?9?8. Re\ iw
artic‘k of Hullcon
t97Ya. Transfczrm;~tional
t 1976). Language
53,
antI non-trawsformational
348-376. accounts
of focus in
man (cd.), .Ighcm grammatical pew in Linguistics
structure,
no. 7. Los Angeles:
ositions and Run-trrlth-condition,lll semantics. London: Academic
dc, ~“.l~repres~Rtat ion of ~~i~~~~~~t~ Brass, Minneapolis, Minn.,
grammatkal structure. 1978. I68 pp. $15.00.
Pridc,tuu, Dept. of Linguistics, crta, Canada. TM; 2 H I . iitry
D.
University
CJniversity
of
of Alberta,
~~~~~t~~act t\ IQ within linguistic: t hcory immediately after 1965 produced ots within the transformationai camp. iLiore recentiy however there ~~r~se~serious chalk ts the transformational paradigm itself, and the book t review here, Kac’s c;gr~~seratutio~r oj’ Grummatical Structure, falls into this f ‘corepresentational grammar’, or COKG as Kac calls it, is htful alternative to transformational grammar (TG), although roblems. The book consists of six chapters, exemplifying COWG‘s basic principles t hrou h an analysis of English. The first chapter details the basic aswmptions and prin ipb of CORG, comparing it in passing with TG. re%ents a set of principles and rules for the analysis of English simplex Me chapters 3 through 5 all deal with properties of complex sentences, nominal modification, and coordination. The book short, pithy chapter devoted to CORG’s metatheoretical foundations. Kltther than dealing in &tail with Kac‘s analysis of English syntax, I have chosen to the nature of CORG itself in an attempt to explicate the theory as I understand it, although 1 shall comment on some aspect% of the particular analysis. I shall clsc,~ with a discussion of a fundamentally miw lderstood issue within linguistic theorizing, namely that of ;he autonomy of ling4stics and its independence from psychology. CORC is a nontransformational theory of grammar which admits no abstract (‘deep structure’) syntactic level. Nevertheless, the theory associates two distinct levels of representation with each sentence: a categorial (or constituent structure) representation and a relational representation, The categorial representation corresponds roughly to the more familiar surface syntactic level, although Kac’s syntactic representations, which dispense with the categories of S and VP, are considerably less elaborate than those found in TG. The relational representation consis!s of an unordered conjunction of statements each of which specifies a relation among the is of the “grammatically significant elements” (p. 9). Each functional statement form X = p( Y) where X and Y are elements of the sentence and p is a relation, such
( 1) Harry
likes vodka.
(2) Vodka,
Harry
(3) Vodka
is what
lihes. Harry
(p. 17)
(4) Vodka
is liked
by Harry.
likes.
Kac ~~~~ts %W(p. 17). CQRG and TG dialer not so much as competing hypotheses of gr,lmmatical systems, b11t mo e as distinct metalanguages within s can be fr;rmed. In order to c;t*rv out his goal of pairing surface ~~tjo~~f representations, Kac proposes three distinct types of rules : rules of c~~represent~~tion, metacondit ions, and domain restrictions. ion (or C-rules) are “. . . language specific conditions inthe relational structures of categorially analyzed sentences” uage specific statements which pair functional with some syntactic (or, presumably, morphological) configuration. Ore e:o/‘~b’~~PI!~~ (OR) states that a NP to the right of a nonpassive predicate le as c)BJ{ P) (pp. 3 I, h-r;Z).The object rule properly identifies t:o&a in (1) c:t cd Eh, but it does not pro\ ide a candidate for object in (I!), since there ht of the kerb. In order to ‘treat such problems, Kac invokes a stinction among vat i~us arguments of a predicate. In parpp, 45tT) that actiw objects and passive subjects be treated as hilt acti\c subjects and passive objects be treated as unmarked the ob~iqtrc~ distcmce corrdiriorl (ODC) which states unmarked argument precede a predicate, they must be --unmarked (pp. 46, 62). Such an analysis correctly treats ct of Mu, since the verb is nonpassive, and the object is therefore marked, while the subject is unmarked. Furthermore, the ODC correctly labels WE&~as object of lib in (3). Kac’s view of subjects as unmarked arguments for actives follows from his treatment of the subject relation as deductively defined. Rather than positing a scrfrjvct rcrk (p. 31) stating that an NP to the left of a nonpassive predicate is analyzable as SUBJ(P), he proposes that subject assignment be carried out by a process of elimination whereby in nonpassive, nonimperative sentences containing either (a) only one argument or (b) several arguments, all but one of which have been assigned a relation, the remaining (or only) argument is deductively assigned the role of subject. Even though rules as OR and ODC are language specific. one still wonders just what kind of analogous rules might be needed for those languages which place less emphasis on linear ordering and more on morphological indicators for grammatical relations. Kac has a great deal to say about subjects ;md objects but he mentions indirect objects only once (p. 73) and he does not provide an analysis of dative sentences at all. Presumably he would treat pairs of sentences such as postulutos
that if both a marked in the or&r ~74murk
(Sa) Harry (Sb) Harry
d an
gave ! arne flowers to Sue. gave !;w some flowers.
sharing a relaticn: ’ representation in which SW is the indirect object and some jtow~ursthe direct cbj ct. There is, furthermore, no discussion of illocutionary distinctions signalled ;)y ,yntactic differences such as, for example, declaratives versus interrogatives; moreover, contextually mptivated notions of topicalization and focus are generally ignored. The major attention in CORG is devoted to grammatical relations, with somb: attention to coreference and coordination, but virtually nothing as
R;v-;c
174
is said
about
are even
omissions that
sentential
surface
Thus,
(6) What it is quite strange
more
syntactic
distinctions.
striking
information
that
relate
of liLp,
wit
just
is, howcker,
that
at the early
for OBJ(likc)
assignments
are carried
;in
must ncccbsarily
it is ~*~~~~~~i
to OR, but clearly
to handle these and other pr domain restrictions to ensure tlh;nt correct
out.
Metaconditions
arc proposed
principles which govern the structure of predications. The three which form the basis of CORG arc (pp. 34tIj the baw 191’ C’ot=rt~s nonexpletive, some
predicate,
element u&s:,
nonvocative
NP in a st’ntencc
the object
in a sentence donjoined
of a preposition
must
btxr
or coreferential,
Assorintior~ (every predicate the exhaustive and unique predications) unique
with
pairing
is not
must
kc idcniif’led
or both).
may
n~~ta~a~~~
;I\
t Iw I_ir\V 01 ICliitirbll;
01~2 argunicnt
as ~~~~~er~a;
bear the same
iind
relation),
must have a subject). The three mt’tit pairing of constituent structure argument\
grammatical possible
more. the metaconditions actives.
anly
be
gin.
according
which should play that role. In order proposes a set of metaconditions and
pric,ilegc. stages of it%
can be an&r
such an assignment
and Sue adores
plus at
(Scd‘ Pri~~~~~~~~
an author’s
for everything
as CORCi
,c./r-word
to note that just because a constituent
Sue is a candidate
d cent
is a question.
l
Selectivity
Pikes vodka
functional
that paired
ths initial
. !ntt’nce
the
it does not follow in the sentence
(7) Harry
iarc
\r+lrtrf as the object
to account
It is important relation, example,
it is rcclillcd signal
like
rules which
such an approach.)
go~fcrn
contcutually
like?
to treat
10 find
expect a theory
when
for a sentcncc
natural
or
configllrations
does Harry
not
sentential
information
,I’.\
relations, for a string,
permit
In a sentence like (I),
and presumably
rod/in
t ‘N Ps or
if such an e~h~~j~t~~~ and
then it is judged
as itl~formed.
Furt
the exclusion
of a subject
assignment
is analyzable
as 0BJ(lik1~),
and since there is
r
other candidate for the role, roriku is assigned the object relation, leaving unassigned. According to Correspondence, Hurty must be given s1911-w gram role, and by Uniqueness, Since
Association
role Htrrry assumes,
whatever
requires
that
every
sentence
deductively
given
that role, and consequently
for actives.
From
one
it is a relation somewhat
which
dift’erent
perspective, is assigned
point
of view,
SUBJ subject
be thrrt of obJect.
be assigned
a subject,
1llrrv.r
no rule of subject
assignment
is needed
appears
to arguments
it cannot
to be primitive
in constituent
assignment
in the theory structure,
or whether
the analysis.
The
notion
a
more profound
of subject
it
can be seen to take place bq Y
generalization
is of extreme
is
since
but from
process of elimination, rendering it the status of an ‘elsewhere‘ relation. clear whether Kac’s treatment of subject assignment for actives is simply a simplification
no
~~~~~.~
It is not
fortuitous is claimed to be lurking in
importance
in many
relational
it does at the top of ;i variety of hierarchies, while Kac seems ss eu,r!ted position. .tin re~tr~~~~~~~~ constitute the third type of CORG rule. These principles, cm-w to identify
the structural
units
is constituted. Kac’s notion of rtancc here and throughout the exposition of his m-e, the predication, which contains a ems, but which is dtstinct from the syntactic coned forms of complex sentences often cntv ~~t~i~t~~~WC!!away from the remainder of the elements with the sentence
~~~~~~~~~~t mucus aspects of the complement clause by placing all the cl;tuse under rl single node S in the deep structure and then fronting ;. hl~wt’\c’r, is noa permitted this luxury,
so Kac needs
unir w/wt . . . Sunt ute.
He labels this a E%,with the predicate clcrtand represents the predication as Pr/eat. It is predication Pr/out as OBJ(tltink). In order to ensure (tlrirrk), even though it satisfies OR, Kac adverts to ent ‘*(pp. 61 ff), which along with principles of priority ents (e.g., Sam, and Pr/uszt) are both candidates for element has priority over any smaller element it hecie conditions automatically select the entire predication Pr/eat as of thirrX both in sentences like (8) and in non-questions such as dkwntinuous
w
tLlrry
thou ht Sam ate the cake.
m~-y, the three types of rule are designed to account for a fairly wide variety phenomena in English. The elaborate treatment of nominal modification, for examp!e, invokes ihe constituent Pabel M (’ odifier’) in place of the usual recon;traint are reviewed in chapter cursive S of TG. The problems of the complex in support of the CORG analysis 4, where crock-linguistic evidence is brought t(: ominal modification as opposed to the inore traditional TG treatment, while coordination is addressed in the penultimate chapter. One has the imprcs\ion that Kac’s methodology involves the postulation of a set of rather straightforward surface interpretive rules, assisted by auxiliary principles such as the 0DC or the maximization principles. Accordingly, he must then find additional support for such principles, with the idea that the principles are auto~~at~ca!!~ justified once such independent motivation has been gathered up. This familiar strategy lies at the heart of TG as well, so that methodo!ogicd!!y the structure of argument% within CORG is virtually identical to that of TG. Of course, this is not surprising, given the autonomy thesis which is common to both CORG and TG : if no external adjudicating criteria are in principle allowed in assessing the claims of a linguistic theory, then some set of arbitrary, internal principles and rules of
176
Retie ws
argumentation must be adopted and adhered to. The point i trary as TG, and principles formulated within any auton ‘universal’ or ‘explanatory’ only in a trivial sense: they are within the confines of the theory and are not derived from or eve phenomena from any other domain. In his closing chapter, Ka of autonomy head on, espousing a strongly Bloomfieldian positio CORG, like other linguistic theories, is ,. . . . a theory of linguistic STRUCTURE, [ori inal emphasis] i.e. o tional principles of linguistic systems which must be undcrsto makes sense to speak of how languages are known. learned. or This view assumes that something called ‘linguistic structure’ pendently of language users, such that ‘sentences’ can be assigned bcorr~c itt racl/o. Yet it is abundantly clear that, once the autonomy p nothing from the domain of language acquisition, psycholinguistics;, socio or any other discipline can in principle be brought to bear in evrtl the theory. Only internal data can be used in assessing an auto which involve the formal apparatus of the theory in conjunc formulation assigns syntactic structure, grammatical relations, autonomy position permits one to choose basic principles (tra transformations, grammatical functions as primitives or as derived relatives solely on the basis of taste. Thus, once the autonomous linguist chooses his basic principles, rule types, etc., he is then free to explore the consequences theory with no responsibility to any other domain. The vacuity of such a p can be highlighted by comparing the autonomous linguist with some stra autonomous astronomer who might formulate a theory of galactic cluster access to, or respect for, independently established physical laws such servation of angular momentum. Our autonomous astronomer could sm some arbitrary principles, select some aesthetically satisfying goals, tind blithely ignoring the rest of physics. When such personalities do emcrg:r: in the natural sciences, they are usually laughed off the stage, even though the:, ?~~ig}~t obtain a following from the lunatic fringe. Yet the view that linguistics is an autonomous discipline seems to be widely, although thankfully not universally, accepted More remarkably, the autonomy position is sometimes even canonized as a fundamental principle. The autonomous linguist sets himself up in an enviable position: until he completes his analysis of ‘structure’ others must wait patiently before attacking the multitude of issues involving language acquisition and use. Of course, the past twenty years’ history in psycholinguistics has demonstrated the sterility of the autonomy position; many linguists now realize that a theory of ‘mguage structure cannot in principle be divorced from problems of language acquisition and use (cf. Derwing, in press). One obvious place where the failure in principle of the autonomy position is easily exposed is in the treatment of passives by various competing theories. CORG relates passives to actives by special riders placed on relation assigning rules. If Kac had chosen to treat codka in (4) as the grammatical subject of Ii&, with Ndrry in some sort cf ob!ique relation, he could have still related actives and passives, although at the semantic rather than relational level. That is, a rule can be formulated which
t assignment is concerned, but which ctives and passives in terms of, for ac arbitrarily rejects this possibility even though ‘8 number of his rules, eliminating the need for te’ and ’ pas6ve predicate’ riders which dot the CORG sl;ives in terms of ’ logical’ rather than ‘surface’ Hems which must be addressed in ad hoc ways by r there is, in fact, a great variety of htfoot (1979) have distinguisned stative, a~~_~~~ti~~-~i~~) passives m ‘transformational’ (e.g., action, rmer are the rrn-passives (The island is Ii as stat i\e p~si\~es (The jar is broken, of the latter. Although he does mention nsformat ional ’ passives, stating that in Subject and Object are treated in the same way as MN~Srh~cct rc-qXMi\;cly in act ivcs” (p, 45). Consequently, the sketch of a tip of the passive iceberg. Of course, stic theory has been constructed on the unsure basis of the transformas, only to founder on the sharp edges of the lexical passives. Classical he passive as the granddaddy of all and relational grammar, trace theory, and CORG all offer a. Given the autonomy position, it is crssible te, say that any of the analyses is correct or incorrect ; the best one can do s of the range of its descriptive power, The autonomy position serves only at the same time insulating it from e~crnal, empirical relevance. Autonomous theories are scientific only in their I!‘, on the other hand, a nsnautonomous position is espoused in which grammatical rults are understood as candidates for empirical, psychological content, then nonarbtrary adjudication is possible. For example, if a particular rule is taken as a candidate for psychological reality in the sense that it can be judged by external fitclors, then one can evaluate cer+ ,in treatments of grammatical phenomena on nds. Within the integrated (nontiutonomous) approach, external facts ccp$rical g acquisition and use inform linguistic theory, ,lust as the theory is an from langu t’sst:ntial contributor to empirical study. Such interaction is now common in psycho~in~~uistic~,although many autonomous linguists seem not to have noticed. And that * list might wake up to find himself so is a pity, for 01.1 day the autonomous Iin,.. entirely dispensable. autonomous as G is an interesting and potentially fruitful alternative to TG, In summary, although its autonomous stance makes it just as suspect empirically as the other windmills against which it tilts. Its major contribution, in my opinion, is that it focuses .ittcntion squarely on the syntactic surface and attempts to establish generalizations at that level without adverting to more abstract levels of syntactic representation, while at the same time incorporating a limited kind of functional information into the theory. I am somewhat suspicious of the abundance of principles and
conditions
which
based theory, other, hopefully C-rules
hinge on linear ordering.
and consequently,
I look forward
non-Indo-European,
of a language
yielding
languages.
like Japanese,
the imprcckn
to an application It would
if for no other
rctison
perturbations
such rules would
of change,
as all theories
are, and in a real sense it does complement
position
and more traditional is adopted,
References
only
internal
TG.
c;1usc. COKE
But the problem
t
t htin to disc
thwretical grammar
of t
bc interesting
is no doubt in t
remains:
tidiness makes a JifTcrenw.
hot
WNX the a~1~~~~6~~~~y