LINGUIST List 3.486

Fri 12 Jun 1992

Disc: Comparatives

Editor for this issue: <>


Directory

  1. mark, Re: 3.472 Comparatives
  2. Richard Sproat, 3.478 Comparatives

Message 1: Re: 3.472 Comparatives

Date: Thu, 11 Jun 92 14:33:45 ESRe: 3.472 Comparatives
From: mark <markdragonsys.com>
Subject: Re: 3.472 Comparatives

A slight correction to Barbara Need's reading of my comment on
"My car is redder than orange": I didn't mean to say that its
readings were *reduced* forms of the somewhat longer sentences
(her 1a, 1c) by means of which I distinguished those readings. I
meant those sentences only as *paraphrases* of the readings. The
misunderstanding may be because the two sentences
 My car's color is closer to red than to orange (1a)
and
 My car's color is closer to red than orange is (1c)
can both reduce to
 My car's color is closer to red than _ orange _ ,
which is therefore ambiguous.

I am not currently interested in theories of how (or whether) one
sentence is reduced to another!

 Mark A. Mandel
 Dragon Systems, Inc. : speech recognition : +1 617 965-5200
 320 Nevada St. : Newton, Mass. 02160, USA
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: 3.478 Comparatives

Date: Thu, 11 Jun 92 14:13:46 ED3.478 Comparatives
From: Richard Sproat <rwsmbeya.research.att.com>
Subject: 3.478 Comparatives

Thanks to Sue Blackwell for this substantially better set of data:

 easy 2704 easier 1042 uneasy 320 uneasier 2
 happy 2455 happier 229 unhappy 513 unhappier 3
 likely 3328 likelier 1 unlikely 777 unlikelier 0
 lucky 846 luckier 24 unlucky 107 unluckier 2
 tidy 208 tidier 14 untidy 115 untidier 3
 worthy 225 worthier 3 unworthy 61 unworthier 1

Given the simple-minded method I suggested in the previous posting
the `expected' frequencies of the _un-X-er_ forms, assuming
E(_un-X-er_) = F(_un-X_)/F(_X_) * F(_X-er_)/F(_X_) * F(_X_)
	 = F(_un-X_)/F(_X_) * F(_X-er_)

	 Actual Expected
 uneasier 2	 123
 unhappier 3	 47
 unlikelier 0 0.2
 unluckier 2 3
 untidier 3	 7
 unworthier 1	 .8

_unlikelier_, _unluckier_ and _unworthier_ are about as frequent as
expected. _untidier_ is only half as frequent as expected.
Interestingly, _unhappier_ is indeed an order of magnitued less
frequent than expected. _uneasier_ is two orders of magnitude less
frequent, but note that _uneasy_ and _easy_ are not the same lexeme,
and given that _uneasier_ must be derived from _uneasy_ (following the
earlier discussion there is simply no other possible conclusion in
this case), it is pretty clear that the multiplicative model
*shouldn't* be the right model in this case: given that _easy_ and
_uneasy_ are simply different words, we wouldn't expect to find
_uneasier_ sensitive to the relative frequency of _uneasy_ and _easy_.
But why is _uneasier_ *so* infrequent. Is it possible that _uneasy_
has drifted lexically so far from _easy_ that people are beginning to
lose the fact that it is morphologically derived via _un-_ prefixation
from _easy_? In that case, maybe it is beginning to be treated as an
unanalyzable trisyllabic adjective and thus has a stronger resistance
to _-er_ suffixation. If that is right, then it suggests that we still
need the prosodic condition on _-er_ affixation (contra Beard).

That leaves us with _unhappier_ and to a lesser degree _untidier_ as
going against the multiplicative model. But note two things: we don't
have the frequencies of the _more un-X_ constructions here, so the
frequency of the operation of comparison of the negative adjectives is
unknown. Secondly, for both _unhappier_ and _untidier_ obvious close
synonyms (_sadder_, _messier) with monomorphemic bases spring
immediately to mind, but not so clearly for the other cases. (Note
that I am *not* saying that there aren't possible synonyms in these
other cases, just that they don't spring to my mind as clearly as in
these two cases. And note that this doesn't explain why _unhappy_
-- close synonym _sad_ -- should be as common as it is.)

Things become slightly _untidier_ then:

1) E(_un-X-er_) = F(_un-X_)/F(_X_) * F(_X-er_) is roughly the right
 model, BUT
2) the frequency of the F(_un-X-er_) forms may be weighted by the
 observable likelihood of forming _more_ comparatives of the _un-X
 form, AND
3) obvious synonyms may probabilistically `block' formation of the
 `_un-X-er_' forms.

Again, if this is right, then we might have an answer as to why the
_un-X-er_ forms appear to be rare. Crucially, I do not believe that
they are rare for formal *morphological* reasons; not that anyone was
necessarily suggesting that, but I just want to make that point clear.

Richard Sproat
Linguistics Research Department
AT&T Bell Laboratories			tel (908) 582-5296
600 Mountain Avenue, Room 2d-451	fax (908) 582-7308
Murray Hill, NJ 07974			rwsresearch.att.com
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue