediction in the first place.

I suspect that the predictions about idioms involved the second
possibility (but maybe the third). If it failed, no big deal. And maybe
some people will find that frustrating.

---joe stemberger

In light of Martin Everaert's latest posting (as well as some
earlier ones by other people), I just thought it might help
focus the discussion to point out that there are two
different definitions of subject idioms floating around.

One involves any idiom which is less than a whole clause
but includes the subject, e.g., The jury is out on X, where
only the object X is non-idiomatic (free, variable, or whatever
you want to call it). It is true that for many years the
existence of such idioms was quite generally denied by generative
syntacticians who wrote about idioms and constituent structure,
but since 1982 counterexamples such this one have been well known
(thanks to Bresnan primarily). The other definition, which arose in
response to Bresnan's work, requires a subject idiom to consist
of a subject and a simple transitive verb, and again this kind
of idiom is often claimed to be impossible (e.g., by Chomsky
in "Knowledge of Language" and by Marantz).

Thus, 'The jury is out on X' is NOT a subject idiom under the
more recent definition. Only examples like the ones I cited
in earlier postings (like 'The spirit moves X', 'X-acc reitet der Teufel',
etc.) are thus counterexamples to the more recent claims.