INTRODUCTION

What would count as a complete theory of conditionals? One goal for such a theory is to answer the following two questions:

1. (i) What do conditionals mean?

2. (ii) What are their logical properties?

These are matters of logical and linguistic analysis: they concern human competence. However, a complete theory of conditionals should also answer two psychological questions:

1. (iii) How do people understand them?

2. (iv) How do people reason with them?

These are matters of human performance that call for the investigation of mental processes.

There are a number of theories that provide answers to some of these four questions. Yet, despite the conceptual analyses of philosophers and logicians, the semantic and syntactic studies of linguists, and the experimental investigations of psychologists, there is no single existing theory that provides a unified and complete account of both competence and performance. My aim in this paper is accordingly to make progress towards such a theory – a theory that concerns the everyday interpretation and use of conditionals, not an idealized philosophical concept, and one that is intended as a contribution to cognitive science.

The paper has four parts. The first part considers how ordinary individuals reason with conditionals, and it describes the main approach that psychologists have taken to deductive reasoning – the theory that there are formal rules of inference in the mind. It argues, however, that this view is mistaken and that inference depends instead on a search for ‘mental models’ of premises that are counterexamples to putative conclusions.

INTRODUCTION

Terms like ‘conditional’, ‘temporal’, ‘causal’, ‘concessive’ are part of the terminological inventory that traditional grammar makes available for the characterization of adverbial clauses. The distinctions drawn by these terms seem clear enough until an attempt is made to explicate them in a way that would have crosslinguistic validity, or to apply them in an exhaustive characterization of all kinds of data within a single language. To begin with, there are nonspecific constructions to which several of these terms or none at all seem to be applicable. Examples are nonfinite constructions such as adverbial participles in Russian, the gerundio in Italian, the construction tout en V-ant in French or participial constructions in English. A sentence like the following may have a conditional or a causal interpretation depending on the context:

1. (1) Lacking that, the movement is dead

Or, to give an example of a different type, the construction Adj as NP be in English merely expresses factuality and is open to both a causal and concessive interpretation:

1. (2) a. Rich as he is, he spends a lot of money on horses (causal)

2. b. Poor as he is, he spends a lot of money on horses (concessive)

But even cases where the adverbial relation in question is marked by a conjunction are sometimes difficult to assign to one rather than another category.

INTRODUCTION

For those of us involved in the attempt to spell out the relation between statements and those aspects of reality they are about, conditionals are a thorny issue. Within this semantic tradition, common wisdom can be summarized rather contentiously as follows: classical model theory gives us the semantics of the material conditional. It works fine for mathematical conditionals, but is a disaster if applied to ordinary language conditionals, especially counterf actual conditionals. Within the possible worlds framework, there are various treatments, some of which are quite successful for certain types of natural language conditionals, including counterfactuals, but they are all a disaster when applied to mathematical conditionals. I will give examples of both sorts of failures below.

My own opinion is at odds with this view of where things stand. I think that the language of mathematics is continuous with ordinary language, since discourse about mathematical objects and mathematical activity takes place in English or some other natural language. While it would not be appropriate to argue for this commonsensical but unfashionable position in this paper, it is appropriate to point out the consequences for a theory of conditionals. To one who takes this line, mathematical conditionals simply are natural language conditionals, so an adequate account of mathematical conditionals must be part and parcel of an adequate account of natural language conditionals. Hence, neither of the approaches to the conditionals mentioned above is at all satisfactory and the lack of an adequate account of the semantics of conditionals is a major embarrassment.

INTRODUCTION

The historical study of conditional sentences in a particular language or language family is a complex and difficult task. One of the major reasons for this is the nondiscrete nature of the category involved, in that the meaning of conditional sentences seems to shade off imperceptibly into adjacent semantic areas, in particular those of concession, cause and time. Equally, even where, as in the case of Romance, there is one favoured structure for conditional sentences (a biclausal sentence incorporating a protasis introduced by the conjunction si), this will not always carry the relevant value, while conversely there will be other structures with diverse functions which can and do in certain circumstances serve to mark a hypothetical antecedent–consequent relation.

INTRODUCTION

What I want to show in this chapter is that conditionals do not belong to the static domain of mathematical logic, but to the dynamic domain of discourse where individuals with different belief systems confront each other now.

I will demonstrate that we must consider discourse factors in (i) the preceding context and (ii) the speaker's attitude; and also that there is a connection between p and q, that is, every construction with the meaning ‘if p, q’ shares an abstract, grammatical meaning similar to ‘correlation/correspondence between p and q’. The evidence will be developed as follows: section 2 will examine two types of English conditionals, both of which have generally been regarded as counterexamples to the ‘connection’ theory; section 3 will show that consideration of factors in (i) and (ii) leads us to distinguish information and knowledge; section 4 will show that this distinction leads us to reject Haiman's (1978) view that conditionals are givens; section 5 is a conclusion.

INTRODUCTION

The recurrent interchangeability or identity of conditional and interrogative markers has been noted in a number of unrelated languages, among them the members of the Uralic family (Beke 1919), Germanic, French and Greek (Havers 1931: 21) and Chinese (Chao 1968: 81–2). The phenomenon is explained on the assumption that conditional protases are the topics of the sentences in which they occur (Haiman 1978; for further discussion, see Akatsuka, Ford and Thompson in this volume). As topics constitute information whose validity is (perhaps only provisionally) agreed upon by all parties to the discourse, it is natural for a speaker to establish their given status by asking for assent or recognition from his interlocutor. In some languages, the semantic equivalence of protasis and topic is directly reflected by the identical morphology and syntax of these two categories: representative examples are Turkish (Lewis 1967: 217), Tagalog (Schachter 1976: 496), Tabasaran (Magometov 1965: 271), Korean (Martin and Lee 1969: 146, 159), Vietnamese (Hoa 1974: 103, 341), Middle Egyptian (Gardiner 1957: 125) and, once again, Chinese (Chao 1968: 81–110).

INTRODUCTION

This paper is concerned with the semantic interpretation of generic expressions in conditionals and with their interaction with temporal adverbs and tense, assuming Situation Semantics as a general framework for a model-theoretic semantics of natural language. As a theory of meaning and interpretation which attributes to the context of an utterance an important role in its interpretation, this recently developed semantic theory will provide new and fruitful concepts for analysing the use of generics in natural language and their role in structuring meaning as a relation between expressions and situations in the external world. The issues concerning generics are presented as informally as possible at first, to clarify the underlying intuitions. Reilly's notion of ‘protogeneric’ (this volume) is discussed as a form of contextually restricted generic expression and the general question is addressed of the conditions under which when and whenever are interchangeable with the conditional if–then without distortion of meaning. Conditionals and generics are shown to share an important semantic property: persistence of expressed information, or insensitivity to putative counterexamples.

INTRODUCTION

I shall be concerned with the use of conditionals in inducements, conditional promises and bribes, and their use in deterrents, conditional threats and warnings. This paper will examine the logic and possible phrasing of such conditionals the principal function of which is purposive, i.e., constitutes an attempt on the part of the speaker to get the addressee to do something (If you fix the car I'll give you $100) or to refrain from doing something (If you come any closer I'll shoot). It is hoped that the account to be developed here will provide an analysis for this special class of speech act conditionals, and serve in some measure as a model for approaching the study of other sorts of conditionals; also that some of the kinds of considerations that emerge as critical here, e.g. the importance of knowledge of the contents of the p and q propositions involved in the conditional, will be of more general relevance.

Conditional promises and threats clearly involve something more than the statement of a contingency between the p and q propositions involved, more even than the statement of some causal connection between these propositions.

INTRODUCTION

This chapter focuses on an analysis of the conditional in Classical Greek, generally excluding the preceding Homeric period and the following Koine, both of which show differences in the relevant constructions from the intervening Classical period. Of course, many of the properties of the Greek conditional are not unique to that language. However, it does command special interest for two reasons, its complexity and the fact that it has been so intensively investigated. Apart from the specific hypotheses, the central point is that it is incumbent on the linguist to account for the formal similarities among constructions, and to employ in addition to hypotheses stemming from formal logic those arising from semantic similarities based on the typical factors found in semantic change in general.

INTRODUCTION

Conditional (if–then) constructions directly reflect the characteristically human ability to reason about alternative situations, to make inferences based on incomplete information, to imagine possible correlations between situations, and to understand how the world would change if certain correlations were different. Understanding the conceptual and behavioural organization of this ability to construct and interpret conditionals provides basic insights into the cognitive processes, linguistic competence, and inferential strategies of human beings.

The question of what a conditional construction is may be answered in many different ways, and from many different perspectives. The linguistic characterization of conditionals in different languages provides the basis for linguistic universals, which presumably at least in part constrain the way we reason. The diachronic point of view provides knowledge of the possible adaptations that a system of conditionals may undergo, and may detect dependencies on developments in other linguistic domains. Studies of language acquisition provide additional perspectives on a linguistic system, offering not only developmental data but also insights into the basic components and relationships of the adult system.

INTRODUCTION

The aims of this paper are, first, to attempt a characterization of conditionals that has crosslinguistic validity and, second and more important, to try to identify the significant parameters in the crosslinguistic description of conditionals, looking both at properties that are common to all languages and at properties that show significant crosslinguistic variation. The claim that a given parameter of variation is significant is, of course, an empirical claim, and it may well be that in further work on this topic other parameters, of which I am unaware or which I consider insignificant, will need to be added to my list.

Two general remarks are necessary before embarking on the characterization of conditionals and crosslinguistic variation within conditionals: these relate to the general problem of isolating a given construction, both intralinguistically and interlinguistically, and to the general problem of identifying the meaning of a construction. I assume that a given construction is to be identified, in general, in terms of a prototype rather than in terms of necessary-and-sufficient conditions. Thus, I will not be surprised if some sentences having the form of prototypical conditionals in a given language do not in fact receive the interpretation of conditions (as when English If you do that, I'll hit you is interpreted as a prohibitive), nor if sentences that do not have the form of prototypical conditionals nonetheless receive a conditional interpretation (cf. the parallel interpretations in English of If he came late, he was punished and Whenever he came late he was punished).

INTRODUCTION

The literature on natural language conditionals, including many of the contributions to this volume, has contributed much to our understanding of the internal structure of conditional sentences and of their ‘meanings’. What has been less well discussed is the discourse function of conditionals. Two grammars of English are exceptions: Modern English by Marcella Frank (1972) and The grammar book by Marianne Celce-Murcia and Diane Larsen-Freeman (1983), both of which begin to characterize conditionals with reference to their patterns of occurrence in discourse. Mead and Henderson (1983) also provide an enlightening discussion of conditionals in a particular context, looking at how they function in an economics textbook. Winter (1982) discusses some of the general factors involved in the positioning of various adverbial types, including conditionals. Linde looks at some of the factors which play a role in the positioning of if-clauses either before or after a main clause. Her basic finding is that, with the exception of certain irrealis if-clauses, the order of clauses does not ‘reverse the order of events in real time’ (Linde 1976: 280).