On Case Grammar

Prolegomena to a Theory of Grammatical Relations

John Anderson

• Publisher: Taylor & Francis
• Year: 2018
• Print ISBN: 9781138624351
• eBook ISBN: 9780429864988
• Edition: 1
• Topic: Literature
• Subtopic: Literary Criticism
• Index: Literature

Originally published in 1977, On Case Grammar, represents a synthesis of various lines of research, with special regard to the treatment of grammatical relations. Arguments are assessed for and against case grammar, localism, lexical decomposition and relational grammar. The book surveys the important evidence to support the validity of the choice of a case grammar as the most satisfactory of current accounts of the notion of grammatical relations. This evidence is derived from a detailed examination of various processes in English and from a typological comparison of other languages, notably Dyirbal and Basque. The book also looks at the establishment of principled limitation on the set of case relations. Lexical, syntactical, semantic and morphological evidence suggests that the set of cases is in conformity with the predictions of a strong form of the localist hypothesis, which requires that case relations be distinguished in terms of source vs. goal vs. location.

1 GRAMMARS OF CASE

1.1 A Traditional Notion of Case

The grammatical terminology of most languages which incorporate the European tradition in such matters displays a systematic ambiguity in the use of the term ‘case’. Usually, it is employed to refer both to a certain inflexional category (and the forms that manifest it) and to the set of semantic distinctions carried by the forms of that category. We can differentiate these as case-forms and case-relations or case-functions respectively. Thus in Lutetiam veni, it might be said that the noun is in the ACCUSATIVE form and that in this instance it indicates, or functions as, the ‘goal’. Much controversy has depended simply on the confusion of these two senses. Certainly, such an ambiguous usage has the disadvantage that case-functions clearly can be expressed in other ways, notably by prepositions or postpositions, by word order or in the morphology of the verb rather than the noun. In what follows I shall use the term CASE-FORM (henceforth CF) more inclusively, to cover any form that serves to express a CASE-RELATION (henceforth CR), where the latter are interpreted as labels for the semantic role that a particular NP fulfils in the predication.

Most traditional accounts posit a complex mapping between the set of CRs and the set of CFs (either in the narrow sense or interpreted, as I propose, more widely). For instance, the same accusative form of Latin we noted above as a marker of the goal relation can also express the (DIRECT) OBJECT, as in puellam amo. And the goal can alternatively be associated with a form which includes a preposition: ad urbem veni There have been rather few attempts to arrive at more NATURAL accounts, that is, descriptions of the CR/CF relationship which involve less of a discrepancy. Notable, however, have been various LOCALIST proposals, the character of which we shall return to below.

All such accounts encountered considerable difficulties in attempting to provide a unitary, or even unified function for such CFs as (particularly) the accusative and NOMINATIVE. Even if the goal/object distinction is disregarded, it is apparent that it is difficult to attribute to the objective accusative itself a constant semantic value, as reflected in a set like that in (1):

(1) a. The policeman struck the student

b. Marilyn gave John the whisky

c. My uncle built a chalet

d. John killed Bill

e. Columbus discovered America

f. The procession crossed the square

g. Fred left home

Compare the even greater diversity of the SUBJECTIVE nominatives in (2):

(2) a. The student was struck by the policeman

b. The student tickled the policeman

c. Nobody knew the truth

d. That trunk contained eight books

e. My dentist suffered terribly

f. John received the whisky from Marilyn

g. The truth was known to nobody

In each instance, the set of NPs partake of a range of semantic roles, some of them traditionally distinguished as subtypes of subject or object, such as ‘object of result’, ‘indirect object’, etc. But what (if anything) these different types of subject or object have in common semantically has remained uncertain.

Thus, to be more particular, grammars written within the classical tradition almost invariably have extensive sections which under one guise or another document at some length the often multifarious ‘uses of the cases’. So in Gildersleeve & Lodge (1895), for instance, out of fewer than 200 pages devoted to the ‘simple sentence expanded’, the description of the cases in their role of ‘qualification of the prediction’ is accorded almost 100. And Woodcock’s (1959) more recent though still traditionalist treatment devotes five of its total of twenty-five chapters to ‘the functions’ of the cases. For the individual cases Gildersleeve and Lodge are typical in providing a detailed and intricate classification of ‘uses’. The accusative, for example, may mark either an inner or outer object (or both in the same clause); and the former divides into the object of result, the cognate accusative and the accusative of extent, the latter may be partitive or not; and so on.

Often, however, such classifications appear to impute to the cases (case forms) distinctions which are signalled elsewhere. For example, it could be argued that rather than an object of result in Latin, we have ‘creative’ verbs (like English make) whose objects (or rather their denotata) come into existence as a result of the action denoted by the verb. Is it then necessary to recognise a distinct relation/use/function? Many, perhaps all, of the object ‘uses’ are reducible in this way. A more interesting hypothesis would involve (as we have observed) minimising such recourse to syncretism. And it is some such natural view which has led to attempts to discover where possible a Grundbedeutung (or at least a single source, be it semantic or syntactic) for linguistic elements, including the case forms (cf. e.g. Jakobson, 1936). For the Latin cases this is illustrated already by the work of Key (1958) and Laurie (1859). The latter in particular pursues essentially a localist strategy such as we shall investigate in ch.2. Typically, however, the form that marks the subject in a language appears to be non-reducible. Apart from correlating with subject-hood such a form simultaneously neutralises uses distinguished in other (non-subject) constructions (as in (2)). Nevertheless, it would seem obvious that the least we can expect of any grammatical theory as far as cases are concerned is that it should provide a principled articulation of the relationship between case relation and case form, distinguishing neutralisation and syncretism from the natural. But this is lacking throughout much of the history of studies of case. The two characteristic polarisations are aptly summarised by Haudry (1968, 141):

Le problème central de 1’étude des cas est dans la difficulté qu’on éprouve à les décrire en termes de ‘signes’, c’est-à-dire à poser en face de chaque signifiant (la désinence), un signifié correspondant. La grammaire historique oscille entre deux attitudes: admettre une polysémie du cas, en la justifiant par un syncrétisme, ou tenter de réunir sous un concept les emplois les plus divers.

It is, however, not my intention here to discuss such traditional preoccupations at any length. Rather, I take as familiar such a context for the following investigations, which are concerned in the main with a consideration of specific proposals made within current frames of reference.¹ My concern is simply to sketch in something of this immediate background as a preliminary to an examination of relevant aspects of contemporary linguistic theory. In particular, we shall be concerned with some developments in what has come to be called ‘case grammar’; and I shall attempt to maintain the adequacy of a variant of this especially in the face of objections that have been raised to particular aspects. In such a discussion, nevertheless, the traditional theme that we have broached in the present section, viz. the articulation of the relationship between CFs and CRs, will underly much of the debate.

1.2 Case relations and the Aspects theory

1.2.1. It might have been anticipated that the development of generative grammar, and in particular the elaboration of the distinction between deep and surface structure, would lead, if not to more natural accounts, at least to an articulation of how the neutralisation of roles typical of subjects and objects might arise. But this was not the case. In the framework established in the early 1960s and culminating in the position formulated in Chomsky (1965) (henceforth ‘the Aspects theory’), the ambivalences associated with surface subjects and objects were carried over into their deep equivalents. Transformations like the passive simply redistribute the NPs in a sentence without increasing or decreasing their ambivalence with respect to CRs. And this is not surprising. The passive, for instance, was conceived of as having purely syntactic motivation. What remains mysterious, however, is exactly what this motivation might be.

The arguments invoked by Chomsky (1957, §§5.4, 7.5) involve the avoidance of a duplication of selectional restrictions (in ‘corresponding’ actives and passives) and the difficulties in specifying restrictions on the distribution of the passive marker be … -en, together with the problem created for an analysis whereby passives are not derived from the structure more immediately underlying actives but rather vice versa by there being ‘no structural way to differentiate properly between [The wine was drunk by the guests] and [John was drunk by midnight] if both are taken as kernel sentences’ (1957, 80). I cannot see that this last assertion can be accorded any force in the absence of any supporting arguments for an analysis which fails to make such a differentiation. Why should we accept that there is ‘no structural way to differentiate properly’ between such sentences? Moreover, even if it is just, it and the other arguments carry weight only within a grammar one of whose goals is the establishment of a set of ‘kernel sentences’, which are chosen ‘in such a way that the terminal strings underlying the kernel are easily derived by means of a [phrase structure] description, while all other sentences can be derived from these terminal strings by simply statable transformations’ (Chomsky, 1957, 48). And they are decisive only against the alternatives envisaged by Chomsky, viz. an analysis whereby ‘the passive of “John loves Mary” would be “John is loved by Mary” ’ (1957, 78) and one in which actives are derived transformationally from passives.

But other possibilities can be and have been envisaged. The original considerations adduced by Chomsky do not enable us to select the simplex-sentence transformational accounts of the derivation of passives over a lexicalist formulation (Freidin, 1972a, 1975; Starosta, 1973) or a complex source in which the ‘active’ predication is embedded as a complement to the verb be (Hasegawa, 1968; R. Lakoff, 1968) or one in which neither the active nor passive variant is structurally closer to their source (Fillmore, 1968a). In Aspects, Cho...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Original Title Page
• Original Copyright Page
• Table of Contents
• Dedication
• Preface
• 1. Grammars of Case
• 2. Localist Case Grammar
• 3. On the Nature of Grammatical Relations
• Notes
• Bibliography
• Index