2 Trier and the ‘mosaic’ model of a lexical field

90



LANGUAGE, MEANING AND SENSE



According to Trier, a change occurred in the structure of the lexical field associated with this conceptual area during the

course of the thirteenth century. At the beginning of the century the lexical field was structured as follows (after Lehrer 1974:

15–16):

Wîsheit

Kunst



List



Kunst referred to ‘courtly knowledge, including social behaviour’ (Ullmann 1957:166); List referred to technical

knowledge and skill; Wîsheit was a general term, that is to say a superordinate, of which the other two were hyponyms. By the

end of the century, the structure of the field had changed:

Wîsheit



Kunst



Wizzen



List moved out of the conceptual field altogether: Wîsheit came to be used for religious, mystical knowledge only; Kunst

lost its association with courtly knowledge and came to denote special technical skill or knowledge, not unlike the former

meaning of List; at the same time, a new term, Wizzen, came into the field to designate the most ordinary, everyday type of

knowledge.

Many criticisms were made of Trier’s assumptions and methodology (none of which, however, diminish the importance of

his influence on lexical studies). For instance, Trier’s contention that the whole vocabulary of a language was structured as a

single global field has not proved tenable. Likewise, his claim that the members of a field form a tightly-structured set leaving

no gaps has not been possible to maintain (see Lehrer 1974:97–102). Other problems have been the precise delimitation of

conceptual areas, and the identification of the ‘same’ conceptual area at different periods of time— both essential to Trier’s

method. Trier has also been criticised for concentrating exclusively on paradigmatic relations; an alternative version of fieldtheory based on syntagmatic relations was advanced by Porzig (1950).

Recently, the appropriateness of the ‘mosaic’ metaphor, with its emphasis on the boundaries between concepts, has been

called into question. Work by Berlin and Kay (1969) on colour vocabularies suggests that boundaries between meanings are

less important than the locations of ‘core’ meanings. In particular, they found that while different languages located the

boundaries between colours in unpredictable and idiosyncratic places, there was much greater agreement as to the location of

the ‘best’ red, blues, greens, and so on.

8.

SEMANTIC FEATURES

8.1

Some history

The basic idea of lexical decomposition—explaining the meanings of words in terms of simpler units of meaning (semantic

features, or components)—is not, in essence, a difficult one. A lexicographer who gives the meaning of stallion as ‘male

horse’ is doing just this. Hardly controversial, one might think. Yet there is still very little in the way of a consensus amongst

linguists as to how far and in what way lexical decomposition should be carried out. The difficulties arise when we try to be

precise about these simpler units of meaning: What is their nature? How do we detect their presence? How are they related to

one another? Is there a finite number of them? Are they universal or language-specific? How much of the vocabulary can be

decomposed?

Probably the first thorough-going attempt at a systematic componential approach to meaning was that of the philosopher

Leibniz (see discussion in Wierzbicka 1980:1–37). What he hoped to do was express complex meanings in terms of ever simpler

meaning elements until he arrived at an inventory of unanalysable primitive semantic units. These, he thought, would

constitute the ‘alphabet of human thought’.

Leibniz’s pioneering work was not taken up by any followers and it was not until the present century that interest in the

problem revived. One of the first to pick up the trail was the Danish linguist Hjelmslev (1953). He started out from the Saussurean

picture of the linguistic sign as an arbitrary association between a meaning and a phonetic form. He reasoned that since the

phonetic form of the word could be analysed into simpler units (phonemes and, ultimately, distinctive features), the structure

of language being essentially symmetrical (an article of faith with Hjelmslev), it must likewise be possible to analyse the

meaning aspect of signs. Ultimately Hjelmslev envisaged an inventory of ‘content figurae’ (elements of meaning) in terms of

which the meanings of all vocabulary items in the language could be expressed. The analysis of word meanings into content

figurae would be a ‘reduction’, like the analysis of word-forms into phonemes, in the sense that the inventory of elements in



AN ENCYCLOPAEDIA OF LANGUAGE



91



terms of which an analysis could be expressed would be much smaller in number than the set of items being analysed. Unlike

Leibniz, Hjelmslev considered that any inventory of content figurae would be specific to a particular language.

Independently, in America, a componential approach to meaning was developing by certain anthropologists with a strong

interest in kinship terminologies (see Leech 1974, chapter 11, for examples and discussion). A later more sophisticated

version of componential theory was put forward by Katz and Fodor (1963) as part of a comprehensive theory of language

structure, the syntactic part of which was being developed by Chomsky and his followers. Most recently, Wierzbicka, taking

inspiration directly from Leibniz, has suggested a radically new version of lexical decomposition, using only thirteen

primitive notions (1980).

8.2

Two examples of semantic feature analysis

In 1963 Katz and Fodor published ‘The structure of a semantic theory’, an article in which they attempted to sketch the

outlines of the semantic component of a generative grammar. This was the part of the grammar that would assign a semantic

interpretation to a string of words on the basis of its lexical content and its syntactic structure. The interpretation would

indicate various semantic properties of the string, such as how many meanings it had, what its semantic relations (such as

entailment) were with other strings, whether it was semantically anomalous, and so on. The semantic component envisaged by

Katz and Fodor comprised a dictionary, in which the meaning of each lexical item was specified in a uniform way, and a set of

rules for combining the separate word meanings, according to the syntactic structure, to give a global meaning. The following

was offered as an example of a dictionary entry (slightly modified):

The items in parentheses indicate ‘semantic markers’. These are features which have systematic patterning in the language;

they are used in specifying the meanings of other words, and are therefore involved in semantic relations between words. For

instance, it should be possible to gauge the closeness of two meanings by the number of semantic markers that they have in

common. The items within square brackets are ‘semantic distinguishes’. These take care of any idiosyncratic aspects of the

meaning of a word—the residue not accounted for by the markers. This analysis is designed to give specific kinds of

information about the meaning of a word. For instance, the number of branch-endings gives the number of ways in which the

word is ambiguous. The analysis of each sense can be read off as the items traversed in going to the end of the branch, so the

analysis of the most likely meaning of bachelor in John is still a bachelor is: (Human) (Male) [who has never married]. As

we have noted, the degree of resemblance between senses can also be read off. (There is also provision in the system—but not

illustrated in our example—for indicating selectional restrictions by means of markers.)

The second illustrative example of componential analysis is in the Continental tradition of structural semantics: Table 5 is

adapted from Baldinger (1980:67) (which in turn derives from Pottier). It presents a feature analysis of various types of

furniture designed for sitting on:

Table 5

with a back

chaise/chair

fauteuil/armchair

tabouret/stool

canapé/sofa

pouf/pouffe



raised above ground



for one person



to sit on



with arms



of solid material



+

+

−

+

−



+

+

+

+

+



+

+

+

−

+



+

+

+

+

+



−

+

−

+

−



+

+

+

+

−



A plus sign in the table indicates the presence of a feature; a minus sign indicates its absence. Obviously the features which

are possessed by all the items in the set cannot be used to distinguish members of the set from one another; they will serve,

however, in an analysis of a wider range of items. For instance, the feature [to sit on] distinguishes members of this set from other



92



LANGUAGE, MEANING AND SENSE



types of furniture such as tables and beds. This analysis differs from the Katz and Fodor example in two important ways. First,

it is not hierarchically organised; there are reasons for believing that insistence on hierarchical organisation makes things

unnecessarily difficult. Second, there is nothing corresponding to Katz and Fodor’s distinction between markers and

distinguishes; the features here are more like markers, except that they are intended to give a full analysis, without residue, of

the meanings of the lexical items.

8.3

Some issues in componential theory

This sort of analysis has an obvious intuitive appeal. Indeed, it seems likely that some kind of feature analysis is inescapable

for any adequate semantics, whether theoretical or descriptive; it is certainly hard to see how the meanings of words could be

specified without recourse to something of the sort. However, all concrete proposals to date have met with strong criticism,

and there does not seem to be much agreement as to the best way forward. Obviously a full discussion of these complex

matters is not possible here; but it is perhaps worth indicating briefly some of the more salient issues.

There are a number of related questions concerning the proper scope of componential analysis. Is it the case, for instance,

that all or most lexical meanings should be considered complex and therefore susceptible of analysis into features? Or are a

significant proportion of word meanings atomic in the sense that they cannot be broken down any further? Sampson (1979)

believes that all word meanings are atomic, and thus rejects the whole basis of componential analysis. Wierzbicka (1980)

treats all words denoting natural species, such as cat, dog, crab, robin, oak, tulip, as unanalysable. On the other hand, a Katz/

Fodor-type analysis would presumably assign markers like (Animate) and (Insect), (Bird), (Tree), or whatever, to such items.

Another question concerns how much of a word’s meaning is to be accounted for by a feature analysis. Katz and Fodor, it

will be remembered, allowed for an unanalysed residue. Their distinction between markers and distinguishes was severely

criticised, but there is perhaps some advantage in allowing for a degree of indeterminacy in word meaning, and it is not

obvious how to do this within a componential framework. A feature-analysis which aims to account for the whole of a word’s

meaning runs the risk of an infinite proliferation of features. Although few componentialists nowadays share Hjelmslev’s

desire for a reductive analysis (Wierzbicka is an exception), most would balk at the idea of an infinite inventory of features.

Probably most componentialists (certainly those who are looking for the ‘alphabet of human thought’) would agree that

semantic features are not to be seen merely as a convenient descriptive device, but correspond to something real in the minds

of language users; they would also agree that we should not expect each language to yield a unique set of features. None the

less, both topics—psychological reality and universality—have given rise to a great deal of discussion (see, for example,

Leech 1974:231–62; Lyons 1977:331–3; Baldinger 1980:93–109).

It is common practice to use ordinary words like human, male and cause to label semantic features, and an analysis often

looks like a translation into some sort of Pidgin (scornfully referred to as ‘Markerese’ by critics); so kill, for instance, might

be rendered as (Cause) (Become) (Not) (Alive). Such an analysis raises some serious theoretical problems. For instance, what

exactly is the relationship between the word cause, say, and the semantic feature (Cause)? Are semantic feature synonymous

with the words which label them? If they are, then one must somehow explain why it is that John killed Bill on Thursday is

not exactly paraphrased by John caused Bill to become not alive on Thursday (if John had given Bill a slow-acting poison

some time earlier the latter sentence, but not the former, would be applicable). One possible way of circumventing this

problem is to say that semantic features are abstract entities, not necessarily synonymous with any concrete lexical items, and

that in the case in question (Cause) has whatever meaning is necessary to account for the difference between the two

sentences. But this only raises a further series of problems. If semantic features are purely abstract, how do we find out what

features there are? How do we discover their meanings (do they indeed have the same kind of meaning as words)? How do we

know whether they are present or not in particular cases? In short, how can we ensure that the facts of language exert strong

enough constraints on the imaginations of linguists? At least one linguist committed to a componential approach, Wierzbicka,

maintains that only features whose meanings are expressed in ordinary language can be brought under proper empirical

control.

Problems of a different kind are posed by some recent work in cognitive psychology. In classical componential theory the

meaning of a word is given by a list of features all of which are necessary and which collectively are all that is necessary. The

semantic relation of hyponymy is represented by the inclusion of all the features defining the superordinate in the featurespecification of the hyponym. Thus if a lexical item X is defined by the set of features (A) (B) (C), and a second lexical item

Y is defined by the set (A) (B), then X will be a hyponym of Y, and the class of Xs will be a sub-category of the class of Ys.

This model of meaning would lead us to expect that all members of a category would have the same membership status, since

they all must possess the same set of qualifying features; or, to put it in semantic terms, all immediate hyponyms of a

superordinate would have the same relationship with the superordinate term. Take the category of birds: the classical

componentialist picture would imply that ostriches, kiwis, turkeys, blackbirds and humming-birds were all on a par, since they

all possess the defining characteristics of birds (whatever these are). However, psychological research has shown that not all



AN ENCYCLOPAEDIA OF LANGUAGE



93



sub-categories of a superordinate category have the same status within the category: some are judged by informants to be

‘better’ examples of their category than others. For instance, blackbirds are judged to be better examples of birds than are

ostriches. The correlated semantic observation is that some hyponyms are judged—on various measures of semantic distance

—to be closer in meaning to their super-ordinates than others. The best examples of a category are called the ‘prototypical’

examples: componential theory has no obvious explanation of prototypicality. (For a thorough discussion of the semantic

significance of psychological research on prototypicality see Pulman 1983.)

EPILOGUE

This survey of descriptive and theoretical issues in the study of meaning has necessarily been extremely cursory. It is hoped,

however, that it has succeeded in giving some idea of the scope, inherent interest and possibilities of semantic—especially lexical

—studies. Although semantics was for many years the Cinderella among the core areas of language study, this is much less

true today, and it is certain that we can look forward to considerable advances in understanding in the near future.

REFERENCES

Alston, W.P. (1964) Philosophy of Language, Prentice-Hall, Englewood Cliffs, NJ.

Baldinger, K. (1980) Semantic Theory: Towards a Modern Semantics, translated by W.C. Brown and edited by R.Wright, Blackwell,

Oxford.

Berlin, B. and Kay, P. (1969) Basic Color Terms: Their Universality and Evolution, University of California Press, Berkeley.

Bloomfield, L. (1935) Language, Allen & Unwin, London.

*Cruse, D.A. (1986) Lexical Semantics, Cambridge University Press, Cambridge.

Geckeler, H. (1971) Strukturelle Semantik und Wortfeldtheorie, Fink, München.

Hjelmslev, L. (1953) Prolegomena to a Theory of Language, translated from the Danish by F.J.Whitfield, Indiana University, Bloomington,

Ind.

Katz, J.J. and Fodor, J.A. (1963) ‘The Structure of a Semantic Theory’, Language, 39: 170–210. Reprinted in Fodor, J.A. and Katz, J.J.

(eds) (1964) The Structure of Language: Readings in the Philosophy of Language, Prentice-Hall, Englewood Cliffs, NJ: 479–518.

*Leech, G.N. (1974) Semantics, Penguin, Harmondsworth.

Lehrer, A.J. (1974) Semantic Fields and Lexical Structure, North Holland, Amsterdam.

Lyons, J. (1963) Structural Semantics, Blackwell, Oxford.

* Lyons, J. (1977) Semantics (2 vols), Cambridge University Press, Cambridge.

Morris, C.W. (1946) Signs, Language and Behaviour, Prentice-Hall, Englewood Cliffs, NJ.

Ogden, C.K. and Richards, I.A. (1923) The Meaning of Meaning, Routledge & Kegan Paul, London.

Palmer, F.R. (1976) Semantics: A new Introduction, Cambridge University Press, Cambridge.

Porzig, W. (1950) Das Wunder der Sprache, Franke, Berne.

*Pulman, S.G. (1983) Word Meaning and Belief, Croom Helm, London.

Sampson, G. (1979) ‘The Indivisibility of Words’, Journal of Linguistics, 15:39–47.

*Ullmann, S. (1957) The Principles of Semantics, 2nd edn, Blackwell, Oxford.

Ullmann, S. (1962) Semantics, Blackwell, Oxford.

Wierzbicka, A. (1980) Lingua Mentalis: The Semantics of Natural Language, Academic Press, Sydney.



FURTHER READING

The items marked with an asterisk in the list above will be found especially helpful for readers who wish to pursue these

topics in greater depth.



6

LANGUAGE, MEANING AND CONTEXT: PRAGMATICS

GEOFFREY LEECH AND JENNY THOMAS



Compared with the branches of linguistics dealt with in earlier chapters, pragmatics has only recently come on to the

linguistic map. Some may doubt, in fact, whether it has become a respectable branch of linguistics, or even if there is any

legitimate field of study called ‘pragmatics’. It nevertheless became a significant factor in linguistic thinking in the 1970s, and

since then has developed as an important field of research. In this chapter, we trace the main trends in the modern

development of pragmatics, from its beginning as a ‘fringe subject’ on the borders of philosophy and linguistics, to its present

broad concern with linguistic communication in its social and cultural context. In the English-speaking world, pragmatics is

more narrowly defined in its concerns than in continental Europe. We begin with the ‘Anglo-American’ tradition of

pragmatics, and later make reference to a somewhat broader and distinct European tradition.

1.

PRAGMATICS WITHIN THE ANGLO-AMERICAN TRADITION

We may roughly describe pragmatics as the study of the meaning of linguistic utterances for their users and interpreters. It is

important, as far as its origin is concerned, to see pragmatics as part of a triad of studies distinguished by the American

philosopher Charles Morris (1938), and later used by logicians such as Rudolf Carnap (1942, 1955). Pragmatics, according to

this line of thought, is the study of signs (and sign systems) in relation to their users; whereas SEMANTICS is the study of

signs in relation to their designata (what they refer to), and SYNTAX is the study of signs or expressions in relation to one

another. The three fields are subdivisions of SEMIOTICS, the study of signs and sign systems, and may therefore be just as

fittingly applied to the study (say) of artificial signs such as traffic lights, or of signs used in animal communication, as to

human language. But, in practice, work in pragmatics has principally been carried out on human language, or ‘natural

language’ as logicians are accustomed to call it.

Before we turn from non-human communication, however, let us use it to give a simplified illustration of the difference

between syntax, semantics and pragmatics as components of semiotics. Our example, from the famous studies by von Frisch,

is the ‘linguistic’ dance by which honey bees indicate the position of a source of pollen and nectar to other bees. The actual

structure of the dance message—the sequence of movements—is a matter of syntax. The referential content of the dance—

signalling the direction, distance, and so on, of the food source—is a matter of semantics. The way the dance functions as an

aspect of bee behaviour—as a means of summoning bees to a food source—is a matter of pragmatics.

For philosophers and logicians like Carnap, pragmatics was the Cinderella of the three studies. It was important, in the

empiricist and positivist thinking of the mid-twentieth century, to formalise precisely the relation between symbols and what

they represented, so that conditions of truth and falsehood could be established or verified. For this enterprise, both syntax and

semantics were important. But pragmatics introduced a messy, unformalisable element—the attitudes, behaviour and beliefs of

symbol-users. Hence, for many philosophers and linguists, pragmatics became a convenient waste-tip where one could dump

(and frequently forget about) aspects of language and communication which were awkward, because they did not fit a nice

neat world of syntactically well-formed sentences and their semantically well-behaved truth conditions.

As the previous paragraphs suggest, pragmatics was born out of the abstractions of philosophy rather than of the descriptive

needs of linguistics (and this, it will be argued below, accounts in part for the difficulties which were later experienced by

linguists when they tried to apply pragmatic models to the analysis of stretches of naturally-occurring discourse). Even when

pragmatics started to become important for linguistics, it was again, at least in the English-speaking world, informed by the

work of philosophers. Three philosophers in particular—J.L.Austin, J.R.Searle and H.P.Grice—must be mentioned as the

inspirational sources of linguistic pragmatics as it developed in the 1970s, although strangely enough, none of them adopted

the term ‘pragmatics’ in their own work.

These three philosophers were closely associated with the Oxford philosophical tradition (although Searle and Grice ended

up teaching in California). They all belonged to the ‘ordinary language’ school of philosophy, rather than the ‘formal

language’ school represented by Carnap: this is, they were interested in the way natural human language conveys meaning, as



AN ENCYCLOPAEDIA OF LANGUAGE



95



a way of understanding the nature of thought, logic and communication. We may indeed call them ‘philosophers of

communication’: for the term communication, associating language with its use to convey messages by users for interpreters,

is at the heart of their work, and is at the heart of pragmatics. Yet, it is important to understand that in their own thinking,

these philosophers were heavily influenced by the orthodox truth-based approach to meaning. In fact, their contribution was

precisely in finding out and exploring problems, limitations and failures of the truth-based paradigm.

1.1

J.L.Austin

Austin’s seminal book (a posthumous reworking of the notes he had prepared for a series of lectures delivered at Oxford and

Harvard) was How to Do things with Words (1962). The emphasis which this title places on ‘doing things’ is shown in his

exploration of PERFORMATIVE utterances, such as:

I resign.

I name this ship Boniface.

I bet you £1,000 you will be acquitted.

I hereby give you notice of dismissal.

Such utterances are problematic because, although they have all the outward signs of being declarative sentences, or

statements, they do not appear to have a truth value. They thus seem to lack what is normally treated as a necessary property

of statements. Austin denied that such sentences could be true or false in arguing that their nature was PERFORMATIVE

rather than CONSTATIVE, in that their meaning was to be identified with the performance of an action. In saying ‘I resign’, a

person does in fact resign; in saying ‘I hereby give notice’, a person actually performs the action of giving notice, etc.

Conditions of the kind required to elucidate their meaning are not truth conditions, but rather ‘felicity conditions’, as Austin

called them—or conditions of appropriateness.

In addition to their declarative form, performatives generally have well-recognised syntactic characteristics, such as a verb

(of a particular kind) in the present tense, a first-person subject and the possibility of adding the adverb hereby. But Austin’s

investigation of them led him to conclude that not only performatives, but all utterances, partake of the nature of actions.

Hence one could bring out the action-like qualities of a statement, a question, a request, etc. by prefixing to it an implicit

performative:

You’ll definitely get paid tomorrow.

(i.e. I promise you that you’ll definitely get paid tomorrow.)

What’s the time?

(i.e. I ask you to tell me what the time is.)

Be quiet when I’m talking.

(i.e. I order you to be quiet when I’m talking).

Even a colourless statement of fact such as Beavers build dams arguably has an implied performative such as ‘I state that…’.

The approach to meaning which says that linguistic phenomena are basically actions or deeds has an advantage over truthbased approaches, in that it invites us to go beyond the traditional logician’s limited concern with declarative or

PROPOSITIONAL meaning. Austin took this idea further, in claiming that the same utterance could at the same time constitute

three kinds of act:

(1) a LOCUTIONARY ACT (or locution): the act of uttering some expression with a particular sense and reference.

(E.g. He said to me ‘Shoot her!’ meaning by ‘shoot’ shoot and by ‘her’ her.)

(2) an ILLOCUTIONARY ACT (or illocution): the act performed in, or by virtue of, the performance of the locution,

(E.g. he urged, or requested, or invited me to shoot her.) —such that we may say that what was said had the FORCE

of that illocution (e.g. of a request, or an invitation).

(3) a PERLOCUTIONARY ACT (or perlocution); the act performed by means of what is said:

(E.g. He persuaded me to shoot her).

Austin focused on the second of these. The locution (1) belongs to the traditional territory of truth-based semantics. The

perlocution (3) belongs strictly beyond the investigation of language and meaning, since it deals with the effect, or result, of

an utterance: whether my words persuade someone to lend me £10 depends on factors (psychological, social, or physical

factors) beyond my control, and is only partly a matter of what I said. The illocution (2) occupies the middle ground between



96



LANGUAGE, MEANING AND CONTEXT



them: the ground now considered the territory of pragmatics, of meaning in context. The verbs used to describe illocutions—

such as claim, promise, beg, thank and declare—can generally be used as performative verbs, and by this connection, Austin

and his student Searle were able to think of the performative prefix (I beg you…) as an illocutionary force indicating device or

IFID (a term now widely used to include not only explicit linguistic indicators of illocutionary intent, but also paralinguistic

and non-linguistic indicators). In contrast, it is not possible to use a perlocutionary verb such as persuade or incite as a

performative: I (hereby) persuade/incite you to vote against the government is a nonsensical utterance, implying as it does

that the addressee is totally under the ‘thought-control’ of the speaker.

1.2

J.R.Searle and ‘Classical Speech-Art Theory’

Austin’s book was rambling, conversational, and (as books on philosophy go) entertaining. It was left to John Searle to bring

greater systematicity to the ideas which Austin had so perceptively explored. In another surprisingly readable book Speech

Acts: An Introduction to the Philosophy of Language (1969), Searle took even further than Austin the idea that meaning is a

kind of doing: he went so far as to claim that the study of language is just a sub-part of the theory of action. Although Searle’s

method was still the informal method of ordinary language philosophy, he crystallised the concepts of illocutionary act and

illocutionary force to the extent where one can reasonably speak of Searle’s ‘Speech-Act Theory’ as the ‘classical’ account

which subsequent work on speech acts treats as its point of departure. The interest of his work for pragmatics, like that of

Austin, centres around illocutionary acts and illocutionary force (understood as the functions or meanings associated with

illocutionary acts). Therefore, when we use the term ‘Speech Act Theory’ in pragmatics, we in practice refer to illocutionary

acts.

Searle offered definitions of various speech acts in terms of the conditions which were required to be present if a given

speech act was to be effectively performed. Comparable to Austin’s ‘felicity conditions’, these were described as four kinds

of rules:

(a) PROPOSITIONAL CONTENT rules specify the kind of meaning expressed by the propositional part of an utterance (for

example, a promise necessarily refers to some future act by the speaker).

(b) PREPARATORY rules specify conditions which are prerequisites to the performance of the speech act (e.g., according to

Searle, for an act of thanking, the speaker must be aware that the addressee has done something of benefit to the speaker).

(c) SINCERITY rules specify conditions which must obtain if the speech act is to be performed sincerely (e.g. for an

apology to be sincere, the speaker must be sorry for what has been done. Other analysts, however, have argued that I

apologise is a self-verifying performative, which is always felicitous, whether uttered sincerely or not).

(d) ESSENTIAL rules specify what the speech act must, conventionally, ‘count as’ (e.g. the essential rule for a warning is

that it counts as an undertaking that some future event is not in the addressee’s interest).

Searle argued that, on the basis of these four rule types, different speech acts can be easily distinguished (although more recent

work in pragmatics— see section 5 below—indicates that the establishing of discrete categories of speech acts is far more

problematical that Searle leads us to believe). Consider, for instance, the rules for requesting and advising:

[Note: These rules are taken from Searle (1969:66–7): S=speaker; H= ‘hearer’ or addressee; A=act; E=event.]

(I) REQUEST

Prepositional content: Future act A of H

Preparatory: 1. H is able to do A, S believes H is able to do A.

2. It is not obvious to both S and H that H will do A in the normal course of events of his own accord.

Sincerity:

S wants H to do A.

Essential:

Counts as an attempt to get H to do A.

(II) ADVICE

Prepositional content: Future act A of H.

Preparatory: 1. S has some reason to believe A will benefit H.

2. It is not obvious to both S and H that H will do A in the normal course of events.

Sincerity:

S believes A will benefit H.

Essential:

Counts as an undertaking to the effect that A is in H’s best interest.

To elucidate the differences between these definitions, Searle adds the comment:



AN ENCYCLOPAEDIA OF LANGUAGE



97



Advising you is not trying to get you to do something in the sense that requesting is. Advising is more like telling you what

is best for you.

The descriptive force of these rules or conditions, like that of Austin’s ‘felicity conditions’, can best be appreciated by finding

examples where they are violated. For example, a request would be ‘infelicitous’

(a) if it did not refer to a future act:

Could you please phone me by 5 o’clock last Tuesday?

(b) or if H were unable to do A:

Would you mind translating this letter into Swahili?

(spoken to someone who knows no Swahili).

(c) or if S did not want H to do A:

Please phone me at the office tomorrow.

(spoken by someone who does not want to be phoned, and indeed will not be at the office tomorrow; in this case the

request would be effectively performed, but would not be sincere)

(d) or if the utterance did not count as an attempt to get H to do A:

Would you kindly refrain from laughter?

(spoken by a TV comic in a situation where there was a clear intention to provoke laughter).

We can all agree that there would be ‘something odd’ about the quoted utterances if they were spoken under the conditions

described.

Speech-act theory lends itself to establishing systems of classification for illocutions. Austin proposed one such

classification, in an attempt to reduce to some kind of order the multitude of different speech acts—which he estimated to

number somewhere between 1,000 and 9,999. Searle improved on this classification, when he divided speech acts (Searle

1979) into five categories:

(i) ASSERTIVES commit S to the truth of some proposition (e.g. stating, claiming, reporting, announcing)

(ii) DIRECTIVES count as attempts to bring about some effect through the action of H (e.g. ordering, requesting, demanding,

begging)

(iii) COMMISSIVES commit the speaker to some future action (e.g. promising, offering, swearing to do something)

(iv) EXPRESSIVES count as the expression of some psychological state (e.g. thanking, apologising, congratulating)

(v) DECLARATIONS are speech acts whose ‘successful performance… brings about the correspondence between the

propositional content and reality’ (e.g. naming a ship, resigning, sentencing, dismissing, excommunicating, christening).

1.3

H.P.Grice: Logic and conversation

The third of our trio of philosophers, H.P.Grice, like Searle attempted to face up to the problem of how meaning in ordinary

human discourse differs from meaning in the precise but limited truth-conditional sense. Whereas Searle, however, proposed

subsuming the truth-based paradigm in an action-based one, Grice was interested in explaining the difference between what is

said and what is meant. ‘What is said’ is what the words mean at their face value, and can often be explained in truthconditional terms. ‘What is meant’ is the effect that the speaker intends to produce on the addressee by virtue of the

addressee’s recognition of this intention (see Grice 1957). There can often be a considerable gap between these two types of

message, one of which consists of only ‘explicit meaning’, while the other contains inexplicit meaning too. Consider an

exchange between two people as follows:

A: Where’s Janet?

B: Uh—she was walking in the direction of the post office five minutes ago.

B’s reply simply reports the behaviour of Janet five minutes before the conversation. But actually it conveys, by implication,

more than that: it implies that B thinks that, seeing that A wants to know where Janet is, the post office, or thereabouts, would

be a good place to look for her. If we ask how that implication is conveyed, the answer must take account of such matters as

‘general knowledge’ and ‘shared contextual knowledge’. Thus, the expression the post office implies that B expects A to share

knowledge of the location of a particular post office (presumably the nearest one to where they are standing). Moreover, the

implication that Janet may be at the post office now rests on common knowledge that the post office is the sort of place you might

be expected to walk to in a few minutes. (There would be no comparable implication if B had said: ‘…she was walking in the

direction of the setting sun …’!)



98



LANGUAGE, MEANING AND CONTEXT



But even assumptions like these do not explain the process of inferring such conversational meanings entirely. To give a

reasonable explanation, we have to assume that interactants in a conversation have regard to what Grice (1975) calls ‘the Cooperative Principle’ (CP):

Make your contribution such as is required, at the stage at which it occurs, by the accepted purpose of direction of the talk

exchange in which you are engaged.

The Co-operative Principle (CP), stated in its most general terms above, can be expanded into four constituent maxims:

The maxim of Quality:

Try to make your contribution one that is true, specifically:

(i) do not say what you believe to be false

(ii) do not say that for which you lack adequate evidence

The maxim of Quantity:

(i) Make your contribution as informative as is required for the current purposes of the exchange

(ii) Do not make your contribution more informative than is required

The maxim of Relation:

Make your contribution relevant

The maxim of Manner:

Be perspicuous and specifically:

(i) avoid obscurity

(ii) avoid ambiguity

(iii) be brief

(iv) be orderly.

If we now return to the example, we can explain the inference that Janet may be at or near the post office, in terms of the

maxims, as follows:

(a) At face value, B’s reply is not an answer to A’s question. It therefore appears (at face value) to be irrelevant. But the

maxim of Relation leads A to expect that B is being relevant, in spite of appearances. Hence, A looks for an interpretation

whereby ‘I saw her walking towards the post office about five minutes ago’ can be relevant. The maxims of Quantity and

Quality lead A to expect that what B says will give the right amount of information to answer the question, if B can

truthfully give that information. But what if B doesn’t know the answer to the question? Then B will give whatever

information is truthfully possible, to help A find the answer.

(b) On the basis of the above argument, as well as from general and contextually shared knowledge, A can reasonably infer

that B does not know where Janet is now, but has suggested, on the basis of what is known, that Janet may be at or near

the post office. Moreover, by the Co-operative Principle, A can assume that it was B’s intention to convey this implicit

message.

This illustration shows how extra meaning can be ‘read into’ what people say, on the assumption that people not only know

the meaning of expressions in their language, but have general knowledge and general human rationality at their command.

Like Austin, Grice writes in an informal, conversational style, and leaves a great deal of scope for subsequent interpreters

and misinterpreters. For example, many commentators have assumed that Grice’s Co-operative Principle is built on some a

priori notion of human benevolence and co-operativeness; that Grice is therefore making some kind of ethical claim about

human behaviour (see, for example, Apostel 1980, Kasher 1976, 1977 Kiefer 1979, Pratt 1977, 1981 and Sampson 1982). But

nothing is further from the truth. The CP is simply a device to explain how people arrive at meanings. There is certainly no

assumption that people are inevitably truthful, informative and relevant in what they say (for a full discussion of this, see

Thomas 1986, Chapter 2). Hence, as Grice points out, we can simply ‘OPT OUT’ of the CP, in that, for example, we may

decide to withhold whatever information we possess: ‘No comment!’ We may inadvertently infringe a maxim or we can secretly

violate a maxim—e.g. A could maliciously and falsely tell B that Janet had walked in the opposite direction from the post

office. Or—more importantly—we can make a blatant show of breaking one of the maxims (Grice terms this flouting a

maxim), in order to lead the addressee to look for a covert, implied meaning. This last kind of exploitation of the CP is basic



AN ENCYCLOPAEDIA OF LANGUAGE



99



to what Grice called CONVERSATIONAL IMPLICATURES, i.e. pragmatic implications which the addressee figures out by

assuming the speaker’s underlying adherence to the CP. It is the very blatancy of the flouting of the maxims which leads to

the generation of a conversational implicature in each of the following examples:

Quantity

At the time of recording, all the members of the cast were members of The BBC Players.

[Implicature: One or more of them are no longer members of The BBC Players.]

Manner

Interviewer: Did the U.S. Government play any part in Duvalier’s departure? Did they, for example, actively encourage

him to leave?

Spokeswoman: I would not try to steer you away from that conclusion.

[Implicature: The U.S. Government did play a role, although the speaker is not in a position to make a commitment to that

effect.]

Relation

Female Guest:

Has the doctor been?

Basil Fawlty:

What can I get you to drink?

Female Guest:

Basil, has the doctor been?

Basil Fawlty:

Nuts?

[Implicature: Basil does not want to answer the question.]

Similarly, implicit meanings of irony or of metaphorical interpretation can be explained, at least in part, by reference to the

CP. For instance, at its face value interpretation, the following example (taken from a ‘Peanuts’ cartoon) breaks the maxim of

Quality; literally speaking, older siblings (of whatever sex) are not coarse grass:

Big sisters are the crab grass in the lawn of life.

The covert interpretation ‘Big sisters are unpleasant and have a tendency to take over’ depends on the assumption that what is

intended is related to the face-value meaning, but is also relevant, truthful and informative.

Levinson, in his seminal article ‘Activity types and language’ (1979), suggests that interactants’ expectations of the degree

to which the maxims will be observed vary according to the type of interaction in which they are engaged:

‘…there may in fact be some relation between Grice’s maxims of conversation and particular expectations associated

with particular activities. Grice’s maxims of quality, quantity, relevance and manner are supposed to outline

preconditions for the rational co-operative exchange of talk. But one thing we can observe is that not all activity types

are deeply co-operative. Consider an interrogation: it is unlikely that either party assumes the other is fulfilling the

maxims of quality, manner and especially quantity.

(Levinson 1979:374)

Harnish (1979) and Holdcroft (1979) make similar observations:

There are…many garden-variety counter-examples, social talk between enemies, diplomatic encounters, police

interrogations of a reluctant suspect, most political speeches, and many presidential news conferences. These are just

some of the cases in which the maxims…are not in effect, and are known not to be in effect by the participants.

(Harnish 1979:340 n. 29)

[Our emphasis]

The implications of the activity-type specificity of the maxims are taken further by Martinich (1984:33), who argues that it is

necessary to introduce a further category of non-observance of maxims, namely ‘suspending’ a maxim (see also Thomas 1986:

44–7):

‘A person can suspend a maxim…When a person opts out of a maxim, that maxim remains operative; but when a

person suspends a maxim, that maxim becomes inoperative. Institutions that allow filibustering suspend the maxims

“Be brief” and “Be relevant.”’

Thus, if a speaker says: Did you hear the one about…, we may safely assume that the Maxim of Quality has been suspended.



100



LANGUAGE, MEANING AND CONTEXT



1.3.1

Implicatures

The notion of implicature (or ‘pragmatic implication’) has been Grice’s most important contribution to the development of

pragmatics. Its significance is that it clearly marks a departure from the kinds of inferences allowable in the truth-based study

of logic: notably, material implication and entailment. Unlike these, which are definable entirely in truth-conditional terms,

implicature depends on, or refers to, factors of context. However, Grice recognises different kinds of implicature.

Conversational implicatures (which depend on the assumption of the CP) are distinct from conventional implicatures, which

are simply associated, by convention, with the meanings of particular words. For example, but carries the implicature, for any

utterance X but Y, that Y is unexpected, given X:

She lives alone, but she has an active social life.

But here implicates that, given that she lives alone, it is not to be expected that she has an active social life. Without this

implicature, but would have the same meaning as and. More recently it has been argued (cf. Morgan 1979) that the distinction

between conversational and conventional implicatures is not wholly sustainable. Utterances such as the polite request formula

Can you X …?, whose force can at first be calculated only by means of conversational implicature may, through repeated use,

become conventionalised.

Another example of this passing from conversational to conventional implicature is provided by the expression ‘to be

economical with the truth’. When this was uttered by Sir Robert Armstrong in the High Court in Sydney during the Spycatcher

case, a meaning could have been worked out by means of a conversational implicature generated by a flouting of the maxim of

Manner. The saying was repeated so often that it rapidly became a ‘frozen metaphor’, even though he has since disavowed

any intended implicature.

Grice makes another distinction between GENERALISED and PARTICULARISED conversational implicatures. The former

is a type of inference which is made generally, without reference to a specific situation. Thus ‘not all X…’ is generally taken

to implicate ‘At least some X’. Suppose, for example, that A is a famous author, whom B meets at a party. If B says ‘I

haven’t read all your books’, A can reasonably conclude that B intends to imply ‘but at least I’ve read some of them’: and can

justifiably feel misled if it then turns out that B hasn’t read any at all.

Cases like this contrast with particularised implicatures, which arise from specific utterances taking place in specific

situations. For example:

A:

B:



Who finished off the brandy?

Geoff was using your room while you were away.



B’s reply implicates that Geoff may have finished off the brandy, but this inference depends on a highly specific context,

namely that provided (among other things) by A’s question.

Even generalised conversational implicatures, however, are distinguished from cases of logical inference in that they can be

cancelled out by a statement which is inconsistent with them. (To use the technical term, they are ‘defeasible’.) For instance,

in our ‘famous author’ example it would be possible for B, without self-contradiction, to add: ‘…In fact, I haven’t read any of

them’, thus cancelling the implicature normally derived from ‘not all’.



2.

DEVELOPMENTS OF SPEECH ACT THEORY IN LINGUISTICS

In the preceding section, we have surveyed the philosophical underpinning of pragmatics as a branch of linguistics. Now, we

consider what linguistic issues brought pragmatics into prominence in the 1970s. Our standpoint is still that of the ‘AngloAmerican tradition’, but, as far as linguistics goes, a leading role is taken by the predominantly American paradigm of

TRANSFORMATIONAL GENERATIVE GRAMMAR (TGG)—on which see Chapters 3 and 4.

The particular variant of TGG which was instrumental in developing linguistic pragmatics was the school of thought known

as GENERATIVE SEMANTICS, led by Chomsky’s disaffected former students Ross, McCawley and George Lakoff. The

main tenet of the generative semantics school was that the base component of a generative grammar was to be associated with

semantic structure. A corollary of this was that syntax and semantics were ultimately non-distinct: semantics was simply the

‘deepest’ level of syntax. But the same spirit of adventure which led to the placing of semantics at the heart of grammar also

led to the inclusion of pragmatic concerns within semantics, and hence, by extension, as part of syntax. Thus the whole of

meaning was to be dealt with in terms of the phrase structure formalism of ‘deep syntax’.