LINGUIST List 3.567

Tue 14 Jul 1992

Disc: Comparatives

Editor for this issue: <>

Directory

1. "Bruce E. Nevin", Comparatives

>From: Allan C. Wechsler <ACWRIVERSIDE.SCRC.Symbolics.COM>
> From: jcjextro.ucc.su.OZ.AU (Jason Johnston)
>
>I agree that these sentences are all acceptable. But if this is true,
>what derivation explains (1)? I haven't studied enough syntax to know
>how to say this technically, but there is something here that needs
>to be explained.
>
> (1) I think this is rarer than Allan does.
>
> (2) I think this is rarer than Allan thinks it is.
> (3) I think this is rarer than I used to believe [it is].
> (4) I think this is rarer than people generally admit [it is].
>
>(5) Moroccans think cheese is more acceptable than Algerians do.
>(6) Moroccans think cheese is more acceptable than Algerians are.
>(7) Moroccans think Libyans work harder than Algerians do. (Ambiguous.)
>(8) I think Mary believes John to be more intelligent than I do.

You are right that this is not the most usual form of comparative, viz:

(9) Marble is harder than chalk.

For ordinary comparatives such as (9), there is available a set of other
sentences such as the following:

(9) Marble is harder than chalk.
(9a) Marble is hard more than chalk.
(9b) Marble is hard more than chalk is.
(9c) Marble is hard more than chalk is hard.
(9d) Marble is hard to a degree more than the degree to which chalk is hard.
(9e) Marble is hard to a degree which is more than the degree to which
 chalk is hard.
(9f) Marble is hard to a degree; said degree is more than a degree; to
 said degree chalk is hard.

These sentences differ from one another by a series of small
differences. These differences are amenable to the interpretation that
a word or words in one such sentence are reduced to zero in the next
sentence in the series. (Between 9a and 9, "more" is reduced to "-er".)

Sentences (9d-9f) are awkward and one is unlikely to hear or read them
in actual usage precisely because less awkward and more conventional
alternatives like (9a-c) are available. However, they are undoubtedly
English, and they have the merit of being more explicit than their
fellows.

Now look at sentence (1):

(1) I think this is rarer than Allan does.
(1a) I think this is rare more than Allen does.
(1b) I think this is rare more than Allen thinks this is rare.
(1c) I think this is rare to a degree more than the degree to which
 Allen thinks this is rare.
(1d) I think this is rare to a degree which is more than the degree
 to which Allen thinks this is rare.
(1e) I think this is rare to a degree; said degree is more than a
 degree; to said degree Allen thinks this is rare.

This makes the differences between (9) and (1) explicit. A different
kind of thing is being compared. In (9) etc, we see a first-order
assertion "marble is hard" (an operator word entering on a zero-order
word or noun) being said to be to a degree, which is then compared to
the degree asserted of the same first-order operator "hard" entering on
a different noun, "chalk". The following dependency tree expresses this
more clearly by showing the operator-argument dependencies graphically:

 degree --- hard --- marble
 /
more
 \
 degree --- hard --- chalk

(The tree is turned on its side to make it easier to draw in ASCII, and
it omits the operator indicator "is" and argument indicator "than".)

Compare this with the sentence set of (1). Here, we have a second-order
assertion, the operator word "think" asserted of the pair "I, hard", and
it is this "think" that is said to be to a degree, which is then
compared to the degree asserted of a repetition of "think" on a
different argument pair (only the first member of the pair differs, "I"
vs. "Alan"). Again, the dependency tree is clearer than the verbal
explanation:

 		 I
 			 /
 degree --- think
 / \
 / rare --- this
 /
more
 \
 \ Alan
 \		 /
 degree --- think
 \
 rare --- this

The argument under "more than" need not be "degree." For example, the
appropriate word is surely "amount" or "number" in e.g. "This hen lays
more eggs than that one." Indeed, in (1) one might also say "hard in an
amount," "to an amount," etc. Also consider:

(10a) A 6' tall woman is taller than a 6' tall man.
(10b) A 6' tall woman is tall in a sense which is more than the sense
 in which a 6' tall man is tall.
(10b) A 6' tall woman is tall in a sense; said sense is more than a
 sense; in said sense a 6' tall man is tall.

Absent the word sharing under more, the more conventional forms of the
comparative are not available. However, the unreduced sentences can be
quite sayable:

(11a) ??I think this is rarer than John is willing to pay for it.
(11b) I think this is rare more than John is willing to pay for it.
(11c) I think this is rare to a degree which is more than the degree
 to which John is willing to pay for it.

Usually, such sentences without word repetition are uninterpretable:

(12a) * I think orchids are rarer than Dewey understood the importance of
 feedback.
(12b) * I think orchids are rare more than Dewey understood . . . feedback.
(12c) ? I think orchids are rare to a degree which is more than the degree
 to which Dewey understood . . . feedback.

The differences between (11) and (12) show that there are two sorts of
factors affecting acceptability. The reduced forms (12a,b) are starred
because word repetition is a formal requirement for the reduction of
redundant words to zero. The fully explicit (and awkward,
unconventional) form (12c) is only queried and not starred because the
problem here is not grammatical but pragmatic: how does one compare the
degree or amount of perceptions so disparate as the rareness of orchids
(or my thinking that orchids are rare--the sentence is ambiguous) and
Dewey's understanding of the importance of feedback?

This analysis is based on work of Zellig Harris, e.g. _A Grammar of
English on Mathematical Principles_. This is not transformational
grammar (TG). It might be called operator grammar (Stephen Johnson's
term), assertion-reduction grammar, or entry-and-reduction grammar.
Reductions take place (usually optionally) when an operator word enters
on its arguments. For example, the reductions yielding (1) might be
graphically represented as follows (though I do not suggest that such a
graphical metalanguage is essential):

 degree --- hard --- marble 0 --- hard --- marble
 / /
more than ==> er than
 \ \
 degree --- hard --- chalk 0 --- 0 --- chalk

The dependency structure of operators and arguments represents the
information in sentences. Additional work of sublanguage analysis
results in a representation of information in discourse. (Harris,
_Language and Information_, _A Theory of Language and Information: a
mathematical approach_, Harris et al. _The Form of Information in
Science_, Ryckman, _Grammar and Information: an investigation in
linguistic metatheory_.)

	Bruce Nevin
	bnbbn.com