Grammatical Relations

The Evidence Against Their Necessity and Universality

D. N. S. Bhat

• Publisher: Routledge
• Year: 2002
• ISBN:
  9781134923755
• Edition: 1
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This book argues that the assumption that grammatical relations are both necessary and universal is an unwarranted generalization. The grammatical relations of subject and object are required in the case of the Indian language of Kannada. Furthermore, the notion of transitivity or transference which forms the basis for postulating grammatical relations does not play the expected central role in all languages: in the case of another Indian language, Manipuri, it is volitionality and transitivity which plays the central role in clause structure. Dr. Bhat argues against the universality and necessity of grammatical relations; his provocative hypothesis will be a challenge to all those concerned with the nature of language.

1
Introduction

1.1
CONTEMPORARY POSITION

Grammatical relations like subject, direct object and indirect object are regarded, either overtly or covertly, as language universal entities by most of the contemporary linguistic theorists (see Dixon 1979, Chomsky 1981, Bresnan 1982a, Perlmutter 1983, Gazdar et al. 1985). Even the few who do not consider them to be language universals do claim that the notion of transitivity or transference which forms the basis of the notion of grammatical relations, plays the central role in the clause structure of all languages. For example, the postulation of the three core arguments, symbolized as S, A and O (or P), as language universal entities, represents such a claim (see Comrie 1981, Foley and Van Valin 1984, Andrews 1985).

Grammatical relations are generally postulated as intermediary ‘abstract’ entities whose primary function is to relate semantic roles like agent, patient, experiencer, etc. with their formal representations like case markers (nominative, accusative, dative, etc.) or distinct positions in the sentence. Such intermediary entities are considered to be necessary because the linkage between semantic roles and their formal representations (case roles) has been found to be too complex to state otherwise in most of the familiar languages. It has also been found that the postulation of such intermediary entities is helpful in providing explicit and economical descriptions of the various morphosyntactic processes that occur in these languages.

Grammatical relations are also found to be helpful in establishing cross-linguistic generalizations. For example, according to Bresnan (1982b:5), individual languages differ in the ways in which these universal grammatical relations (which she calls grammatical ‘functions’) are encoded; configurational languages like English use constituent structure positions, whereas non-configurational languages like Malayalam use morphological case features. The use of universal concepts like subject and direct object, according to her, helps us easily to visualize the commonness between such varied uses of formal devices (see also Perlmutter (1983) for a similar opinion).

1.2
TWO QUESTIONS

There are two important questions that we might ask in connection with this postulation of grammatical relations as abstract intermediary entities:

1. Are they really necessary for describing the clause structure of all languages? If they are not, why have they become necessary in many of our familiar languages?
2. Is it possible to make use of them while describing the clause structure of all languages (in order to make cross-linguistic generalizations or in order to establish a Universal Grammar) even if they are not actually needed in some of them? That is, does the notion of transitivity, which forms the basis of these grammatical relations, play the central role in all languages, such that the latter notions can be established in all of them without any difficulty?

My answers to both these major questions are in the negative. In the first part of this monograph, I point out that there are languages like Kannada which do not need any intermediary entities for describing their clause structure; and in the second part, I point out that there are languages like Manipuri (and also Kannada) in which the notion of transitivity does not play the expected central role.

1.3
THE NEED FOR GRAMMATICAL RELATIONS

The crucial point that I will be trying to establish in the first part is that the need to establish grammatical relations as intermediary entities arises in languages like English mainly because the representations of two distinct types of sentential relations, namely semantic and pragmatic, have been mixed together and ‘grammaticalized’ in these languages (see 2.1 and 2.2). For example, English uses the preverbal position for encoding semantic relations like agent, patient, instrument, etc. (in different types of sentences) and also for encoding the pragmatic relation of being the sentential topic. Because of this combined encoding of the two, the actual relationship between form and meaning has become rather complex in the case of semantic relations in this language. Linguists have therefore found it useful to establish grammatical relations as intermediary entities for describing this relationship.

Whereas languages like Kannada use distinct sets of devices for encoding semantic and pragmatic relations, and hence the relationship between form and meaning in either of these two cases is simple and direct. In Kannada, for example, case suffixes and postpositions are used for encoding semantic relations, whereas word order is used for encoding pragmatic relations (see 3.2 and 3.3). There is, therefore, no need to establish any intermediary entities for describing these relationships. In fact, the postulation of such entities would only make the description unnecessarily complex.

Languages like English, which jointly encode semantic and pragmatic relations, have been found to use a single set of rules and principles for their various morphosyntactic processes, whereas the ones like Kannada which encode them differently have been found to possess two distinct sets of morphosyntactic processes (of semantic and pragmatic relevance) which use two different sets of rules and principles. It has been found convenient, in the case of the former type of languages, to use grammatical relations for describing the various morphosyntactic processes, whereas in the latter case, such abstract entities are completely unhelpful; one can, rather, use the semantic and pragmatic factors themselves for this prupose (see 3.4 and 3.5).

Linguists have generally found it convenient to exclude aspects of pragmatics from their theories of grammar, but in the case of languages like English (called ‘configurational languages’), they have not been able to exclude all the aspects of pragmatics from their theories. This is because some of the aspects of pragmatics, like the sentential topic, have got fused with semantic aspects in these languages as a result of their joint encoding with the latter as mentioned above. Whereas in languages like Kannada (called ‘non-configurational languages’) linguists have been able to exclude all the aspects of pragmatics, because these are encoded separately from semantic aspects. This has, however, given rise to certain distortions in the current formulations of linguistic theories, as seen, for example, in the postulation of configurationality for the ‘deep structure’ representation of semantic relations in non-configurational languages in spite of the fact that configurationality is not used overtly by them for this purpose (see 2.3 and also 4.4.1).

I also point out in this second chapter that the two major types of case-marking systems that have been generaly recognized by contemporary scholars, namely accusative and ergative, actually represent two different ways in which languages can combine together the encoding of semantic and pragmatic relations (see 2.3.2). However, such systems do not include languages like Kannada, in which the two are encoded distinctly.

In order to illustrate these various claims regarding the necessity (or otherwise) of grammatical relations, I present a detailed description of Kannada clause structure in the third chapter.

1.4
THE PRIMACY OF TRANSITIVITY

In the second part of this monograph, I try to show that the notion of transitivity or transference, which forms the basis for the postulation of grammatical relations or of other related concepts like S, A and O, does not play the expected central role in all languages. As pointed out by Hopper and Thompson (1980), transference is only one of several components that co-vary; there are contexts in which two or more of these components (like transference and volitionality) may conflict with one another. My contention is that in such contexts, languages differ from one another in resolving the conflict in favor of one or the other of these components (see 4.2). For example, Kannada resolves such conflicts in favor of volitionality and not transference. Hence, we can only regard volitionality to be playing the central role in Kannada.

Secondly, attempts to analyze the single argument of non-volitional verbs as underlying objects (4.3.1) and of certain psychological predicates as underlying subjects (4.3.2) are in conflict with the actual situation as it exists in Kannada. This is also true of the attempt to regard the so-called ‘active’ languages as having a ‘split-S’ system (4.3.3). The failure of such attempts can be regarded as resulting from the fact that the notion of transference does not play a central role in these languages.

Whereas the notion of ‘external’ argument (Williams 1984) is shown to be basically a pragmatic notion. It has actually no place in theta-theory. Languages may differentiate between sentences which externalize one of their arguments, on the one hand, and ones which do not on the other hand; this distinction corresponds to the one made by some logicians between categorical and thetic judgements (4.4).

In order to illustrate some of these claims about the notion of transitivity and external argument, I present a detailed description of Manipuri clause structure in the fifth chapter.

1.5
TWO APPENDICES

There have been two different recent attempts to justify the postulation of grammatical relations in Dravidian languages, both of which are of direct relevance to this study. They are (a) by Sridhar (1976, 1979), who argues that the dative noun phrases occurring in certain predicates of mental and physical states are to be regarded as underlying subjects in Kannada; and (b) by Mohanan (1982), who argues that the thematic (semantic) roles and case features are to be linked through the intermediacy of grammatical relations in Malayalam (a language which is closely related to Kannada).

I propose to examine these two claims in the appendices which immediately follow the second part of this monograph. My contention is that neither of these claims can be regarded as tenable.

Part I
Evidence against the necessity of grammatical relations

2
The need for grammatical relations

2.1 DIFFERENTIATING BETWEEN SEMANTIC AND PRAGMATIC RELATIONS

Sentences can generally be analyzed as consisting of a set of arguments (noun phrases) and a predicate. There are some exceptions to this general rule in some languages, such as the equational sentences of Kannada, in which there is no predicate, or the ‘weather’ sentences of the same language in which there is no argument (see 3.2.1). The arguments occurring in a sentence have to represent two main types of relations called ‘semantic’ relations and ‘pragmatic’ relations, of which the former relate the arguments with the predicate and the latter with the speech context.

Semantic relations deal with the basic ‘conceptual’ meaning of a given sentence. Their number and type is determined by the kind of predicate that occurs in the sentence, or rather by the kind of event or state that the predicate denotes. For example, the verb give would generally require three different semantic relations, namely the giver (agent), the receiver (recipient) and the object given (patient), whereas the verb cry would require only one: the person who cries (agent).

Pragmatic relations, on the other hand, deal with an entirely different type of meaning: namely, the way in which these different arguments are related to other arguments that occur in the speech context (for example, the ones occurring in the preceding sentences) and also with the participants themselves of the speech act, such as the speaker and the addressee. This latter type of relation is regarded as dealing with the organization or ‘packaging’ of the arguments concerned (Foley and Van Valin 1984). For example, a given argument may have to refer to an individual or object that is already being talked about in the speech context or to one that is being newly introduced into the conversation; it may have to refer to an individual that needs to be specifically emphasized or to one that may be left in the background; and so on.

The difference between these two types of relations can be seen clearly in the following set of Kannada sentences:

See Table

All these sentences contain the same set of three semantic relations, namely the agent (avanu ‘he’), recipient (nanage ‘to me’) and patient (pustaka ‘book’); they are represented by the same set of case markers in these sentences: the nominative (unmarked), dative (ge) and accusative (annu) respectively. (Notice that the accusative marker is overt, in the case of inanimate nouns, only when the noun is immediately followed by a word or particle other than the verb (as in (1b)) or when it is being emphasized.) However, the four sentences (1a–d) are different from one another in the kind of pragmatic relations that get associated with their arguments; for example, the topic of the sentence (i.e. the argument which connects the sentence with the speech context) is avanu ‘he’ in (1a), pustaka ‘book’ in (1b) and nanage ‘to me’ in (1c); the whole sentence, exclusive of the argument pustaka ‘book’ is the topic in (1d); in this latter case, the excluded argument receives special emphasis through a syntactic process called ‘clefting’, and functions as the focus.

Languages may use two distinct sets of grammatical devices for denoting these semantic and pragmatic relations as seen in the above-mentioned sentences of Kannada, or they may mix them up in different ways and use, mostly, the same set of grammatical devices for denoting both of them. Since the former situation would lead to greater clarity in a speech act, one would normally expect it to be the preferred one, and the latter to be the ‘marked’ one.

However, linguists have generally considered the latter situation to be the most common (and ‘unmarked’) one among natural languages. For example, Givón (1984) finds most languages to be shying away from double-marking a nominal argument for semantic and pragmatic relations; Andrews (1985:63) considers it difficult to provide a coherent account of how the semantic and pragmatic functions are signaled in terms of a direct connection between coding features and the functions they express. The question that I wish to raise here, especially in vie...

Table of contents

• Cover Page
• Title Page
• Copyright Page
• Preface
• Abbreviations
• 1 Introduction
• Part I Evidence against the necessity of grammatical relations
• Part II Evidence against the universality of grammatical relations
• Appendix 1 Dative noun phrases as underlying subjects in Kannada
• Appendix 2 Grammatical relations in Malayalam
• References