Let us return to synchronic linguistics. By the end of the 1930s, the description of sound systems had a theoretical foundation in the work of Trubetzkoy and others, and its method was increasingly codified. But the phonology of a language is again just one part of the total ‘language structure’. How do phonemes relate to, in particular, units of meaning? By what methods can these in turn be identified securely? What kinds of relation does each unit of meaning bear, in turn, to other units of meaning?

These were technical questions, and the elaboration of techniques of description, which is the most striking feature of linguistics in the decades that follow the Second World War, was not at first accompanied by new general or philosophical ideas. For the most part European linguists tended to found their work on those of Saussure, and linguists in America on Bloomfield's. The most exciting problems were of method, and those that concerned the detailed structure of a language system. For Trubetzkoy the ‘language structure’ had already been not one system but ‘several partial systems’ (‘mehrer Teilsysteme’) (Trubetzkoy, 1939: 6). These components ‘hold together, complement each other, and stand in mutual relations’ (‘so daß alle Teile einander zusammenhalten, einander ergänzen, sich aufeinander beziehen’). The main task for his successors was to say exactly how they do so, and exactly what the components are. It was in America especially that answers were given, and much of what was worked out in this period has found its way into textbooks and been taught for decades to generations of students.

The term ‘semantics’ was coined in the 1880s, by the French linguist Michel Bréal. As ‘phonétique’ was in its widest sense a science of the sounds of speech, so ‘sémantique’, from the Greek verb for ‘to mean’, was a ‘science des significations’ (Bréal, 1911 [1897]: 8). How Bréal saw this discipline need not concern us. But for a structuralist a language is a system in which determinate forms are related to meanings that are also determinate. ‘Structural semantics’ is accordingly concerned with meanings as terms in such systems.

Its beginnings lie in Saussure's theory of ‘linguistic signs’ (2.1). Each sign, as we have seen, had two sides; and just as on the level of expression, to use Hjelmslev's terminology, units were identified in abstraction from the sounds that speakers physically utter, so those on the content level had to be abstracted from the passing meanings that were intended. This second task was therefore, in principle, as important as the first. But when Hjelmslev wrote he was himself among the few who had approached it seriously. We have seen how, by his test of commutation, both expressions and their contents were divided into ‘figurae’: ‘[ram]’, for example, into ‘[r]’ plus ‘[a]’ plus ‘[m]’; ‘ram’, in parallel, into ‘he’ plus ‘sheep’ (5.1). Another parallel, as we will soon see, was between semantic oppositions and distinctive phonetic features in Prague School phonology (3.2).

From the 1940s onwards structural linguists were increasingly distinct from linguists in general. Structuralist schools were not established equally in all countries: scarcely at all, for example, in Italy or in Germany. But in the United States especially to be ‘a linguist’ was increasingly to be a structuralist. Above all, it was to be a ‘descriptive’ linguist. The term ‘descriptive’ has many resonances, and can easily be misunderstood. But for Americans its sense was like that of Saussure's ‘synchronic’. In a leading introduction to linguistics, ‘descriptive or synchronic linguistics’ was the study of ‘how a language works at a given time, regardless of its past history or future destiny’ (Hockett, 1958: 303). It deals with ‘the design of the language of some community at a given time’ (321).

But Saussurean linguistics also had a ‘diachronic’ or historical branch. How was that affected?

When the Cours appeared this branch was dominant. Saussure himself had been young in Leipzig in the 1870s, when the so-called ‘Junggrammatiker’ or ‘neogrammarians’ formulated basic principles of change in language. It was through their work above all that, we are told, ‘linguistics in the true sense’ (‘la linguistique proprement dite’) had been born (Saussure, 1972 [1916]: 18f.). It is not surprising therefore that the chapters of the Cours devoted to its diachronic side (Part 3, 193ff.) say so little that was new. Of the structuralists who followed, Bloomfield had been trained in the tradition of the neogrammarians, and his masterly account of language change, which forms most of the latter half of Language, was based firmly on their principles.

What is ‘structural linguistics’? Do most linguists still accept its principles? Or are they now believed in only by old men, clinging to the ideas that were exciting in their youth? Who, among the scholars who have written on language in the twentieth century, was or is a structuralist? Who, by implication, would that exclude?

It may seem, at the outset, that the first of these questions should be fundamental. We must begin by asking what, in general, we mean by ‘structuralism’. There are or have been ‘structuralists’ in, for example, anthropology; also in other disciplines besides linguistics, such as literary criticism and psychology. What unites them, and distinguishes them from other theorists or practitioners in their fields? In answering this question we will identify a set of general principles that structuralists subscribe to; and, when we have done that, we will be able to ask how they apply to the study of language. From that we will deduce the tenets that a ‘structural linguist’ should hold; we can then see who does or, once upon a time, did hold them. But an inquiry in this form will lead us only into doubt and confusion. For different authorities have defined ‘structuralism’, both in general and in specific application to linguistics, in what are at first sight very different ways. There are also linguists who are structuralists by many of the definitions that have been proposed, but who would themselves most vigorously deny that they are anything of the kind.

Let us look, for a start, at the definitions to be found in general dictionaries. For ‘structuralism’ in general they will often distinguish at least two different senses.

The chief, if not the only, interest of the Cours de linguistique générale was as an account of the foundations of linguistics. Of its five numbered parts, the last three deal conventionally with sound change and changes in grammar, the distribution in space of languages and dialects, the reconstruction of prehistoric languages, and other topics natural in a manual of its day. These together form more than a third of the whole (193–317). The chapters everyone now cites are part of the introduction, most of Part 1 (‘General Principles’) and most again of Part 2 (‘Synchronic linguistics’), a third again. But the earliest reviewers, as Keith Percival has shown, did not see in these the revolution that was later proclaimed. The book was seen more as old-fashioned (Percival, 1981).

That is perhaps not so surprising. For work on the foundations of a discipline need not have immediate repercussions on the way it is practised. When the Cours appeared, most linguists worked on Indo- European or some other family, on the history or grammar of particular languages, in dialectology, and generally in fields to which it offered nothing new. Even a ‘synchronic’ linguist could learn little. The treatment of speech-sounds, for example, was based on lectures given in 1897 (editors' note, 63) which were already dated. For the rest, we might be tempted to recall a remark by Delbrück cited in an earlier section (2.3). Provided that their methods are not disturbed, practising linguists can live happily with whatever any theorist says about the philosophical principles that underlie what they are doing. Only other theorists, of whom there are at any time few, need respond.

Is structural linguistics still a living movement, and if not when did it die? The answers will depend on what is meant by ‘structuralism’; and, as we found at the outset, that is not straightforward. But for many commentators its creative phase has long been over. Its heyday lasted from the 1930s, when it was named, to the end of the 1950s; and, throughout that period, linguistics was dominated by it. But structuralism in America is said to have been overturned by Chomsky, and by the 1970s his hegemony was world-wide. Other programmes have emerged since, some opposed to and some simply independent of his. But few have reasserted views that he rejected. Therefore structuralism, as an active source of ideas, is indeed dead.

This account should not be seen as an Aunt Sally. Most linguists do not now describe themselves as structuralists, even if they work in fields like phonology, or with units like the morpheme, that the movement defined. When they refer to ‘structuralist approaches’ to their subject, they will mean ones that have been superseded since the end of the 1950s. To follow these overtly would not, therefore, be a good career move. But if linguistics is no longer officially structuralist, many linguists are still strongly influenced by structuralist ideas. These concern, in particular, the notion of a language as a system and the autonomy of linguistics as defined by it.

The structuralist bible is, by long consent, Saussure's Cours.

Linguistics is said in dictionaries to be ‘the branch of knowledge that deals with language’ (New Shorter Oxford English Dictionary) or ‘the scientific study of language’ (Collins). But for structuralists it has been much more the study of, in the plural, languages. This was true at the outset, for Saussure, and is still true for many as we enter the twenty-first century. What then constitutes ‘a language’? It is easy to give examples: English is one, Japanese another, and so on. But what, in general, are they?

Let us look again at dictionaries. For the first editor of The Oxford English Dictionary (Murray et al., 1933 [1884–1928]), the earliest sense of ‘language’ (§1) was that of ‘the whole body of words and of methods of combination of words used by a nation, people, or race’; alternatively, ‘a tongue’. The dictionary itself was thus an account of the ‘whole body’ of words that constitute the lexicon of English. The second definition (§2) adds a ‘generalized sense’: ‘words and the methods of combining them for the expression of thought’. But where Murray saw a ‘body’, The New Shorter Oxford Dictionary speaks of a ‘system’. Language is ‘a system of human communication using words … and particular ways of combining them’; it is ‘any such system’, the definition adds, ‘employed by a community, a nation, etc.’ (§1a). In the Collins dictionary, it is ‘a system for the expression of thoughts, feelings, etc. by the use of spoken sounds or conventional symbols’ (§1); also in general (§2) ‘the faculty for the use of such systems’.

It was noted in 1.2.1 that, in many languages, particularly the Native American languages and those of Papua New Guinea, mood is described in terms of the grammatical markers of realis and irrealis. Although the distinction is basically the same as that between indicative and subjunctive, both being markers of the typological categories Realis and Irrealis, there are sufficient differences to warrant discussion of them in a separate chapter. Yet there is not always a clear distinction between them – see 7.1 for discussion.

Joint and non-joint marking

There are basically two ways in which realis and irrealis markers function. In some languages their main function is to co-occur with other grammatical categories. In others they mainly occur in isolation and are themselves the only markers of specific notional categories.

For instance, in Amele (Papuan – Roberts 1994: 372) an irrealis marker is required whenever a future marker is present in the sentence:

By contrast in another Papuan language, Muyuw (Bugenhagen 1994: 18, quoting a personal communication), the irrealis marker is itself the indication of future:

Strictly, these might seem not to be comparable in that, in the Amele example, there is a syntactic relationship between two grammatical markers, irrealis and future, whereas in the Muyuw example there is simply a specific grammatical marker, irrealis, that indicates futurity.

The modal systems described in the last two chapters share a number of features, not only in the systems themselves, but also, for many languages, in the use of modal verbs and the association with possibility and necessity. A detailed discussion of these has been left to this chapter in order to avoid too much repetition and cross-referencing. These issues hardly affect evidential, so that the discussion is almost entirely concerned with epistemic, deontic and dynamic modality.

Modal systems

Formal identity of different systems

Notionally, epistemic modality and deontic/dynamic modality might seem to have little in common. As suggested in 1.2.2, epistemic modality is concerned solely with the speaker's attitude to the truth value or factual status of the proposition (propositional modality), whereas deontic and dynamic modality refer to events that are not actualized, events that have not taken place but are merely potential (event modality). Yet in English (and many other languages) the same forms are used for both types. The following, for instance, can all be interpreted either epistemically or deontically:

Where the same verbs are used for the different types of modality, there are often slight differences in the forms, which will be summarized in 4.2.3. Such differences suggest that the types are grammatically as well as notion-ally different, but do not explain why the forms are basically the same.

This chapter will deal with a number of issues relating to both subjunctive and irrealis that have not been addressed, or not fully addressed, in the preceding two.

Similarities and differences

The decision to use different terminology for indicative/subjunctive and to deal with them in separate chapters rests upon a number of considerations.

It is partly a result of different traditions. The terms ‘indicative’ and ‘subjunctive’ are the traditional terms used in the description of the classical and modern languages of Europe, though, not unreasonably, they have been used by many scholars for other languages. According to Bybee et al. (1994: 236) the earliest use of the terms ‘realis’ and ‘irrealis’ noted in an extensive corpus is to be found in the description of the Australian language Maung in Capell and Hinch (1970). These terms have been preferred by scholars working in the Native American languages and in the languages of the Pacific, particularly those of Papua New Guinea. However, since it is unlikely that the scholars working in these languages were unaware of the classical tradition, it is probable that this decision was influenced by perceived differences in the nature of the data being investigated.

One possible difference that might be suggested is that mood in the European languages is a morphosyntactic category closely integrated with person, number, tense and voice. The four categories are not independently marked, the form being simultaneously the marker of all the grammatical categories, as well as of the lexical item.

Since the publication of the first edition (Palmer 1986) there has been considerable interest in modality (as well as in grammatical typology in general). A symposium on Mood and Modality held in the University of New Mexico in 1992 was successful in bringing together over forty researchers, and resulted in the publication of eighteen papers (Bybee and Fleischmann 1995a). Yet, in contrast, a workshop on modality at the International Congress of Linguists only ten years before had attracted only four scholars. This symposium was followed in 1993 by a symposium on Modality in Germanic languages, which resulted in a further eight papers (Swan and Westvik 1997). Indeed, Bybee and Fleischmann (1995b: 1) suggest that the first symposium succeeded in establishing for modality the kind of status that had been established for tense and aspect by a symposium on those subjects ten years before (Hopper 1982). It should, however, should be pointed out that the first edition of the present volume was published six years before the symposium in New Mexico and that a volume on modality in English appeared thirteen years before that (Palmer 1979).

It was recognized in the first volume that the most appropriate name for the relevant category is simply ‘modality’, and that ‘mood’ is more appropriate as the traditional name for indicative, subjunctive, etc., in both classical and modern European languages.

It has come to be recognized in recent years that modality is a valid cross-language grammatical category that can be the subject of a typological study. It is a category that is closely associated with tense and aspect in that all three categories are categories of the clause and are generally, but not always, marked within the verbal complex.

In notional terms all three are, in some way, concerned with the event or situation that is reported by the utterance (though for simplicity the term ‘event’ will, throughout this volume, be used to cover events, actions, situations, states, etc.). Tense, rather obviously, is concerned with the time of the event, while aspect is concerned with the nature of the event, particularly in terms of its ‘internal temporal constituency’ (Comrie 1976: 3). Modality is concerned with the status of the proposition that describes the event.

Basic concepts

Realis and irrealis

Modality differs from tense and aspect in that it does not refer directly to any characteristic of the event, but simply to the status of the proposition. One possible approach to its analysis is to make a binary distinction between ‘non-modal’ and ‘modal’ or ‘declarative’ and ‘non-declarative’, and to associate this distinction with the notional contrast of ‘factual’ and ‘non-factual’, or ‘real’ and ‘unreal’.

However, these terms are not really satisfactory (see 1.1.2), and in recent years ‘realis’ and ‘irrealis’ have been used for this distinction.