LINGUIST List 3.605

Mon 27 Jul 1992

Disc: Comparatives

Editor for this issue: <>

Directory

1. Dan Finer, Comparative Reconstruction
2. "Bruce E. Nevin", Comparatives: more on more

The examples in John Frampton's recent posting highlight the nature
of comparative reconstruction (filling in the ellipsis in the
comparative construction, that is). One of his examples is (1)
(= his (4c)):

(1) *I wonder whether there are more birds in the yard than Allan
 does.

According to Frampton's account, _more_ (or -er) moves across the
island boundary at an abstract level of representation (LF), and
this movement triggers a violation. While I agree with him that
LF-movement is the culprit here, one might wonder why it is that
_more_ must move out of the island in this case; why can't it
simply adjoin to the subordinate clause if it must take scope, and
if so, why is the sentence blocked? More generally, what forces
wide scope? The answer is related to the nature of the VP ellipsis
in the subordinate than-clause, and in these cases at least, the
ellipsis involves Antecedent-Contained Deletion (ACD). In order
for the sentence as a whole to be interpreted, the elided VP must
be assigned an interpretation as well, but the interpretation of
the missing VP depends upon the structure that it is in turn
contained in (the matrix VP; note the occurrence of _does_ in the
comparative clause). The way the interpretations of these VP's
unwind can be held responsible for the particular status of these
examples.

For concreteness, assume an analysis of ACD along the lines of May
'85 (_Logical Form_) in which the consituent containing the ACD
site undergoes LF-raising (see also Baltin (LI 18.4) for a
competing analysis and Larson & May (LI 21.1) for a reply. The
examples discussed here favor L&M, I believe). ACD looping is
avoided if, prior to interpretation, the comparative clause is
moved out of its containing VP and adjoined higher in the
structure. The remaining material in the vacated VP is then copied
into the empty VP. Since it is the matrix VP (headed by _wonder_)
that is recruited for the interpretation of the elided VP, the
comparative clause must be raised out of the island (hence **),
regardless of whether _more_ subsequently moves independently,
moves as part of the same operation that moves the comparative
clause, or simply remains in situ (there are probably proposals out
there for these three positions at least. Rochemont and Culicover
'90, and interpretive versions of Bresnan '73, and McCawley '88 are
possibilities). If the clause is not raised out of the island, on
the other hand, a reconstructive regress occurs.

Other relevant examples are in (2)-(4). Although (3) is ambiguous
(either the upper or lower VP can be reconstructed into the
comparative clause), (4c) has only one interpretation, for the same
reasons that (4b) and (1) are ungrammatical.

(2) a. Edward thinks Thomas is happier than Henry is.
 b. Edward thinks Thomas is happier than Henry does.

(3) Edward thinks Thomas works harder than Henry does

(4) a. Edward wonders why Thomas is happier than Henry is.
 b. *Edward wonders why Thomas is happier than Henry does.
 c. Edward wonders why Thomas works harder than Henry does.

(4a) and (4c) show that the island has no effect if it is possible
to reconstruct the elided VP locally (which yields narrow scope for
the comparative here).

 Dan Finer
 SUNY Stony Brook

In Linguist 3.590 (7/18/92), John Frampton writes:

| In response to Nevin's "incremental paraphrase" (my term) analysis of
| comparatives in Linguist List 3-567.
|
| The analysis of
|
| (1) I think this is rarer than Allan does.
|
| can't be correct. Consider
|
| (2) I think there are more birds in the yard than Allan does.
|
| Nevin's proposal would require the paraphrase
|
| (3) I think there are birds in the yard more than Allan does.
|
| But this is not a paraphrase.

This is true. In the analysis I gave, the amount of "my" thinking is
compared with the amount of Alan's thinking; sentence (1) compares the
number of birds as "I" think it to be with the number as Alan thinks it
to be. Let me wipe the egg off and try to do better.

First, an assertion without the "I think" (which comes only near the end
of the derivation):

11a. There are birds in the yard (mounting) up to a number.

Then the comparison with what Alan thinks about the number of birds:

11b. There are birds in the yard up to a number;
 said number is more than a number;
 (amounting) to said number Alan thinks there are birds in the yard.

(I will comment on the odd word order of the last conjunct after the end
of the derivation.)

Low-information zeroing (Z. S. Harris (1982) _A Grammar of English on
Mathematical Principles_ 3.5 [GEMP] pp. 150-186) reduces (11b) to (11c)
and thence to (11d):

11c. There are birds in the yard up to a number which is more than a number
 to which Alan thinks birds are in the yard.

11d. There are birds in the yard more than a number
 to which Alan thinks there are birds in the yard.

"More" is transposed to the left of birds as its modifier, as though an
adjective:

11e. There are more birds in the yard than a number
 to which Alan thinks there are birds in the yard.

(This follows the ordinary derivation of adjectives from relative
clauses.) Additional low-information zeroing yields:

11f. There are more birds in the yard than
 Alan thinks there are birds in the yard.

Only at this point is "I think" asserted of the whole sentence, with the
comparison and transposed position of "more" already in place:

11g. I think there are more birds in the yard than
 Alan thinks there are birds in the yard.

The last step is a familiar low-information reduction:

11h. I think there are more birds in the yard than Alan does.

Sentence 11h = sentence 1.

The dependency schematic that I offered was misleading because it
confounded operator-argument dependency with head-modifier dependency

I promised to say something about the inverted word order seen in (11b).
The basis for this is the possibility of placing a sentence modifier at
various positions, even ahead of a later-entering (higher) operator.
Thus:

12a. I saw birds amounting to a dozen.
 b. Amounting to a dozen I saw birds.

13a. I thought I saw birds amounting to a dozen.
 b. Amounting to a dozen I thought I saw birds.

Somewhat peculiar the word order of the (b) sentences seems, poetic
perhaps, or suggesting a Yoda-ish or maybe Yiddish-ish dialect, but
English surely it still is. (More on marginal sentences at the end of
this post.)

This is relevant because only those things may be relativized that may
appear in front position in a sentence conjoined under a paratactic
operator whose form is the reduced intonation of an interruption or an
aside (written with dash for the former, semicolon for the latter).
Contiguity or near contiguity is a condition for reduction to relative
pronoun.

| I think Nevin is correct in thinking that what is at stake here is the
| ability of /more/ to move out of its clause and appear, in some
| representation, in the main clause. (This is implicit in the sequence
| of paraphrases which he gives.) Consider

| (4) a. *I regret that there are more birds in the yard than Allan does.
| b. *I think there aren't more birds in the yard than Allan does.
| c. *I wonder whether there are more birds in the yard than Allan does.

| Verbs like /regret/, negative elements, and wh-islands are typical
| obstacles to movement.

In the non-sentences of (4), "regret" etc. enter on I plus one of the
following non-sentences, respectively:

 (4') a. *There are more birds in the yard than Allan does.
 b. *There aren't more birds in the yard than Allan does.
 c. *There are more birds in the yard than Allan does.

The problem is not with the higher operator (regret, think, wonder),
which must enter late in the derivation as in (11g), but rather with the
lack of a derivation for the "more" in the non-sentences of (4').
(I have a long ms. on island phenomena etc. in operator grammar that I
really should polish up for publication, but have not had time.)

| The problem seems to be analagous to "neg raising".

The examples involving negation are not entirely parallel because of
some well known peculiarities with the placement of the negative
particle. In operator grammar, the negative particle is taken as a
reduction of "I deny" entering on a sentence. (See GEMP Chapter 7,
Forms due to metalinguistic operators, especially at p. 326.)

Several others sent interesting and helpful comments off-line, for which
I thank them. These comments suggest that more needs to be said about
low acceptability at certain points in some of these derivations. It
may be helpful to bear in mind that the aim is what one could call a
semantic representation, using only resources that are well established
for the language. There are gaps in the network of elementary sentence-
differences resulting from analysis. If one can fill the gap using only
vocabulary, operator-argument dependencies, and reductions that are well
attested elsewhere (with fully acceptable sentences), then it is
justified, even if all the sentences (or "infra-sentences") of the set
thereby produced seem peculiar or unlikely. The criterion for mapping
from one subset of sentences to another (where members of each set are
all of the same form, and the two sentence forms differ by an elementary
sentence difference) is that any pairwise acceptability differences in
one set are not reversed for the corresponding sentences of the other
set. This criterion is not violated when all the members of a set
together are reduced to marginal acceptability.

This provides a principled way of dealing with the fact that there is no
well defined "set of sentences" for a language. Language change through
time and geographic and social variation involve forms at the margins of
acceptability. Frequently it is possible to identify forms that are on
their way out of favor or which may be on their way in to full
acceptability. I recall Harris drawing a diagram with two concentric
circles. The inner circle defined an "informationally complete
sublanguage" that was the product of operators being asserted of their
arguments. The outer circle included all the paraphrases yielded by
reductions (and all the informational degeneracies such as ambiguity).
A wavy line within the outer circle indicated the (ill defined) bounds
of acceptability, subject to change over time and variation from one
area or community to another, and even from one subject-matter domain to
another. (A closer approach to well-definition is possible within the
sublanguage grammar for a particular subject matter.)

By regularizing beyond the language as given at its present historical
stage or in certain dialects, one is able to account for its
informational capacity (and for the information in particular texts) in
a detailed and precise way, in terms of the assertion of operator words
on argument words plus reductions of informationally redundant forms.
(The reductions are essentially for compactness and efficiency but
concurrently for conformity to social conventions of various sorts.)

Underlying this is Harris's important methodological point that there is
no standpoint outside of language from which one may describe language.
Whatever means one devises for describing language depends upon the
informational capacity of language itself for its function. To account
for that capacity, one must use the resources of language itself. And
because of this, one must take great care that the machinery of
description does not introduce extraneous redundancies and dependancies
that would obscure those by which language constitutes information.
(See _Language and Information_, New York: Columbia University Press
1988, and _A Theory of Language and Information: a mathematical
approach_, Oxford: Clarendon 1991.)

	Bruce Nevin
	bnbbn.com