Introduction

A number of linguists have used the notion of a ‘hearsay’ particle in analysing particular languages, e.g. Ballard (1974), Barnes (1984), Chafe and Nichols (1986), Derbyshire (1979), Donaldson (1980), Givón (1982), Haviland (1987), Hewitt (1979), Höhlig (1978), Laughren (1981), Levinsohn (1975), Lowe (1972), Palmer (1986), Slobin and Aksu (1982), Thomas (1978), Willet (1988). While the exact uses of ‘hearsay’ devices vary from language to language, it is said that their main function is to mark information which the speaker got from somebody else. The data on hearsay particles have generally been rather fragmentary, and the notion of a hearsay particle is typically left unanalysed. My aims in this chapter are twofold: first, to provide a fuller range of data on one particular hearsay particle – rέ, from Sissala: and second, to use rέ to choose between two competing accounts of the nature and function of hearsay particles in general.

What is the minimal hypothesis one might make about hearsay particles, given only the informal observation, noted above, that hearsay particles are used to mark information that the speaker got from somebody else? The minimal hypothesis would be, I think, that they should be used only for reporting actual speech. Reported thought would be excluded, and the status of paraphrase, or speech that is attributed by inference without actually being heard, would be unclear. Typically, work on hearsay particles reports a number of uses which do not fit the minimal hypothesis. My data provide clear counter-examples to it.

Apart from the minimal hypothesis, I know of two main accounts of the nature and function of hearsay particles.

Introduction

Sissala has three different forms of ‘and’, whose use is syntactically conditioned: ká is used to conjoin Ss, a is used to conjoin VPs, and rí, or arí, is used elsewhere. The different co-ordination constructions also differ in their pragmatic effects: for example, sentential co-ordination with ká might be analysed as suggesting that the event described in the second conjunct was unexpected, whereas non-sentential co-ordinations carry connotations of stereotypicality. The question I shall consider in this chapter is how these pragmatic differences can be explained.

In principle, such differences in pragmatic effect might be traced to any of three sources. First, the various co-ordinating conjunctions might differ in their truth-conditional meaning: use of ká, for example, might entail that the event described in the second conjunct was unexpected. Second, the co-ordinating conjunctions might have a common truthconditional meaning but differ in their non-truth-conditional meaning: ká, for example, might carry a constraint on relevance (as discussed in chapter 4), specialising it for use in contexts in which an element of unexpectedness was presupposed. Third, the pragmatic differences among conjoined structures might arise not from the lexical meanings of the co-ordinating conjunctions, but from syntactic factors. This is the idea I shall pursue.

My claim will be that where the speaker has a choice between the co-ordination of non-verbal constituents (NVCs), VP co-ordination and S co-ordination; the form he chooses will follow from considerations of relevance. Since S co-ordination requires more linguistic processing effort, it will only be chosen to achieve some extra effect; i.e. some effect that could not have been achieved by the less costly NVC or VP co-ordination.

Introduction

In the last two chapters I have shown that while some linguistic expressions can be analysed in purely truth-conditional terms, others – the semantic constraints on relevance – cannot. This raises an interesting theoretical question: how does one decide whether a given linguistic expression is truth-conditional or not? In this chapter, I will propose tests for distinguishing between truth-conditional and non-truth-conditional uses and look at a particle – baa – which appears to have both types of use. The existence of such particles raises a second interesting theoretical question: must they be analysed as genuinely ambiguous between truth-conditional senses, or can they be treated as merely vague?

My answer to these questions will be as follows. First, baa is not multiply ambiguous: it has two core meanings, one truth-conditional and the other non-truth-conditional, from which the full range of uses can be predicted. Second, it is genuinely ambiguous between truth-conditional and non-truth-conditional uses. However, the two core meanings are clearly related, and this type of ambiguity or polysemy between truthconditional and non-truth-conditional senses is quite common. I shall end with some speculations on why such ambiguities are found.

Uses of baa

Temporal uses

Baa has a variety of temporal uses. In all of these, it is positioned consistently before the VP. Its semantic interpretations are discussed below.

The iterative and restitutive use

Both baa in Sissala and again in English may have an iterative or restitutive interpretation. What exactly I mean by this will be explained by reference to the examples. The following examples have an iterative interpretation.

Introduction

In chapter 5 I showed how a particle can be ambiguous between truth-conditional and non-truth-conditional uses, and vague between different truth-conditional uses. In some respects wuu, which can mean ‘all’, ‘every’, ‘each’, ‘any’, ‘whole’, ‘very’ and ‘always’ raises similar issues. What makes it interesting to raise the question of vagueness versus ambiguity again is the fact that while some of the truth-conditional uses can be represented within predicate calculus, others cannot – which underlines the inadequacy of a purely predicate-calculus treatment of truth-conditional quantificational phenomena.

Though there is no question that all the quantificational (nonidiomatic) uses of wuu make a contribution to truth conditions, the question arises how the differences between the ‘all’, ‘every’ and ‘each’ interpretations are to be represented. They cannot be represented in terms of differences in quantifier, or in quantifier scope: rather, they appear to involve differences in the way the given quantifier is viewed: collectively or distributionally; ‘one by one’ or ‘one amongst many’. I will argue that these differences in interpretation are genuinely truth-conditional, and cannot be seen as involving constraints on relevance, as discussed in chapters 4 and 5. Since such differences cannot be captured in predicate-calculus terms, this provides further confirmation of the inadequacy of predicate calculus for natural-language semantics.

A related issue has to do with the domains of wuu – that is, the sets of individuals quantified over – how they are established and how they are interrelated in discourse. I shall discuss the fact that they seem to be indeterminate in some cases, and that speakers and hearers may not have the same assumptions about a domain; I shall also discuss the case where domains are ‘loosely’ fixed.

Introduction

The area of non-truth-conditional particle phenomena as discussed by Grice (1975), Karttunen and Peters (1975) and Blakemore (1987) is a relatively new field of research. Little is known about how such phenomena arise, their range of possible functions and their distribution across the languages of the world.

If it is true, as was suggested in chapter 1, that all humans have the same logical abilities, and that communication creates a presumption of adequate contextual effects for the minimum justifiable effort, should we not expect to find in every language similar linguistic phenomena which save the hearer processing effort by guiding him towards the intended range of contextual effects? In other words, could the very fact that humans are constrained by considerations of relevance be the basis for a typology of non-truth-conditional particle phenomena? This chapter is an attempt to show that this could well be the case.

After reviewing some theoretical accounts of non-truth-conditional particles in English, I will look in detail at the behaviour of a variety of such particles. I will discuss at some length the analysis of also in English, auch in German and its equivalent má in Sissala. I will argue that auch and má have a range of functions not shared by also in English, and I will offer an explanation of this fact in relevance-theoretic terms.

Other particles I will look at more briefly are so, after all, you see and even in English, and their equivalents in Sissala. I will try to show that relevance theory sheds light on the analysis of such particles.

Introduction

Sperber and Wilson (1986a) point out that, as a general tendency or in given situations, humans pay attention to some phenomena rather than others; they represent these phenomena to themselves in one way rather than another; they process these representations in one context rather than another. What determines which phenomena they attend to, and what representations and contexts they construct? Sperber and Wilson suggest that there is a single general answer to these questions. Humans tend to pay attention to what is relevant to them; they form the most relevant possible representations of these phenomena, and process them in a context that maximises their relevance. They claim that relevance, and the maximisation of relevance, is the key to human cognition. This has consequences for the communicator: by demanding attention from the audience she suggests that the information she is offering is relevant enough to be worth the audience's attention. Thus relevance is the key to communication too. In this chapter, I hope to explain and justify these claims.

By showing why speakers might be expected to aim at a certain standard of relevance if they want to keep their audience's attention, Sperber and Wilson remedy one of the main inadequacies of Grice's pragmatic approach: Grice provides little justification or explanation of his own pragmatic maxims. I will show later on in this chapter that their criterion for utterance interpretation is strict enough to eliminate all but a single interpretation, thus remedying another of the main weaknesses of Grice's approach.

In a later section, I will show how relevance theory provides a better theoretical foundation for comprehension and textuality than coherence models.

The main aim of this book, which developed out of my Ph.D thesis (Blass 1988), is to show how relevance governs discourse.

Although the book makes a contribution to the semantic and pragmatic analysis of a particular unstudied language – Sissala – this is not the only or most important aim of my study. My main intention is to show, using Sperber and Wilson's relevance theory (1986a), that connectivity in discourse is a pragmatic rather than a semantic matter: it results from relevance relations between text and context rather than from relations linguistically encoded in the text.

In the first two chapters, I discuss a variety of approaches to the analysis of discourse, and argue that in the current state of knowledge, relevance theory offers the only possibility of a genuinely explanatory account. In the remainder of the book, I look at a variety of semantic and pragmatic phenomena, and try to show not only that they can be analysed perceptively in terms of relevance theory, but also that their analysis has interesting implications for the theory I have used.

If it is not my main aim to provide a study of some phenomena in Sissala, why did I choose a so-called ‘exotic’, unanalysed, language rather than English or German (my mother tongue) to make my point? The importance of the book lies primarily in what it has to say about pragmatic theory, and in particular in its demonstration that although there may be enormous variation in cultural backgrounds, the principles by which hearers use contextual information in interpreting utterances in discourse are universally the same.

Introduction

Every language with referential expressions seems to have some means of indicating to hearers not only that a referent of a certain type is required, but also where to search for appropriate referents, and how easily accessible they are. Within the framework of relevance theory, such expressions, in at least some of their uses, should be analysable as semantic constraints on relevance, contributing not to truth-conditional content but to reducing the hearer's processing load. In this chapter, I will look briefly at some Sissala data of this general type.

It has long been thought that defining expressions – determiners – play an important role in this aspect of processing (see Clark and Clark 1977:466–7; Clark and Marshall 1981; Hawkins 1978; Nunberg 1977). Thus, according to Hawkins a speaker using a definite article has the following in mind:

1. a. He introduces a referent (or referents) to the hearer.

2. b. He instructs the hearer to locate the referent in some shared set of objects.

3. c. He refers to the totality of the objects or mass within this set that satisfy the referring expression. (1978:125)

In English, a noun phrase containing the indefinite article a(n) can be used with specific or non-specific reference. Either can be the basis for further definite reference. Thus in an example like (1)

1. A colleague of mine has just given me a call,

the speaker might have wanted either to introduce a specific colleague or just to introduce the referent in a non-specific way.

Discourse and discourse analysis

We all know that hearers have intuitions about the grammatically of sentences. It is also true that they have intuitions about the wellformedness of discourses. However, are we talking about the same kind of well-formedness judgement? Some seem to think this is the case. I quote Werth (1984:17): ‘We will argue that connectivity at both levels – that is, both between and within sentences – is crucially semantic, and that the obvious formal links which exist are in fact the overt manifestations of semantic connections.’ Those who propose discourse and text grammars, such as Harris (1952), Longacre (1983) and proponents of discourse-production models, think along similar lines.

However, the fact that utterances of single grammatical sentences can themselves constitute acceptable or unacceptable, well-formed or illformed discourses, can be used to show that there is a fundamental difference between judgements of grammaticality and judgements about the acceptability or well-formedness of discourses.

Consider examples (1)–(4):

1. A dog is for life.

2. Trespassers will be prosecuted.

3. She can't.

4. Every chair has a nose.

All four are grammatical sentences of English: but which of them constitute complete and well-formed discourses? This question cannot be answered without taking background or contextual assumptions into account. Although it may be hard to understand in isolation, we realise that (1) is a complete and acceptable discourse when, for instance, we see it on a sticker on the back of somebody's car.