Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 02 No. 62
NL-KR Digest (6/29/87 18:25:35) Volume 2 Number 62
Today's Topics:
Re: S/O Raising
----------------------------------------------------------------------
Date: Fri, 26 Jun 87 01:51 EDT
From: Michael P. Smith <mps@duke.cs.duke.edu>
Subject: Re: S/O Raising
In article <1176@houdi.UUCP> marty1@houdi.UUCP (M.BRILLIANT) writes:
>In article <9799@duke.cs.duke.edu>, mps@duke.cs.duke.edu (Michael P. Smith) writes:
>> A quick review: in S/O raising, the subject of a that-clause
>> complement is "raised" to the direct object of the matrix clause, and
>> the verb in the complement put into the infinitive. For example, S/O
>> raising applied to
>> Alice expected that Bob would be chosen.
>> yields
>> Alice expected Bob to be chosen.
>
>which in turn can be turned into a passive with the original subject,
>now an object, as subject again:
> Bob was expected to be chosen
by Alice.
>> The puzzle is that some verbs in the matrix permit S/O raising, and
>> others do not.
>A further puzzle is that some that do not permit S/O raising do permit
>the passive transformation. Here's the original sample list with Y
>after those that allow the passive, and N for those that don't:
>
>> PERMIT RAISING PROHIBIT RAISING
[...]
I don't agree with all your decisions above, but if I understand the
phenomena you're describing, it is very interesting. 'Say' seems a
clear case: contrast
(1) * Alice said Bob to be chosen.
and
(2) Bob was said to be chosen by Alice.
(We should probably prefer "Bob was said by Alice to be chosen." to
make the scope of the prepositional phrase clear, but certainly a
meaning of this sentence is the same as the original. And were the
sentence "Bob was said to be rude by Alice." there would be no ambiguity.)
What's interesting is that on the standard theory, (2) would be
derived from (1) by passivization. (I say subject to correction by
those who know more about it than I do.) I don't know about the newer
theories, but I should think that deriving (2) from (1) by passivization
would be a natural move for any theory in which 'Bob' in (1) is a
raised object. Now, as Jeff Goldberg points out in <307@su-
russell.ARPA>, this is *not* the case for the GB analysis. So I won't
pretend to know how (2) would be derived according to GB. You want to
help us out here, Jeff?
Anyway, if (2) is derived from (1) by passivization, but (1) is deviant,
what we have here is a case of an obligatory passivization. This is
interesting to me because I had thought active/passive was only a
matter of emphasis and tone.
>> The problem is to find a principled distinction between the two
>> classes of verbs.
Now four, Marty notes.
>What about just saying that there are these classes of verbs? After
>all, what is it that defines a verb, anyway, other than that it is
>allowed in certain constructions? That is, can you find a "principled
>distinction" between verbs and non-verbs? Don't just fall back on the
>grade-school definition that "a verb is a word that expresses an action
>or state of being": it's not very accurate.
Gee, what's wrong with Aristotle's definition:
A verb is what additionally signifies time [i.e., it has tense],
no part of it being significant separately [i.e., it's a morpheme],
and it is a sign of things said of something else [i.e., it
functions as a predicate and not the subject in a sentence].
De Interpretatione 3
OK, there's a lot of room for improvement in the third part. My point
is that Aristotle managed to say something more insightful than "a
verb is a word that expresses an action or a state of being" over 20
centuries ago, and I'm sure that we can do better still today.
By 'principled' I didn't mean 'perfect'; I meant one that appeals to
theoretical principles and is not ad hoc.
I get the impression in your reply as well as in Jeff's that you want to
say that it's just a brute fact that some verbs permit raising, or
raising and passivization, and others don't. Well, I'm not happy with
this, but I could learn to live with it, if it were just *one* brute
fact. After all, explanations have to stop somewhere, and I expect that
they end more abruptly for human practices than for say, chemical
properties.
So it *might* turn out that verbs with an odd number of letters
prohibit raising, while verbs with a prime number of letters permit
passivized raised forms, and that there's no more to be said. (Even
as I write this I realize that I could never believe anything like it,
but never mind.) I'm much more reluctant to believe that there could
be dozens of brute facts here, two for each verb: one concerning
whether it permits raising, and another as to whether it permits the
passivized raised form. This stronger disinclination is based on more
than aestheic judgment. I believe that human languages cannot be
theoretically extravagant and still be learnable.
One reason for believing this would be to believe with Jerry Fodor in
LANGUAGE OF THOUGHT that people learn a language by constructing a
theory of it. I don't believe this, even when it is explained that
said theory construction is entirely unconscious. But I do believe
that the situations of the language learner and the linguist are
analogous enough that a language is learnable iff a theory of it is
possible (cf Donald Davidson, "Theories of Meaning & Learnable
Languages"). And if every time a situation like this comes up it is
'explained' through a myriad of subcategories or otherwise involves
dozens of brute (= irreducible) facts, that, it seems to me, would
make the language inaccessible to theoretical linguists and infants
alike.
Some point to the complex interactions of all levels of language --
phonology, syntax, semantics, pragmatics -- as reasons for the
theoretician to despair. On the contrary, I think that the very
complexity of these interactions means that the amount of irreducible
complexity within each cannot be as great as it seems. People do
learn language, after all, and readily.
A taste of the complexity: Marty says that his linguistic intuitions do
not permit 'remember' to have a raised object. Admittedly,
(?) Alice remembered Bob to be (have been?) chosen.
is a little stilted. But what about
Alice remembered Bob to be a wonderful cook.
Actually, the construction works best if you pretend you're reading
P.G. Wodehouse:
Suddenly, Bertie remembered his opponent to be the current
holder of the West End boxing championship.
It seems the naturalness, if not the grammaticality of S/O raising
depends not only of the matrix verb, but the complementizer as well.
Anyway, to my ear, the latter two active raises are ok, but the
passives don't work. Actually the first fares best:
Bob was remembered by Alice to have been chosen.
sounds ok if we sneak a little stress onto the subject. But that's
cheating, since the sentence would only work in the specialized
context of changing or re-affirming a conversational topic.
>M. B. Brilliant Marty
>AT&T-BL HO 3D-520 (201)-949-1858
>Holmdel, NJ 07733 ihnp4!houdi!marty1
-----
Michael P. Smith "The worlds revolve like ancient women
ARPA: duke.cs.duke.edu Gathering fuel in vacant lots." T.S. Eliot
------------------------------
From: M.BRILLIANT
Subject: Re: S/O Raising (long)
In article <9831@duke.cs.duke.edu>, mps@duke.cs.duke.edu (Michael P. Smith) writes:
> In article <1176@houdi.UUCP> marty1@houdi.UUCP (M.BRILLIANT) writes:
> >What about just saying that there are these classes of verbs? After
> >all, what is it that defines a verb, anyway, other than that it is
> >allowed in certain constructions? That is, can you find a "principled
> >distinction" between verbs and non-verbs? Don't just fall back on the
> >grade-school definition that "a verb is a word that expresses an action
> >or state of being": it's not very accurate.
>
> Gee, what's wrong with Aristotle's definition:
> A verb is what additionally signifies time [i.e., it has tense],
> no part of it being significant separately [i.e., it's a morpheme],
> and it is a sign of things said of something else [i.e., it
> functions as a predicate and not the subject in a sentence].
> De Interpretatione 3
I don't want to fight with mps, because we are mostly in agreement, I
think.... Aristotle's definition of a verb is syntactic, at least as
interpreted by mps, and the grade-school definition is semantic.
I tacitly assumed that "principled distinction" meant "semantically
based." A lot of syntactic rules as taught in foreign language courses
have semantic components: words that signify X are inflected Y-wise,
etc. I was suggestng that we were looking at a purely syntactic
phenomenon with no consistent semantic component.
Maybe we do have a controversy brewing:
>
> I get the impression in your reply as well as in Jeff's that you want to
> say that it's just a brute fact that some verbs permit raising, or
> raising and passivization, and others don't. Well, I'm not happy with
> this, but I could learn to live with it, if it were just *one* brute
> fact..... I'm much more reluctant to believe that there could
> be dozens of brute facts ... I believe that human languages cannot be
> theoretically extravagant and still be learnable.
I've been told that from the point of view of historical linguists,
so-called "primitive" languages - including those of contemporary
"primitive" cultures as well as those of cultures ancestral to our own
- have more complicated grammars than the languages we are most
familiar with. That seems to contradict the principle of learnability
through elegance.
M. B. Brilliant Marty
AT&T-BL HO 3D-520 (201)-949-1858
Holmdel, NJ 07733 ihnp4!houdi!marty1
------------------------------
From: Jeffrey Goldberg
Subject: Re: S/O Raising
>In article <9830@duke.cs.duke.edu> mps@duke.UUCP (Michael P. Smith) writes:
>In article <307@su-russell.ARPA> goldberg@su-russell.UUCP (Jeffrey Goldberg) writes:
>>In article <9799@duke.cs.duke.edu> mps@duke.UUCP (Michael P. Smith) writes:
[So, ">>>" prefixes Michael's original posting, ">>" is me, ">" is
Michael again. -J. Goldberg]
>>(1)
>>> Alice expected that Bob would be chosen.
>>>yields
>>(2)
>>> Alice expected Bob to be chosen.
>>What I can tell you, is that in none of the modern theories is (2)
>>derived from (1).
>I don't know if you meant to include the (extended) standard theory,
>but in that theory as well (2) is not derived from (1), but both from
>a common deep structure.
I did not mean to include the Extended Standard Theory (EST) which
is no longer a "current" syntactic theory. It was replaced by the
Revised Extended Standard Theory (REST) which in turn was replaced
by the Over-Extended Standart Theory (GB).
>I don't see the semantic difference between (1) & (2); further, the
>subjunctive seems to be required in (1) by the sequence of tenses
>rather than any subcategorical peculiarities of 'expect'. Maybe I
>should have said
>(1') Alice expects that Bob will be chosen.
>and
>(2') Alice expects Bob to be chosen.
The semantic difference is not a truth conditional one here.
(there is no situation in which one could be false and the other
true.), but there are fairly clear discourse differences here. (At
least to me. Having something to do with the role of the object
"Bob".)
>It's rude to push after Jeff admits that an appeal to
>subcategorization by itself is no answer, but -- how does the
>subcategorization story go? Is it that each failure to raise has its
>own reason, or is it that verbs otherwise nearly synonymous sometimes
>differ along some single dimension that explains why one allows raises
>and the other does not? See my reply to Marty on why I don't think
>every failure to allow raising can have its own story.
It is the second. Verbs that are near synonymous may differ
syntactically. Consdier "eat" (which allows object ellipsis, "He
ate." as opposed to "devour" which does not. *"He devoured." Or
consider Dative Shift (aka, double object constructions) with
"give" and "donate". "He {gave/*donated} the alumni association $100."
>Michael P. Smith "It could also be said that a man thinks
>ARPA mps@duke.cs.duke.edu when he learns in a particular way."
--
Jeff Goldberg
ARPA goldberg@russell.stanford.edu
UUCP ...!ucbvax!russell.stanford.edu!goldberg
------------------------------
From: Jeffrey Goldberg <goldberg@russell.stanford.edu>
Subject: Re: S/O Raising (Too long)
>In article <9831@duke.cs.duke.edu> mps@duke.UUCP (Michael P. Smith) writes:
>In article <1176@houdi.UUCP> marty1@houdi.UUCP (M.BRILLIANT) writes:
>>In article <9799@duke.cs.duke.edu>, mps@duke.cs.duke.edu (Michael P. Smith) writes:
>I don't agree with all your decisions above, but if I understand the
>phenomena you're describing, it is very interesting. 'Say' seems a
>clear case: contrast
>(1) * Alice said Bob to be chosen.
>(2) Bob was said to be chosen by Alice.
>(We should probably prefer "Bob was said by Alice to be chosen." to
>make the scope of the prepositional phrase clear, but certainly a
>meaning of this sentence is the same as the original. And were the
>sentence "Bob was said to be rude by Alice." there would be no ambiguity.)
>What's interesting is that on the standard theory, (2) would be
>derived from (1) by passivization.
Yes, this is an interesting problem.
>I don't know about the newer
>theories, but I should think that deriving (2) from (1) by passivization
>would be a natural move for any theory in which 'Bob' in (1) is a
>raised object. Now, as Jeff Goldberg points out in <307@su-
>russell.ARPA>, this is *not* the case for the GB analysis. So I won't
>pretend to know how (2) would be derived according to GB. You want to
>help us out here, Jeff?
In the normal Raising to Object cases in GB the "Object" is in the subject
position of the infinitival clause. Every overt NP (noun phrase) must be
assigned (abstract) Case by a legal Case assigner. Subjects are normally
assigned Case by the INFL (inflectional element of the clause), but
infinitivals don't have enough meat in them to be legal Case assigners. A
verb like "expect" can subcategorize for an S (as opposed to an S' (S-bar)),
which will allow (given other definitions of the theory) the verb "expect" to
assign Case to the subject of the infinitival complement clause. A number of
object properties depend on being assigned Case. So, in some respects the
subject of the infinitival will behave like an Object of "expect".
Passive in GB is done simply by stating that passive verbs are unable to
assign Case, and do not give the subject position a Thematic Role (a role in
the meaning). The theory is such that this will get passive to work (more or
less) and the important thing is that passive only is concerned with the
assignment of Case and not the grammatical function of the thing being
passivized.
In GB, there is a slighly different deep structure for a passive sentence then
there is for the active. The passive verb is already in place. With its
characteristic inability to assign Case to its object (or subject of its
complement infinitival S).
So, in GB one would need to stipulate a lexical gap in the paradigm for "say".
There is the active "say" that takes an S'.
There is a passive "said" that takes an infinitival S.
There is no active "say" that takes an infinitival S.
Please note that this paradigm gap is similar to the gap suggested
by the facts about the verb "rumor" (left as an exercise to the
reader.) So, if one were willing to allow idiosyncratic syntactic
facts to be stated about the subcategorizations of particular verbs
then there is no immediate problem with "say" for the modern
theories. ST has exactly the problem you raised (pun intended).
>Anyway, if (2) is derived from (1) by passivization, but (1) is deviant,
>what we have here is a case of an obligatory passivization. This is
>interesting to me because I had thought active/passive was only a
>matter of emphasis and tone.
Essentially, you have it right. But the facts are the facts. We have a
passive without the corresponding active. All of the modern theories have
mechanisms that will allow this, but nothing that is particularly satisfying.
>I get the impression in your reply as well as in Jeff's that you want to
>say that it's just a brute fact that some verbs permit raising, or
>raising and passivization, and others don't. Well, I'm not happy with
>this, but I could learn to live with it, if it were just *one* brute
>fact.
An explaination would be nice. All we have now is a mechanism. As your
posting that started all this mentioned Perlmutter and Soames say that the
characterization of the classes of verbs that have particular syntactic
properties is still an open question. They understate the fact here. It is a
complete mystery. The people who work on this stuff work on what is called
"Lexical Semantics". I am sorry that I can't provide a good reference there.
> And if every time a situation like this comes up it is
>'explained' through a myriad of subcategories or otherwise involves
>dozens of brute (= irreducible) facts, that, it seems to me, would
>make the language inaccessible to theoretical linguists and infants
>alike.
[...] We know nothing about what people are capable of learning, and a
prioristic reasoning of this sort is not particular helpful.
Consider how much arbitrary information must be learned (the meanings and
pronunciations) of the entire lexical. Why would a number of syntactic
idiosyncracies suddenly make this to much to learn? [I should point out that I
am in a minority here with this attitude toward learnability arguments.]
However, If you find that similar verbs in other languages show similar
oddnesses, then you have a very strong case for saying that these things are
not purely accidental. There would have to be some systemicity to it.
>>M. B. Brilliant Marty
>>AT&T-BL HO 3D-520 (201)-949-1858
>>Holmdel, NJ 07733 ihnp4!houdi!marty1
>Michael P. Smith "The worlds revolve like ancient women
>ARPA: duke.cs.duke.edu Gathering fuel in vacant lots." T.S. Eliot
--
Jeff Goldberg
ARPA goldberg@russell.stanford.edu
UUCP ...!ucbvax!russell.stanford.edu!goldberg
------------------------------
End of NL-KR Digest
*******************
