Professional Documents
Culture Documents
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Cambridge University Press is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Linguistics
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JL5 (I969) 193-320 Printed in Great Britain
* This work was done in conjunction with a study of the acquisition of some linguistic
and cognitive skills directed by Margaret Donaldson and supported by the Social Science
Research Council. We are grateful to her, John and Mary Marshall, James Thorne and in
particular the Editor, for their helpful comments and advice. Robin Campbell is now at
Psychology Dept, University of Stirling.
2I5
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
NP Predicate-phrase
I/
N Aux VP
Jolhn
Compar Adjective
morthn 4: S 4Fg
clever
NP Predicate-phase
N Aux VP
BilAdjective
clever
Figure I
216
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
certainly constitutes a strong argument for the proposal that SOME comparatives,
e.g. (7), should be derived from complex bases. But sentences (i) and (2) are
unambiguous and a similar analysis therefore requires independent motivation.
An obvious argument, used extensively by Smith (I96I), appeals to generality of
application. Smith argues that all comparatives can be derived from a single
transformational rule of Comparative Conjunction in which deletion of the
corresponding predicate in the constituent sentence is blocked, unless, as in the
case of simple constructions like (I), this predicate is non-distinct from the
predicate in the containing sentence. However, this is only justified in terms of
generality per se. We feel that, although it is desirable to formulate rules which are
as general as possible, this desideratum brings with it a real danger of glossing
over important distinctions. In (i) two OBJECTS are being compared whereas in
(6) it is two PROPERTIES, namely the depth and width of the river, which are
being compared. We could preserve this distinction if we derived (i) and (7)
differently.
Other possible motivating factors are dealt with below. For the present, we
note only that ambiguity and certain semantic considerations provide a basis for
distinguishing (i) from (7) and (6) respectively. Secondly, there is considerable
disagreement in previous work as to what should be the structure of the con-
stituent sentence. Smith (I96I) and Chomsky (I965) opt for Nom - be - Adj
where the Adj is identical to the Adj in the matrix sentence. Lees suggests
Nom - be - that - Adj and Huddleston (I967) Nom - be - than - Adj, with the
same restrictions on the Adj. Of course, if these restrictions are not met, then
'two-variable' comparisons (cf. Huddleston, I967) such as
2I7
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
the plural copula are must be used to delete the singular copula is in the embedded
sentence. Chomsky solves this and similar problems by reformulating the dele-
tion convention in the following way (I965: I82): 'nondistinctness rather than
strict identity is what is involved in deletion - only those features of a formative
that are inherent either to its lexical entry or to the position in the sentence where
it is inserted are to be considered in determining non-distinctness. Formally, we
can say that a formative must be regarded as a pair of sets of features, one mem-
ber consisting of the 'inherent' features of the lexical entry or the sentence
position, the other member consisting of non-inherent features introduced by
transformation. The general principle for erasure operations, then, is this: 'a term
X of the proper analysis can be used to erase a term Y ... just in case the inherent
part of the formative X is not distinct from the inherent part of the formative Y.
Though this seems entirely reasonable it does weaken an important convention
(since recoverability is of paramount importance) and may well have unforeseen
ramifications in other parts of the syntax of English. However, we wish to ques-
tion the use of optional deletion transformations on much more general grounds.
Consider the following pairs of sentences:
Since (io) is derived from the structure immediately underlying (Ii) by opera-
tions of deletion and permutation, it has always seemed unfortunate to us, as
psychologists, that, by any reasonable metric, (io) must be adjudged more
complex than (i i). However much linguists, such as Chomsky, may rail against
the confusion of competence with performance, it is now clear that any attempt
to describe or explain human language performance must make appeal to gram-
mar, and it seems reasonable to suppose that a complexity-metric can be defined
over derivations which will have some correlation, however wayward, with em-
pirical measures of complexity. At any rate, if it turns out to be impossible to
[i] Since we wrote this article, a paper by Doherty & Schwartz (I968) has come to our
attention, in which the authors do in fact adopt this latter suggestion.
2i8
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
[2] In fact, semantic analysis of sentences like (I2) and (I3) is faced with two possibilities
(cf. Bierwisch, I967; Katz, I967: either (i) postulating a relational marker, roughly
[greater in (size) than (an average. . .)] (cf. Katz, I967: I87), or (ii) supposing (I2) to
have been derived from, loosely, John is more clever than A, where A is some dummy
element carrying features which might be realized as the average N where N is some
superordinate of John. Adopting (i) greatly complicates a semantic theory of the type
proposed by Katz. On the other hand, adopting (ii) results in an infinite regress of base
structures, given the current syntactic framework. Thus (2) would be derived from
something like Yohn is more than # Bill is more than # A is more than # ... # . . .
clever # clever # . . . clever. This is obviously an implausible derivation!
[3] Possibly the double-based approach to comparatives was suggested as being compatible
with the simple logic of relations, such that fxy entails fx and, under some circumstances,
fy. By the same token, however, consideration might have been given to the parallelism
between adjectives and verbs (which we discuss below), e.g. Bill hates John, Bill is
greater than John. Such an approach is made explicit by Johnson (I922) in his alternative
title for the chapter on 'Relationals', namely 'Transitive Adjectives'.
zi9
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
This shows, if the current analysis of comparatives is the correct one, that co-
ordination must be ordered fairly late in the sequence of transformational rules:
specifically, it must follow the rules which (obligatorily) delete the matching pre-
dicate in the constituent sentence. This is probably too strong: rather, it shows
that the rule for co-ordination is not ordered with respect to other transforma-
tions but may be applied whenever its structure-index etc. is satisfied. However,
220
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
2. AN ALTERNATIVE FORMULATION
This is not as unsatisfactory as it might at first appear, since it sets (6) apart from
(i), (2) and (4) in terms of relative syntactic complexity (as we suggested above).
The objections raised by Smith to this proposal are not so serious when the
matrix comparative construction is simple, as we are suggesting. However, like
Huddleston (I967) we do not consider that this type of reduction of all compara-
tives to simple comparatives is desirable. For one thing, it leads to extremely
clumsy derivations in many cases involving recourse to elements corresponding
to, for instance, the number of nom, in structures like
22I
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
which do not appear in the surface form. We shall not attempt here to deal with
any possible objections to our suggested analysis which make reference to more
complex constructions. (Rather, in the following section we shall attempt to
show that our suggestions lead to an interesting set of generalizations about
SIMPLE sentences.) However, there is one observation of Huddleston's (which he
attributes to Pilch, I965, and which was also noted by Chomsky, I965: 234,
fn. 36) which we must deal with because it seems to constitute syntactic evidence
Pred phrase
Nom VP
Predicate
Figure 2
[4] The argument of Lees (I96I, p. 172) that the relation between the deviance of
222
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
and
as against
and
However, there are severe technical difficulties with this. The derivation of (23)
from (28) (or rather from whatever underlies (28) ), and from
would be excessively intricate, and hence the simpler or more 'normal' of the
two senses of (24) would have the more complicated analysis.
As an illustration of what we mean by talking of 'severe technical difficulties'
consider how Smith, incorporating Huddleston's suggestion, would derive (23).
First of all, note that the derivation of (23) in this way involves a high degree of
embedding and requires the following four sentential strings:
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Now what is difficult here from a technical point of view is the framing of th
conditions, in a non ad hoc way, under which these transformations operate. In
particular, the conditions for the last two operations pose problems. Secondly, if
the optional deletion deriving (32k) from (32j) is not applied, then we are left
with an unacceptable sentence. Moreover, we cannot delete the final nominal
earlier, since its presence is required to control the order-change at (321). We have
no idea as to whether or not these difficulties could be surmounted.5 In this
respect it is hardly surprising that Thorne (I968: 297) should remark: 'The
reason for this - the partial failure [of the computer-program for syntactic
analysis] to handle and and the total failure to handle as and than - is very simple.
It is directly related to the fact that we ourselves have an incomplete under-
standing of and and only vague ideas about the grammar of as and than.'
However, we have still to show how the observation of Huddleston, Pilch and
Chomsky can be accounted for by our simpler scheme. The solution we will
suggest is a tentative one; it raises difficulties of a different sort. First of all, it ha
been suggested that the subject of the embedded relative need not be FORMALLY
identical to the nominal which it modifies. In fact, we believe that what is
crucial to relativization is not FORMAL identity but identity of REFERENCE. Mor
over, this is true not only of relativization, but also of other transformational
processes, such as reflexivization and pronominalization. We do not intend to go
into this question in any detail (for some discussion see Postal, I967), since we
see no good solution at present. However, if this is a correct view, then the
necessity for FORMAL IDENTITY disappears (for some other arguments for th
proposal see Chafe, I967). Once this restriction is removed it is clear that the
shifting to prenominal position of comparatives and other adjectival expressions
could be made conditional on the feature structure of the subject of the embedded
relative. Loosely, where Bill is not a lawyer we would have
[5] In Doherty & Schwartz (I967) an attempt is made to overcome just these technical
problems. Whether this attempt is successful or not is open to doubt, since they 'solve'
the problems mentioned here by analysing the embedded complement of comparison,
corresponding to (a) A Mini is a wh- big car, as derived from, roughly, (b) A car # wh-
the Mini is a car # Mini is wh- big, without giving any justification for this peculiar
structure. Their article provides some first-rate examples of what we mean by 'clumsy
derivations' and 'severe technical difficulties'.
224
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
whereas
is unambiguous. However, we do not feel that this is the whole story by any
means. For although (35) might be regarded as ambiguous, the same might be
said of
i.e. 'big for an elephant' or 'big for an animal'.6 We have thus a very difficult
problem. Whereas the ambiguity of (37), (38) and even (35) might be dismissed
as syntactically irrelevant, the ambiguity of
(39) I know several lawyers who are more successful than Bill,
says something about my wife as compared with other wives in town, whereas
It is worth noticing that this example, and of course its many analogues, poses
problems for relativization since it too (cf. fn. 2) leads to an infinite regress of
base structures. Let us now summarize these various facts. Consider the follow-
ing sentences:
[6] This example has been extensively discussed by Weinreich (I966), Katz (I967),
Lyons (I968).
C 225
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
(46) That painting is the # wh the painting is the most expensive in my house
# one,
(48) Nom -_ *
[7] Some support for these suggestions is provided by the existence of such forms as
(a) He is too big a man to be overlooked, (b) j7ohn's mother is as bigoted a woman as I know,
(c) How nice a man he is! where, not only the head of the complement noun phrase, but
the determiner also, follows the predicate head.
226
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
+ F2 ] Predicate +F 2
2
? F2
+F_ ~ ~ ~ ~~~~~~F
+ F ~ ~ ~ ~ ~ ~ ~ +
Fq
L Nominal
I
underlying structures associated with (43c) and (43d) differ from that associated
with (43b) in the position of the transformational 'trigger'. Likewise for (45c)
and (45d). The rule will, of course, be optional in the case of (43a) and (44a)
and obligatory in the case of (45a), the obligatorily-introduced one in (45b)
being deleted to produce (45a). The awkward facts of relativization are then
easily accounted for; those Adj + Noun combinations where the Noun is not
introduced transformationally will have the Adj introduced in the normal
manner from a relative clause, the permutation transformation achieving this
being made conditional on identity of feature composition above the asterisk
(in addition to identity of reference, however that is to be marked).
Of course, there are still difficulties in our formulation. Clearly, it is only
relevant to a certain class of adjectival predicates, for instance we do not have
nor
227
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
It is possible that the italicized elements are related in the way we suggest.
Clearly, we need to consider the relationships between proper nouns, pronouns,
full nouns, articles, etc. in much greater detail.8 Since these issues are not
relevant here, we will not pursue them.
It is worthy of note that certain languages (e.g. Burmese, Chinese, Viet-
namese, and other languages with 'classifiers') seem to require a rule of just
this sort. In these languages one can identify classes of noun complements
which, in general, do not occur in isolation (cf. Hla Pe, I965). An interesting
question, to which we have been unable to find an answer, is whether or not
in such languages there correspond to English
a range of structures, differing only in the choice of the classifier, which mark
the ambiguities of the English form. Hla Pe (I965: 170) says: 'A classifier
indicates an attribute of the object which assigns it to a class. Each classifier can
be applied to several objects of similar attribute, and some of these objects may
have more than one attribute. A person may be classed as a human being or an
animal according to his behaviour; and a horse or an elephant as merely an
animal or a mount and a sword as just a straight long thing, or as a weapon.
The choice of classifier is prompted by the occasion.' If this is so, then this
adds some credibility to our suggestions.
[8] Proper names, pronouns and definite articles seem to be related syntactically in a
poorly understood way. For instance, in general they do not co-occur with one another;
that is, we do not find *the John, *the he or *he John. They are also related semantically
in that they all are used to refer to SPECIFIC individuals. Postal (1967) has suggested that
pronouns be considered a special sort of definite article. Though this is perhaps too
strong a claim, it is certain that they are all to be associated with definiteness. This
suggests that another structure marker may be required (introduced after the rule
assigning definiteness to the complex symbol dominated by Nom). Together with our
suggestions in the text, this makes Bierwisch's (i 967) notion of marker-trees seem useful,
not only in semantic analysis, but also in syntax.
228
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
These and many other similar sentences suggest that the correspondence
between PASSIVES and comparatives is far more complete than that between
active transitives and comparatives. This calls in question the parallelism
suggested by Lyons and suggests instead that simple adjectival predicates are
to be related to short passives and comparative predicates to full passives.
Indeed for the former two types of structure, intuition often fails to distinguish
one from the other. For example, which of the following sentences are 'verbal'
(i.e. short passives) and which 'adjectival'?
However, matters are even more complex than this. Logicians have often
marked a distinction between 'contrary' and 'contradictory' pairs of adjectives,
in terms of their different entailments. In these terms, pairs of adjectives, such
as big-small, beautiful-ugly, etc. (in fact, all those adjectives which allow com-
229
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
230
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
This leads to considerable confusion between agency and ergativity, since the
subjects of active transitives are normally, though by no means always, agentive.
If this is an acceptable view then nothing remains to be explained except the
fact that certain languages mark the subjects and objects of transitive verbs
differently. But this is exactly what we would expect if these case-markings are
superficial features of the nominals in question. It is notable that in Turkish
and Russian the subjects and objects are differently marked just in the case
that there IS possibility of confusion, namely when the OBJECT is definite (Turk-
ish) or animate (Russian), both of which are typical features of SUBJECT nominals
(cf. Lyons, I968: 294). It seems, therefore, preferable to consider ergativity in
English to be a property of certain verbs, e.g. open, change, move, etc., which
allow deletion, or rather non-realization of the copula in short passive construc-
tions. We suggest that the so-called intransitives which figure in discussions of
ergativity in English are in fact pseudo-intransitive, those with agentive nominals
being reflexive, those with non-agentive nominals being reduced short passives.
231
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
(C) P N P N
a2 U2
232
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
(73) _ p N N
(ii) N N
case 0
233
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
lnominative, + animate
locative) - agent - agent
- dummy - dummy nominative .
nominative locative
nnominative, + directional { locative
_ dummy - spatiotemporal n
- to I
S S
Predicate Predicate
N I{ locati -N N P locative N
+ directional .1
- spatiotemporal
Figure 3
cf. John was killed by Bill; The fish was cooked by Mary; Mary moved (i.e.
Mary was moved by a); Mary was annoyed (i.e., Mary was annoyed by a)
(77) [agentive [P locative]]
cf. J7ohn went to school; John left (i.e. John left from a)
(78) [locative [P agentive]]
cf. (if acceptable): The room was entered by J7ohn; the school was reached (i.e.
the school was reached by A).'0
[Io] Rule (E.i) may have to be modified to account for the odd behaviour of locative sub-
jects. In fact, it seems as if what we require in such cases is a special rule which rewrites
234
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
Consideration of these and other examples shows that we now have succeeded
in relating and distinguishing the different types of adjectives and adjectival
predicates in the ways suggested earlier in this section. Also we have succeeded
in relating adjective and verb in the manner suggested. Obviously, there is
enough material in this grammar to form the background of a very wide-ranging
discussion and we have not even begun to consider the consequences of deve-
loping some of the N's as E.
There is one distinction which we have still to make, namely that which
distinguishes 'equatives' from 'comparatives'. The complements of both of these
comparatives are derived from a locative nominal. We want to mark these
locative nominals in such a way that in one case, where the expression is 'equa-
tive', the locative constituent (see rule E) is developed as as-as and in the other
case as more-than.
Considering other locatives for the moment (notice that we have defined
the feature [locative] into the Predicator, replacing it in the Nominal by [nominative].
Thus the Predicator CS underlying have, own, etc., differs from the CS underlying be,
belong, etc., in the presence of the feature locative. Likewise, the presence of [locative]
in the CS's of verbs like go, look, come, etc. is realized as go + Locative Particle, etc. If
we allowed this rule to apply optionally under certain circumstances, this might enable
us to account for the wide inter- and intra-idiolectical variation on the constituency of
such particles: cf. (a) Who did you give the book to, (b) To whom did you give the book.
235
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Further, although we have not dealt with such constructions here, we will
require a second distinction to mark off non-spatiotemporal locatives from the
above types. Consider the following sentences:
The locative in (gi) is clearly spatiotemporal whereas the locatives in (88), (89)
and (go) are clearly not (other cases are not so clear). Let us now mark this
distinction by the rule
It should be noted that these rules are unsatisfactory inasmuch as they intro-
duce much redundancy, that is, the selection of these features is controlled by
the environment of the locative nominal to some extent and by the animacy of
the locative nominal.
In terms of these distinctions, the locative complement of comparison will
be marked [+directional, -spatiotemporal, ?to] depending on whether the
locative constituent is realized as as-as, corresponding to [+'to] and more-than,
corresponding to [ - to]. It should be pointed out that little of this set of distinc-
tions is new. In fact, much of it is quite traditional and has been recently
discussed in terms similar to ours by Lyons (I968: 298 ff.).
A distinction was drawn by Donaldson & Wales (I968) between 'functional'
and 'descriptive' comparatives according to whether or not the 'norm' of
236
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
It would seem that modality is involved somehow, since (92) and (93) are related
in an obvious way to (95) and (96):
At any rate, there seems no difficulty in principle with this suggestion and
we will suppose it to be a good one. Corresponding to [+ to] we will have
[locative] realized as 'enough' and corresponding to [-to] as too.
Many of the above suggestions, rules, remarks, etc. argue for a strong and
complete (in the formal sense) parallelism between verb and adjective in deep
structures. Recently, there has been much discussion of this proposal, notably
by Lakoff (I965), Lyons (i966a, I968) and Anderson (I968). To the various
remarks of these authors we have little to add (over and above our proposals
about comparatives), save that we believe the general thesis to be correct. How-
ever, it seems to us that a profitable line of further investigation of this topic
might proceed by way of methods similar to those we have employed here. For
instance, although we have throughout attempted to set up correspondences
between different sorts of adjectives and different sorts of verbs, we have not
carried these distinctions to their limit. A recent paper by Garcia (I967) provides
a useful discussion of verbs in terms of (a) their lexical content and (b) the
freedom with which they combine with different sorts of complements. At one
end of the scale we have the Modals, themselves by no means homogeneous in
terms of their behaviour with respect to (a) and (b), which in these terms take
an obligatory sentence-like complement. Between the modals and 'full' lexical
verbs we have an enormous variety of equivocal verbs like begin, try, succeed,
make, do, happen, etc., 'equivocal' as to their lexical or grammatical nature.
Likewise, English adjectives exhibit the same sort of continuous progression
from 'modal' adjectives like free, able, bound, obliged, etc. through adjectives
237
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Clearly these verbal forms behave much as the comparative morphemes (more ...
than and as ... as) do in English.
The second mechanism involves case-marking of the complement of com-
parison. The case used is commonly the ablative or dative. Frequently this
mechanism co-exists in the language with one or other of the other mechanisms.
(In some languages the 'adjective' is inflected, in others not.) Tswana (cf. Cole,
1955) exhibits such a variety of methods, as the following examples show:
238
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
239
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Although these sentences are vastly different from a formal point of view, they
are synonymous, and (io8) is semantically redundant in that both go and
fast involve the notion of motion. It might not be too far-fetched to imagine
rules, operating after the subject-forming rule, which introduce redundancies
of this sort. The rule suggested in section 2 for forming noun-complements is
an example of such a rule. This would then allow us to maintain a simpler
statement of the meta-rule connecting form and meaning, namely, all and only
those sentences deriving from identical base-structures (that is, before rule D
operates) have identical readings. ('Interpretive' semantics of the kind proposed
by Katz (I967; etc.) has to allow for the possibility that distinct base structures,
such as those normally considered to underlie (107) and (io8) might have the
same readings.) This general view of the nature of the connexion between
semantics and syntax will be implicit in what now follows.
Comparative sentences present a wide array of semantic problems, some of
which we have already discussed. One feature of these sentences which has not,
however, been discussed previously is the different 'sensitivity' of different
comparative structures to contextual factors. For example, simple 'comparative'
structures as exemplified by
240
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
seem to be relatively free from contextual variation in the sense that, loosely,
their meaning is the same, no matter what the context. Also, their application
is relatively unrestricted from a contextual point of view, given of course that
John and Bill are 'comparables' in this respect and that John, say, subtends a
greater angle at the eye. (There are, of course, other phenomenonological factors
which might complicate this a little.) However, sentences such as
seem to present problems. Idiolects vary on this point. Some speakers maintain
firmly that (iI Io) must be glossed by
Others maintain no less firmly that John may, in fact, be taller. It may well be
that both factions are wrong and that the correct view lies somewhere between.
For consider,
Clearly, the 'exactly equal' school of thought has problems here, as they do in
another way with
whereas, the 'greater than or equal' faction have problems in so far as the more
'normal' interpretation of (iio) must be that glossed by (iii), even if that
glossed by
is possible. Clearly, the existence of such sentences as (I14) and similar sen-
tences with, just, exactly, easily, etc., reflects this plasticity of meaning (notice
that we prefer to avoid the term 'ambiguity', since it is difficult to see how such
an ambiguity might be STRUCTURALLY marked). The same sort of problems arise
with negation, e.g.
(cf. Wales & Grieve, forthcoming, for some discussion of this and similar
questions).
Further, other comparative forms present problems in terms of application.
Consider
Sentence (ii6) seems to entail that both John and Bill are 'short'. Put another
D 241
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
That is, (i I9) is semantically marked, in this sense. We shall now put forward
some suggestions as to how these notions might be formalized. Before doing
so, however, it will be necessary to discuss briefly the changing role of trans-
formational rules.
Since the early days of generative syntax there has been a gradual but steady
change in the role of transformational roles. Whereas formerly generative power
resided almost solely in these rules, it now resides in the base rules. Moreover,
in Chomsky's early grammars (e.g. 1957), certain transformations (e.g. question,
negation and passive) formalized semantically important relations between
kernel and non-kernel sentences. This fact provided scholars (e.g. Lyons, I963)
with a technical apparatus suitable for investigating these traditionally important
semantic relations. On the face of it, this 'bonus' of early generative grammar
has now been lost (cf. Katz & Postal, I964: I I8-i I9). However, we should like
to argue here that there is no reason why one should not still investigate these
same semantic relations (and others) using the same sort of transformational
framework.
An apparent objection to this might run as follows. A generative grammar of
a language formally describes the native speaker's knowledge of his language.
If a particular rule does not figure in that grammar, then it does not characterize
any aspect of the native speaker's knowledge and is therefore irrelevant to the
goals of synchronic linguistic theory (as defined by Chomsky). To this objection
we reply that this is a very narrow view to take of linguistic theory, and that,
even if one accepts this view of the purpose of synchronic description, our
242
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
suggestion is still reasonable on two counts: (i) these semantic relations are
crucially involved in the structure of discourse, and attempts should be made
to formalize this aspect of synchronic description; (ii) even in the more limited
context of the description of individual sentences, current semantic theory is
in no position to lay down hard and fast meta-rules about the way semantic
phenomena are to be studied.
We recognize of course that our method of semantic analysis is of a tentative
and preliminary nature, and that the results of applying this method must
eventually be incorporated into base structures before it attains much theoretical
significance. We will first illustrate the method with reference to a (relatively)
simple set of sentences. Passive sentences are sometimes distinguished from
active sentences in deep structure by the presence of a dummy element domin-
ated by a Manner Adverbial (cf. Katz & Postal, I964). This dummy operates
as a 'trigger' which allows the passive transformation to operate, permuting the
nominals and adjoining an extra constituent to the auxiliary verb.
This operation may be broken down into two operations, as follows:
The operations A and B allow us to relate (I24) and (125) to (123), or (I26)
and (I23) to (I25), etc. If we add to A and B a further two operations I and C,
I: 'identity' S >-- S
C: 'passive' apply A, B,
then we can relate (I23), (124), (I25) and (I26) to each of (I23), (124), (125) or
(126). We can represent this completeness as in Table i.
The four operations, or rather relations, thus constitute a GROUP, known as
the four-group (cf. Birkhoff & MacLane, I965: I3I). This much might be
regarded as a trivial consequence of our definitions. But so far we have only
been concerned with relations of form. Consider now the SEMANTIC effects of
these four operations. Whereas A and B effect a change in meaning, I and C
preserve meaning. Moreover, from the point of view of their semantic effects,
243
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Table i
I A B C
I I A B C
A A I C B
B B C I A
C C B A I
{I, C} -+ e; {A, B} -+ f
Although we have, for simplicity, discussed these formal and semantic relations
in terms of operations, we stress again that it is the RELATIONs between these
sentences that we are attempting to characterize and not, for the moment, their
derivation in a grammar. That is the 'operations' are UNDIRECTED.
However, we would now like to refer to our earlier remarks about the passive
transformation and argue that, for a certain class of transitive sentences, each
pair of members of that class contain in their deep structures sequences of
'choices' related as e or as f. That is, in terms of our base grammar proposals
in section 3 (and assuming that lexical selection occurs at a stage before the
244
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
assignment of the case features, [agentive], etc.), sentences (123) to (I26) will
all have the same base structure: the predicator kill combined with the two
nouns John and Bill. When this structure underlies (123), the John is assigned
the feature [agentive] and it is made subject; when it underlies (I24), Bill is
assigned the feature [agentive] and is made subject; and so on. We expect that
when this sort of structure is better understood, then we will be able to set up
relationships between feature-assignments corresponding to the semantic rela-
tions we have discussed.
Let us now briefly consider a more complex example. Consider the following
set of sentences:
(We will not dwell upon the obvious parallelism between the operations of
'alteration of voice', discussed above, and 'polarity reversal': cf. Lyons, I968:
245
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
463-468.) These five operations relate such pairs as (I27) and (I28), (127) and
(I29), (127) and (I30), (127) and (I3i). To continue:
(127) & (132) are related by E AB BA
(I 27) & (I33) are related by F AC CA
(127) & (I34) are related by G AD DA
(I27) & (I35) are related by H BC CB
(I27) & (136) are related by J BD DB
(I27) & (I37) are related by K CD DC
(I27) & (138) are related by L ABD ED AJ, etc.
(I27) & (39) are related by M ACD FD AK -GC, etc.
(127) & (140) are related by N BCD HD, etc.
(I27) & (I4I) are related by 0 ABC EC, etc.
(127) & (142) are related by P ABCD EK - OD, etc.
Thus we have the matrix represented in Table 2.
We would now like to relate the sentences SEMANTICALLY according to the
Roman numerals on the right of the above list. (All the sentences tagged 'I' are
synonymous, as are all those tagged 'IIF, and so on.) Of our five 'basic' formal
relations, four are semantically significant, that is, effect DISTINCT changes in
meaning. As with the simple example with which we began, we can set up
correspondences between formal and semantic relations as follows:
246
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
Table 2
I A B C D E F G H J K L M N O P
I I A B C D E F G H J K L M N O P
A A I E F G B C D O L M J K P H N
B B E I H J A O L C D N G P K F M
C C F H I K O A M B N D P G J E L
D D G J K I L M A N B C E F H P O
E E B A O L I H J F G P D N M C K
FFCOAM
F F C O A M HII K
KEPGNDLB
E P G N D L B J
G G D L M A J K I P E F B C O N H
H H O C B N F E P I K J M L D A G
J J L D N B G P E K I H A O C M F
K K M N D C P G F J H I O A B L E
L L J G P E D N B M A 0 I H F K C
M M K P G F N D C L O A H I E J B
N N P K J H M L O D C B F E I G A
OO H F E P C B N A M L K J GI D
| P N M L OK J H G F E C B AD I
247
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
Table 3
E N C R
I K L O A H J M B F G N C D E P
I I K L O A H J M B F G N C D E P
K K I O L M J H A N G F B D C P E
L L O I K J M A H G N B F P E D C
0 O L K I H A M J F B N G E P C D
A A M J H I O L K E C D P F G B N
H H J M A O I K L C E P D B N F G
N -?
J J H A M L K I O D P E C N B G F
M M A H J K L O I P D C E G F N B
B B N G F E C D P I O L K H J A M
F F G N B C E P D O I K L A M H J
G G F B N D P E C L K I O M A J H
N N B F G P D C E K L O I J H M A
C C D P E F B N G H A M J I K O L
D D C E P G N B F J M A H K I L 0
E E P D C B F G N A H J M O L I K
LP P E C D N G F B M J H A L 0 K I
strate this formally, that we could, in principle, set up relations between feat
assignments in simple comparatives corresponding to the semantic relations we
248
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
Table 4
E N C R
E E N C R
N N E R C
C C R E N
R R C N E
have discussed. Moreover, if, as suggested earlier in this section, there are
grounds for considering one form of the simple comparative construction
semantically primitive in some sense, then other forms could be related to this
basic form by way of these relations between feature-assignments. For instance,
(I27) John is bigger than Bill and (I38) 7ohn is not as small as Bill are related
semantically by e, and formally by L = ABD. In terms of the features discussed
in Section 3, and others not discussed there, the features assigned to (I27)
might include locative-*[-to], P--[+polarity] and Mood-+[-negative], and
those assigned to [138], locative-+[+to], P-+ [+polarity] and Mood-*[+nega-
tive]. In that case we could set up a DIRECTED relationship, [-to, +polarity,
- negative] _ e[+ to, - polarity, + negative], making the selection of the right-
hand side triad of features conditional (though OPTIONAL) on certain contextual
information about Bill, namely, that he is 'small' (though of course we do not
know how to represent this information formally).
Finally, a strong argument for our classification of sentences like
as comparatives is that they, too, exhibit the same set of semantic and formal
relationships as do comparative sentences with -er/more. . . than and as. . . as.
Clearly, much of what we have suggested is speculative. In particular, our
grammatical suggestions require extensive justification which it is not possible
to provide here. It may therefore turn out that some or all of our proposals are
ill-founded. Our main purpose has been to re-open discussion of comparative
structures with a view to revising the older type of analysis, which we believe
to have been unsatisfactory in that it assigned superficial status to the linguistic
expression of what we believe to be a fundamental linguistic, logical and intellec-
tual operation.
249
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
JOURNAL OF LINGUISTICS
REFERENCES
250
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms
COMPARATIVE STRUCTURES IN ENGLISH
25I
This content downloaded from 136.159.235.223 on Sat, 17 Sep 2016 23:05:56 UTC
All use subject to http://about.jstor.org/terms