LINGUIST List 3.472

Tue 09 Jun 1992

Disc: Comparatives

Editor for this issue: <>

Directory

1. David Stampe, 3.465 Adjectives, Compounds
2. Jacob Hoeksema, Re: 3.471 Adjectives and Morphology
3. Allan C. Wechsler, 3.471 Adjectives and Morphology
4. Carstairs-McCarthy, _unhappier_ etc.
5. mark, Re 3.465 Adjectives
6. Allan C. Wechsler, 3.471 Adjectives and Morphology

Regarding stems that can take the English morphological comparative,
Robert Beard <RBEARDflint.bucknell.edu> asks:
 What sort of natural class is "all monosyllabic stems
 and disyllabic stems ending on an open syllable with a light
 vowel" anyway?

It is the class of stems which, with unaccented -er is added to them,
are pronounceable as one beat ("measure", "accent", etc.) in English.
English resists more than three syllables per beat (a musical triplet).
A larger number may be suffered (classifica tory, labiali zation) if
there is no alternative, but not gladly.

David Stampe <stampeuhunix.uhcc.hawaii.edu>, <stampeuhunix.bitnet>
Dept. of Linguistics, Univ. of Hawaii/Manoa, Honolulu HI 96822

Message 2: Re: 3.471 Adjectives and Morphology

The current discussion on Linguist on the proper treatment of un-comparatives
raises a question which I think has not been addressed so far: Why is it that
comparatives such as 'unhappier', 'uneasier' and the like are so exceedingly
rare? While I do not have an answer to this question, I think it is relevant
for the debate at hand.

In an 8-million+ word corpus of English texts, I found more than 14,000
tokens of comparatives, not a single one of which was of the form
'un-A-er'.

For three items with high frequency in the corpus, 'easy', 'happy',
and 'likely', I counted the following numbers of occurrences
(for 'easy', I also counted the relevant numbers for its antonym
'difficult'):

1. Easy.

easy: 798	easier:	334 more easy: 1 less easy: 1
uneasy: 30	uneasier: 0 more uneasy: 2 less uneasy: 1
difficult: 798 -- more difficult: 60 less difficult: 1

2. Happy

happy: 994	happier: 37	more happy: 3	less happy: 1
unhappy: 122	unhappier: 0	more unhappy:0	less unhappy: 0

3. Likely

likely: 1082 likelier: 1 more likely: 143
unlikely: 256 unlikelier: 0	more unlikely: 2

It is clear from these data that there is a great resistance to
use comparatives of un-adjectives. The data for 'difficult' show
(and other data could be added here) that this resistance does not
generally concern the "negative" member of a pair of antonyms, but
more likely precisely the class of un-adjectives. A comparison of
happy/easy and likely shows that the results are comparable for
forms that favor a morphological comparative and those that favor
a syntactic comparative.

If this is so, what causes it? Here, a number of answers could be
given, depending on your general position in the Sproat/Stump
debate. Let me sketch a few, and also indicate what my position
in this debate is.

A. If un-A-er derives from A-er by un-prefixation (Sproat's suggestion)
then one could say that un- does not like to attach to comparatives
(although it CAN, given that 'unhappier' etc are not ungrammatical,
just rarely used). This would put the blame on 'un-'. However, the
numbers for 'more un-A' are also quite low, which would then call
for a separate explanation.

B. The old phonological story: -er only attaches to short adjectives;
un-adjectives are too long. This, however, fails to explain why
some forms are impossible (such as '*acceptabler') while the un-A-er
forms are grammatical but avoided. It also does not tell us why the
periphrastic or syntactic comparative is rare.

C. There is a resistance to forming comparatives of un-adjectives
(both of the form un-A-er and more-un-A). If this is correct, then
un-A-er forms must be derived from un-A, rather than A-er. For now,
I would consider this the most likely option, and thus reject
Sproat's interesting suggestion, although I don't know what causes
the resistance. Maybe someone can enlighten me here.

Some final points:

1. Litotes.

If un- is the outermost operator in un-A-er formations, one would
expect to find un-A-er forms in the well-known 'not-un-A' construction
(cf. a not unfriendly gesture, a not unimportant remark). However,
this does not seem to be the case: *a not unhappier fellow than him,
*a not unlikelier outcome than this one. (Where the result is better,
as in _The outcome was not unlikelier_, we clearly deny something,
rather than weakly assert something, as in a regular litotes case
like _The outcome was not unexpected_.) Again, this pleads against
treating un- here as the outer affix.

2. Actual forms.

As far as I can see, un-A-er is possible only if un-A is possible.
If un- is prefixed to a comparative, one might also expect to find
cases where un-A does not exist. On the other hand, if un-A-er
is derived from un-A, then this follows.

3. Gradability.

One might also use exclamative sentences to test for gradability.
Compare:

Boy, is she smart! and
*Boy, is she asian/unmarried/presbyterian !

or:

What a smart/*presbyterian guy !

At least according to this test, comparatives are not gradable:

*Boy is she smarter!
*What a smarter guy!

------------
Jack Hoeksema
University of Groningen

I confess that I don't understand many of the technical (and, I suspect,
ideological) issues in this debate about "unhappier". Still, I think
there is a misunderstanding developing about the example

(1) ? My car is redder than orange.

Robert Beard stars (1) unambiguously, while Eric Schiller says it is
completely acceptable to him, and cites as a comparable example

(2) [Your clothes will become] whiter than white.

which I imagine we all accept, if only due to repeated inoculation.

But I suspect that Beard and Schiller are attempting to get two
different readings of (1):

(1a) * My car is redder than [it is] orange.

(1b) My car is redder than [the color] orange [is].

In (1a), which I think is the reading Beard tried unsuccessfully to get,
"orange" is an adjective, while in (1b) it is a noun. My intuition is
that (1b) is parallel to (2), and when I try to read (2) as parallel to
(1a), I have to reject it.

I don't want to speak for Beard, but I think his point is that the "it
is" can be deleted from

(3a) My car is more red than it is orange

to get

(3b) My car is more red than orange

without changing meaning; but "it is" cannot be deleted from

(4) My car is redder than it is orange

to get (1) without changing meaning.

Can I add my pennyworth to the debate between Sproat and others?
Much of the puzzlement over bracketing paradoxes stems from the
assumption that, if a morphologically complex word has a meaning
which is not plainly idiosyncratic or opaque, then that meaning must
be derivable from its morphological composition fairly directly.
But this neglects the factor of lexical-semantic pressure (or, in
some cases, inflectional-morphological pressure) to supply a
word(-form) with a certain meaning, by hook or by crook, no matter
how it is formed. For example, if one tries to derive the meaning
of _electorate_ and _constituency_ just from their morphological
components, one misses the fact that there is lexical-semantic
pressure to provide a word which means 'geographical area relevant
for legislative electoral purposes'. Apart from any other meanings
they have, _electorate_ and _constituency_ realise this meaning in
New Zealand and Britain respectively. The fact that, from the point
of view of its morphological composition, _electorate_ especially
seems more or less appropriate for this meaning is mnemonically
useful, perhaps, but not essential - contrast the Canadian
equivalent term _riding_, which is quite opaque. Similarly, there
is pressure (whether we call it lexical or inflectional doesn't
matter) to provide morphological comparatives, if possible, for
_unhappy_ and _uneasy_. And, given the existence of _happier_ and
_easier_, _unhappier_ and _uneasier_ are obvious candidates to do
the job, irrespective of whether their internal bracketing 'fits'
this meaning exactly.

My account here is analogous to Spencer's (1988) account of _nuclear
physicist_ etc. We need a term for 'someone who does nuclear
physics', and in using this term we use a mnemonically appropriate
one without bothering about its bracketing.

I am not suggesting that all bracketing paradoxes or morphosemantic
mismatches involve lexical semantic pressures of this kind, but I
suspect many of them do. I am also not suggesting (a` la Anderson
and Beard) that word-internal bracketings don't exist. What I *am*
suggesting, though, is that, even if such bracketings do exist, they
can easily be overridden by lexical-semantic pressures in
determining a word's meaning; also that morphologists need to pay
more attention to lexical semantic relationships - and not just ones
which are mirrored fairly directly in word-shape.

Some relevant references:
Carstairs, A. 1988. Some implications of phonologically
conditioned suppletion. Yearbook of Morphology 1, 67-94. [On
'meaning-driven' versus 'expression-driven' morphology.]
Spencer, A. 1988. Bracketing paradoxes and the English lexicon.
Lg 64, 663-82.
Carstairs-McCarthy, A. 1992. Current morphology. London:
Routledge. [Pages 47-51, 92-7]

Andrew Carstairs-McCarthy

Two comments, one on Beard/Schiller and one on Beard/Sproat.

Eric Schiller disputes Robert Beard's starring of
 My car is redder than orange.
>From reading the surrounding text in each of their postings, it is
clear that they are taking two different readings of this string.
Beard's starred reading is a paraphrase of
 (*B) My car's color is closer to red than to orange.
while Schiller's acceptable reading is a paraphrase of
 (S) My car's color is closer to red than orange is.

To express the distinction graphically, consider a scale of hue.
(m is "the color of my car"). The unacceptable reading (*B) looks
like this:
 (*B') R m O
 (distance mR < distance mO)

The acceptable reading (S) looks like this:
 (S') R --------m--------- O
 (mR < OR; m can be anywhere in
 the dashed range)
The geometry also points out an alternative interpretation of (S):
 (S") m R O
 (mR < OR; R is between m and O)
In this interpretation, the car may be a purplish-red, which the
speaker asserts is more like (an "ideal") red than (an "ideal")
orange is.

I agree with both Beard and Schiller: I find reading (*B)
unacceptable and reading (S) OK. I'm happier with (S), though, if
there's emphatic stress on "redder" than if there isn't: "My car
is REDDER than orange." I think I could also accept (S) for
geometry (S"), the purplish-red car that is closer to red on one
side than orange is on the other side. For what it's worth, I
don't like reading (*B) for geometry (S") any more than I like it
for geometry (*B'). -- At this point I'm getting scanted out on
the variations, so I'd better drop the subject while I'm still a
native speaker.

 * * *

On Beard and Sproat re "cook" vs. "baker": My first reaction to
 (1) cook of French cuisine
in the context of deverbal nouns and their possible arguments, was
"ugh! unacceptable". Then came the double-take: "yes, that IS
OK"; and then the synthesis: "But it's not an argument of the
noun: it works the same way as
 (2) chef of French cuisine
does." IOW [in other words], the "of" in the acceptable readings
of (1) and (2) doesn't serve the same function as the "of" in
 (3) baker of French pastries
which governs an underlying object; rather, in (1) and (2) the
"of" is a looser connector, to which I don't think we can assign
a relation at any level lower than semantics. In short, with a
brief lemma I agree with Sproat on these judgements.

 Mark A. Mandel
 Dragon Systems, Inc. : speech recognition : +1 617 965-5200
 320 Nevada St. : Newton, Mass. 02160, USA

I confess that I don't understand many of the technical (and, I suspect,
ideological) issues in this debate about "unhappier". Still, I think
there is a misunderstanding developing about the example

(1) ? My car is redder than orange.

Robert Beard stars (1) unambiguously, while Eric Schiller says it is
completely acceptable to him, and cites as a comparable example

(2) [Your clothes will become] whiter than white.

which I imagine we all accept, if only due to repeated inoculation.

But I suspect that Beard and Schiller are attempting to get two
different readings of (1):

(1a) * My car is redder than [it is] orange.

(1b) My car is redder than [the color] orange [is].

In (1a), which I think is the reading Beard tried unsuccessfully to get,
"orange" is an adjective, while in (1b) it is a noun. My intuition is
that (1b) is parallel to (2), and when I try to read (2) as parallel to
(1a), I have to reject it.

I don't want to speak for Beard, but I think his point is that the "it
is" can be deleted from

(3a) My car is more red than it is orange

to get

(3b) My car is more red than orange

without changing meaning; but "it is" cannot be deleted from

(4) My car is redder than it is orange

to get (1) without changing meaning.