In this chapter I shall attempt to clear the way for the consideration, in later chapters, of the various aspects of semantics, first, by discussing (and dismissing) two unsatisfactory views of semantics which, though prima facie plausible, provide no solution to semantic problems and, secondly, by attempting to set out some of the more important distinctions that have to be made.
In an earlier section it was suggested that language might be thought of as a communication system with on the one hand the signifier, on the other the signified. But a basic problem is to establish the nature and relationship of these two.
One of the oldest views, found in Plato's dialogue Cratylus, is that the signifier is a word in the language and the signified is the object in the world that it 'stands for', 'refers to' or 'denotes' (I shall use these terms without distinction - though some scholars have suggested that a useful distinction can be made). Words, that is to say, are 'names' or 'labels' for things.
This is, prima facie, an attractive view for it is surely true that the child learns many of his words precisely by a process of naming. He is often given names of objects by his parents and his first attempt at language will include saying 'Da da' when he sees his father or in producing his name for train, bus, cat, etc., on seeing the relevant objects in real life or in a book.
There are, however, many difficulties with this view. To begin with it seems to apply only to nouns; indeed traditional grammar often defines the noun, as distinct from the adjective, verb, preposition, etc, as 'the name of a person or thing'. It is difficult, if not impossible, to extend the theory of naming to include these other parts of speech. It is possible, no doubt, to label colours, as is done in colour charts, and thus it may be that the colour words (adjectives) can be regarded as names. But this is not at all plausible for most of the other adjectives. Since the beginning of this section I have used the adjectives early, attractive, true, relevant, traditional, difficult and plausible. How many of these could be used as a label to identify something that they denote? The point is even more obvious with verbs. It is virtually impossible to identify what is 'named' by a verb. Even if we take a verb like run and attempt to illustrate it with a boy running (either in a still or moving picture) there is no obvious way in which we can isolate the 'running' part of it. With a noun we can often draw a picture of the object that is denoted. But this is difficult, if not impossible, with verbs. For let us consider the verb run and an attempt to illustrate what it denotes with a picture of a boy running. There are two difficulties that arise (even if we have a moving picture). First, we are not presented separately with a boy and with 'running'. We need a fairly sophisticated method of separating the two. Secondly, even in so far as we can distinguish the boy and 'what he is doing', it is far more difficult to identify precisely what are the essential characteristics of what is denoted by the verb than what is denoted by the noun. For instance, does running involve only the movement of the feet or are the arms involved too? Does it necessarily involve a change of position? Is the speed relevant? Clearly there is not something that can easily be recognised and identified as 'running'. The problem is obviously even more difficult with remember, like or see. Similar considerations hold for prepositions (up, under) and conjunctions (when, because), while pronouns (I, he) raise even more severe problems, since they denote different things at different times.
Can we, however, retain the theory of naming, but apply it to nouns alone? An obvious problem, to begin with, is that some nouns, e g. unicorn, goblin, fairy relate to creatures that do not exist, they do not, therefore, denote objects in the world. One way out of this difficulty is to distinguish two kinds of world, the real world and the world of fairy stories. But this is, of course, to admit that words are not just names of things, and it must involve some fairly sophisticated explanation of the way in which we can, by some kind of analogy, move from giving names to objects in the world to giving names to objects that do not exist. Such an explanation is possible, but such words are evidence of the fact that words are not simply names of the objects of our experience.
There are other nouns that, while not referring to imaginary items do not refer to physical objects at all. Thus we cannot identify the object to be named by love, hate, inspiration, nonsense. When the grammarian speak of nouns being names of things we can ask whether love, hate etc, are things. If they are inclined to say 'Yes, but they are abstract things', it becomes clear that the only reason why they wish to call them things is that they have nouns corresponding to them. But then the whole definition is circular, since things are what are named by nouns (see 7.2).
Even where there are physical objects that are identifiable, it is by no means the case that the meaning is the same as its denotation (the object it 'stands for' or refers to). One of the best-known examples to illustrate this point is that of the evening star and the morning star. These can hardly be said to have the same meaning, yet they denote a single object, the planet Venus.
Yet another difficulty is the fact that even if we restrict our attention to words that are linked with visible objects in the world around us, they often seem to denote a whole set of rather different objects. Chairs, for instance, come in all shapes and sizes, but precisely what is it that makes each one a chair rather than a settee or a stool? Often the dividing line between the items, referred to by one word and those referred to by another is vague and there may be overlap. For when is a hill a hill and not a mountain? Or a stream a river? In the world of experience objects are not clearly grouped together ready, so to speak, to be labelled with a single word. This is a problem that has bothered philosophers from the time of Plato. There are two extreme, but clearly unhelpful, explanations. One is the 'realist' view that all things called by the same name have some common property - that there is some kind of reality that establishes what is a chair, a hill, a house. The second, the 'nominalist' view, is that they have nothing in common but the name. The second view is obviously false because we do not use chair or hill for objects that are completely different -the objects so named have something in common. But the first view is no less invalid. For there are no clearly defined 'natural' classes of objects in the world around us, simply waiting for a label to be applied to them; part of the problem of semantics is to establish what classes there are. Even if there are no natural classes, it might be argued that there are 'universal' classes, classes common to all languages. But this is not so. The classification of objects in terms of the words used to denote them differs from language to language. If, for instance, we take the English words stool, chair, arm-chair, couch, sofa, we shall not find precise equivalents in other languages. The French word fauteuil might seem to be equivalent to English arm-chair, but whereas the presence of arms is probably an essential characteristic for arm-chair, this is not necessarily so for fauteuil. Similar considerations hold for chest of drawers, side-board, cup-board, ward-robe, tallboy, etc. The colour systems of languages appear to differ too (we shall discuss this in more detail later), in spite of the apparently 'natural' system of the rainbow. The words of a language often reflect not so much the reality of the world, but the interests of the people who speak it. This is clear enough if we look at cultures different from our own. The anthropologist Malinowski noted that the Trobriand Islanders had names for the things that were useful to them in their daily life that did not correspond to words in English (see 3.2). Another anthropologist, B.L.Whorf, points out that the Eskimos have three words for snow depending on whether it was falling, lying on the ground or used for igloos, but the Hopi use only one word to refer to 'flier', be it an aeroplane, a pilot or an insect (see 3.4).
We can, unfortunately, be misled by scientific terminology for here we often find that there ARE natural classes. If we go to the zoo, we shall find that each creature has a particular name, and that no creature can be labelled in two different ways, nor is there any overlap between the classes. A gorilla is a gorilla, a lion a lion The same is true, or very largely true, of the names of insects, plants and even of chemical substances. But these scientific classifications are not typical of everyday experience. Most of the things we see do not fall strictly into one class or another. Moreover, we should not be misled into thinking that we can and should tidy up our terminology by seeking the advice of the scientist. Of course, as literate and educated beings we will be influenced by scientific knowledge and may well refrain from calling a whale a fish or a bat a bird (though why could not fish simply mean 'a creature that swims in the water' and bird 'a vertebrate that flies'?). But we can go too far. Bloomfield argued that salt could be clearly defined as sodium chloride, or NaCl. He was wrong to do so. Salt, for ordinary language, is the substance that appears on our tables. It is no less salt if its chemical composition is not precisely that of the chemists' definition. Salt for most of us belongs with pepper and mustard, which do not lend themselves to any simple scientific specification - and neither should salt in its everyday use. Ordinary language differs from scientific language precisely in the fact that its terms are not clearly defined and its classes not rigorously established.
One possible way out of all our difficulties is to say that only SOME words actually denote objects - that children learn SOME of them as labels. The remainder are used in some way derived from the more basic use. This is in essence the proposal of Bertrand Russell who suggested that there are two kinds of word, 'object word' and 'dictionary word'. Object words are learnt ostensibly, i. e. by pointing at objects, while dictionary words have to be defined in terms of the object words. The object words thus have OSTENSIVE DEFINITIONS.
Yet much of what we have already said shows that this can be no solution. For in order to understand an ostensive definition we have to understand precisely what is being pointed at. If I point to a chair and say 'This is a chair', it is first of all necessary to realise that I am pointing to the whole object, not to one of its legs, or to the wood it is made of. That may be fairly easily established, but it is also necessary to know what are the characteristics of a chair if the definition is to be of any value. For someone who does not know already what a chair is may well assume from the ostensive definition that a stool or a settee is a chair. He might not even be sure whether the word chair applied equally to a table, since the ostensive definition does not even establish that we are pointing at a chair as something to be sat on, rather than as a piece of furniture. Pointing to an object itself involves the identification of the object, the specification of the qualities that make it a chair or a table. It requires a sophisticated understanding, perhaps even the understanding of the entire categorisation of the language concerned. As the philosopher, L. Wittgenstein, commented, 'I must already be the master of a language to understand an ostensive definition.'
To return to the child, it is clear enough that he does not simply learn the names of things. For if he did he would be unable to handle all the complexities that we have been discussing. Above all, learning a language is not learning just 'This is a ...', even less is it saying 'book' whenever he sees a book We shall not solve problems of semantics by looking at a child learning language, for an understanding of what he does raises precisely the same problems as those of understanding what adults do in their normal speech.
So far in this section we have talked about the meaning of words. But we shall also have to discuss the meaning of sentences (Chapter 6). It is enough here to point out that a naming theory for sentences is no more satisfactory than one for words. We cannot directly relate the meaning of a sentence to things and events in the world. The strongest view which relates sentences to actual things and events, such that There is a horse on the lawn means that there is a horse on the lawn, is obviously untenable, since we can tell lies or make mistakes (there may be no horse on the lawn). A weaker view is to see meaning in terms of the conditions under which a sentence would be true - the meaning of There is a horse on the lawn being thus stated in terms of 'truth conditions' involving a certain kind of animal being at a particular time on a specially prepared area of grass. But this gets us nowhere. For the truth conditions can be most easily stated in the same words as the sentence - There is a horse on the lawn is true if there is a horse on the lawn (alternatively There is a horse on the lawn means 'There is a horse on the lawn'). The tautology is obvious-we are saying nothing at all. Putting the truth conditions into other words, moreover, e. g. paraphrasing in terms of 'an equine quadruped' does not help; it merely trades on some of the semantic relations within language (between horse, equine, quadruped) and totally fails, as does the naming theory of words, to specify meanings in terms of external things and events.
The view we have just been criticising relates words and things directly. A more sophisticated and, at first sight, more plausible view is one that relates them through the mediation of concepts of the mind. This view in all its essentials has been held by some philosophers and linguists from ancient times right up to the present day. Two of the best-known versions are the 'sign' theory of de Saussure and the 'semiotic triangle' of Ogden and Richards.
According to de Saussure, as we have seen, the linguistic sign consists of a signifier and a signified, these are, however, more strictly a sound image and a concept, both linked by a psychological 'associative' bond. Both the noises we make, that is to say, and the objects of the world that we talk about are mirrored in some way by conceptual entities.
Ogden and Richards saw the relationship as a triangle.
The 'symbol' is, of course, the linguistic element - the word, sentence, etc., and the 'referent' the object, etc., in the world of experience, while 'thought or reference' is concept. According to the theory there is no direct link between symbol and referent (between language and the world) - the link is via thought or reference, the concepts of our minds.
This theory avoids many of the problems of naming-the classifications, for instance, need not be natural or universal, but merely conceptual. But it also raises a completely new problem of its own. For what precisely is the 'associative bond' of de Saussure or the link between Ogden and Richards' symbol and concept?
The most naive answer to the question is to say that it is a psychological one, that when we think of a name we think of the concept and vice versa, i. e. that meaning consists of our ability (and indeed our practice) of associating one with the other, of remembering that chair refers to the concept 'chair'. This view is totally unsatisfactory. It is not clear what exactly is meant by 'thinking of a concept. Some scholars have actually suggested that we have some kind of image of a chair when we talk about chairs. But this is certainly false. I can visualise a chair in 'my mind's eye', but I do not do so every time I utter the word chair. If this were a necessary part of talking, it would be impossible to give a lecture on linguistics. For precisely what would I visualise? The problem is, of course, that of names and things all over again. More reasonably, perhaps, what is meant is that I relate my utterance of the word chair to some more abstract concept. But that will not help either. For what is this abstract concept - what colour is this chair, what size or shape? In any case we ought not to be interested in what happens on each occasion, but with the more general question of the meaning of chair. As a phonetician, I should not be interested in the precise articulation of chair except as material for many more general statements of phonetics and phonology. Similarly, as a semanticist, I want to know about the general meaning of chair, not, what I may or may not do every time I utter the word. As we said earlier we are not concerned with utterance meaning.
A more sophisticated version sees the link not as something we make every time we use a word, but as some kind of permanent association stored in the mind or in the brain. The difficulty with this view is that it really says nothing at all. For how can we, even in principle, establish what the concepts are? There is no obvious way in which we can look into our minds to recognise them, and still less a way in which we can look into the minds of others. In effect all this theory is doing is to set up, in some inaccessible place, entities that are BY DEFINITION mirror images of the words that they are supposed to explain. Wherever we have a word there will be a concept-and the concept will be the 'meaning of that word'. This is, obviously, a completely circular definition of meaning. It involves what is sometimes called a 'ghost-in-the-machine' argument. We wish to account for the working of a machine and present a total explanation in mechanical terms, but for some hypothetical person this is not enough - he cannot understand how this machine could work unless there is some kind of disembodied ghost or spirit inside it. Such an argument accounts for the phenomena by setting up an entity whose existence is justified solely as something that 'explains' the phenomena. Science has had many examples of this kind in its long history. Once scholars explained fire by positing the existence of the substance 'phlogiston' Of course we can never disprove the existence of such entities. We can only point out that they explain nothing at all, and that, therefore, nothing is gained by arguing for them.
It is, perhaps, hardly necessary to point out that, as with naming, the sentence is no more satisfactorily defined in terms of concepts than the word. Neither the naive or the more sophisticated version of the theoiy is at all helpful. Certainly when I say There is a horse on the lawn there is no reason to suggest that I actually 'think of the concept, while a definition in terms of more abstract, timeless concepts is once again to say nothing at all but merely to interpret meaning by its mirror-image, postulated in an inaccessible place
Sadly, there are many linguists today who accept in whole or in part a conceptual view of meaning. This has stemmed from a new mentalism' associated with N. Chomsky and his followers who have, in particular, insisted that intuition and introspection must play a large part in our investigation of language. It is a short and perhaps inevitable step to see meaning in terms of the mental entities called concepts. But this must be rejected for three reasons. First, the ghost-in-the-machine objection is overwhelming - nothing is said by moving meaning back one step to the brain or the mind. Secondly, even if there were concepts in the mind they are in principle inaccessible to anyone but the individual, and we are left therefore with totally subjective views, since I can never know what your 'meanings' are. (Of course, if we had the knowledge to investigate the brain scientifically and could account fully for language in the structure of brain cells, both of these objections might be, thereby, overcome, but we are centuries away from such knowledge.) Thirdly, the arguments about intuition and introspection are irrelevant. We CAN introspect - and ask ourselves questions about our language without actually waiting for empirical data, actual recordings or texts. But in so doing we do not learn more about our language or its structure; we merely produce for ourselves some more examples of our language. As J.R. Firth said, we go 'fishing in our own tank'. What we do NOT do by this process is establish the phonological or grammatical rules or structures; this comes from the investigation and comparison of a great deal of data (even if that data is all introspective). The same must be true of semantics, and it follows that we should not believe that there are concepts that can merely be discovered if we look in the right place. It is perhaps worth considering that if scientists had continued to rely on 'reason' (i. e. to look for answers to their problems within themselves and their own rational processes) rather than-observation, we should still be searching for the philosopher's stone to turn lead into gold, rather than be on the edge of succeeding through nuclear physics.
Finally, in this section it is worth noting that to some extent DUALISM, the view of language described here and in the previous section that sees meaning as part of the signified /signifier relation, is encouraged by the term meaning itself and by the statement that words (and sentences) HAVE meaning. For if this is so it is obviously legitimate to ask what kind of entity meaning is, and to look for it either in the world or in people's minds. But to say that a word has meaning is not like saying that people have legs or that trees have leaves. We are easily misled by the verb have and the fact that meaning is a noun into looking for something that IS meaning.
In practice we all know what it is for a word to have meaning. Knowing the meaning of a word means that we can do a number of things - we can use it properly, we can explain it to others in terms of paraphrases or synonyms. But it does not follow from that that there is an entity that IS meaning or a whole group of entities that ARE the meaning of words. For a word to mean something is similar in some way to a notion that a signpost points somewhere; we can understand the meaning of a word just as we can read the signpost. But it does not make sense to ask what it is that words mean any more than to ask what it is that signposts point to. It is not sense, that is to say, to ask IN GENERAL what words mean or signposts point to. It is sense only to ask 'What does THIS word mean?', 'What does THIS signpost point to?'
The problem of semantics is not, then, nor can it be, the search for an elusive entity called 'meaning'. It is rather an attempt to understand how it is that words and sentences can 'mean' at all, or better perhaps, how they can be meaningful. If we are talking of 'having' meaning, it is rather like talking about 'having' length. Having length is being so many feet or inches long; length is not something over and above this. Similarly, meaning is not some entity that words or any other linguistic entities 'have', in any literal sense of'having'.
Wittgenstein said, 'Don't look for the meaning of a word, look for its use.' This is not a very helpful remark since we are perhaps not much clearer about the 'use' of a word than we are about its meaning. But it has some value; we can investigate use, and we are less likely to think of use as something that words 'have' in any literal sense, and less likely to waste our time in an attempt to discover precisely what it is.
2.3. Sense and reference.
I have already used the term REFERENCE in talking about the denotation of words in 2.1- It is also used in a useful but wider sense, to contrast with SENSE, to distinguish between two very different, though related, aspects of meaning.
Reference deals with the relationship between the linguistic elements, words, sentences, etc., and the non-linguistic world of experience. Sense relates to the complex system of relationships that hold between the linguistic elements themselves (mostly the words); it is concerned only with intralinguistic relations.
It might seem reasonable to argue that semantics is concerned only with the way we relate our language to our experience and so to say that reference is the essential element of semantics. Yet sense relationships have formed an important part of the study of language. For consider the words ram and ewe. These on the one hand refer to particular kinds of animals and derive their meaning in this way. But they also belong to a pattern in English that includes cow/bull, sow/boar, etc. Older grammars of English treated this as a part of grammar, because it was clearly related to sex, and sex was supposedly a matter of gender (since sex and gender are related in some degree in Latin). But there are other kinds of related words e. g. duck/duckling, pig/piglet (involving adult and young), or between father/son, uncle/nephew (involving family relationships), and these are not usually thought to be grammatical. They are rather a part of the 'semantic structure' of English. There are many other kinds of sense relations, too, e. g. those exemplified by narrow/wide, male /female, buy/sell; these we shall discuss in some detail later. The dictionary is usually concerned with sense relations, with relating words to words, though most dictionaries state such relations in a most unsystematic way (Chapter 4). It could be argued though, that the ultimate aim of the dictionary is to supply its user with referential meaning, and that it does so by relating a word whose meaning is unknown to a word or words whose reference is already understood.
We have, then, two kinds of semantics, one that deals with semantic structure and the other that deals with meaning in terms of our experience outside language. But the situation should not surprise the linguist, since he has a similar situation at the other 'end' of his language model, where we had tentatively placed phonetics (12). For linguists distinguish between PHONETICS, which deals with speech sounds as such and describes them in terms of their auditory or acoustic characteristics or of the articulations of the speech organs, and PHONOLOGY, which deals with the sound systems of languages in terms of the internal relation between sounds. We might well look for a similar distinction between 'semantics' and 'semology'. I shall not, however, use these terms because reference and sense are in current usage. Nor would I push the analogy too far. It is enough to see that there may be two kinds of semantics, one that relates to non-linguistic entities, and one that is intra-linguistic.
In recent years some linguists have attempted to limit semantics, both in theory and in practice, to sense relations. One example is to be found in a well-known article by J.J.Katz and J.A.Fodor entitled. 'The structure of a semantic theory'. (Katz and Fodor talk about 'sentences', whereas we have been largely concerned with the meaning of words, but their theory is based upon word meaning - and the question whether the word or the sentence is the basic element of semantics will be left until 6.1.) They state, 'A semantic theory describes and explains the interpretive ability of speakers: by accounting for their performance in determining the number of readings of a sentence; by detecting semantic anomalies - by deciding upon paraphrase relations between sentences; and by marking every other semantic property or relation that plays a role in this ability.' Stripped of its jargon this statement means that a semantic theory must account for ambiguity, anomaly and paraphrase (these will be explained and exemplified shortly).
The last sentence of the quotation is a most unfortunate one. It is a 'catch-all' qualification that effectively allows us to include in semantics all kinds of unspecified semantic properties or relations, but we may, presumably, assume that 'every other semantic property or relation' is of the same kind as the three that are given. Katz and Fodor give some examples of these, but a more complete and orderly list of semantic properties is to be found in an article by M. Bierwisch. He argues (following Katz and Fodor) that a semantic theory must explain such sentences as
(1) His typewriter has bad intentions.
(2) My unmarried sister is married to a bachelor.
(3) John was looking for the glasses.
(4) (a) The needle is too short.
(b) The needle is not long enough.
(5) (a) Many of the students were unable to answer your question.
(b) Only a few students grasped your question.
(6) (a) How long did Archibald remain in Monte Carlo?
(b) Archibald remained in Monte Carlo for some time.
(1) is an example of an anomalous sentence, (2) of a contradictory one and (3) of an ambiguous one (Katz and Fodor would say it has two readings); (4) illustrates paraphrase or synonymous sentences; in (5) one sentence follows from the other, while in (6) the first implies or presupposes the second.
Katz and Fodor quite specifically exclude from a semantic theory any reference to the 'settings' of sentences. Semantics is not, or cannot be, concerned with the way words and sentences are used in relation to the world around us. Bierwisch, however, talks about (i) 'the interpretation of sentences' and (ii) 'how these interpretations are related to the things spoken about', but he gives no indication how we can proceed from the one to the other. Some linguists have even further restricted semantics and defined it in terms of truth-relations between sentences, i. e. those involving logical and analytic truth (see 6.4). Only some of the suggested sense relations will still belong to semantics. All meaning that does not belong to semantics is 'pragmatics'.
This is a really extraordinary state of affairs. What seemed at first to be the essential aspect of meaning, the relation between language and the world, is to be ignored or given second place. Moreover, only a tiny part of what is usually to be regarded as meaning can possibly be stated. For although dictionaries are concerned with defining one word in terms of others and so with sense, only a minute part of dictionary definition could be handled in terms of the sense relations we have listed (but see Chapter 4). Sadly, one is tempted to conclude that when scholars have concentrated on sense to the exclusion of reference (in its widest sense), they have done so because it is easy to describe. It has structure and can be accurately and precisely determined. But this reminds one of the drunk who lost his key at his front door but was found looking for it under the street-lamp ten yards away 'because it's lighter here'.
There are some further difficulties. It is not always possible to distinguish clearly between sense and reference for the simple reason that the categories of our language correspond, to some degree at least, to real-world distinctions. Whether language determines the shape of the world or vice versa is probably a 'chicken and egg' problem, though it will be discussed in 3.4. The fact that we have ram/ewe, bull/cow may be part of the semantic structure of English, but it also clearly relates to the fact that there are male and female sheep and cattle. But we have to remember (1) that not all languages will make the same distinctions, (2) that there is considerable indeterminacy in the categorisation of the real world - as we saw in our discussion of names, some things (e. g. the mammals) fall into fairly natural classes, others do not. It is because of this that we can (a) distinguish sense and reference, yet (b) must allow that there is no absolute line between them, between what is in the world and what is in language.
Some scholars have been very concerned by the fact that if we deal with meaning in terms of the world, then semantics must include the sum total of human knowledge and for this reason have restricted their attention to sense. This argument is discussed in detail in 3.1, where it is argued that the problem of the sum total human knowledge is no less a problem for sense. Moreover, there are some terms of language that are not reducible to other terms, but interpretable ONLY in terms of the events around us. Most important are the DEICTICS (what philosophers have called 'indexical expressions'), the pronouns, I, you, etc., the demonstratives this and that and time markers such as now and tomorrow (7.4). For these cannot be paraphrased by any other forms that do not themselves indicate the real world, the present time or the relevant speakers and hearers. A theory of meaning in terms solely of sense, of intralinguistic relations, cannot even in principle handle these terms. Only a theory that accepts the relation of language to the world can do so, There is, of course, the added bonus that such a theory can handle other kinds of meaning too (next section).
2.4. Kinds of meaning.
Implicit in the view of semantics as a study of sense relations (and even more obviously as a study of truth conditions) is the assumption that it is concerned with factual information or with 'propositions' that can be either true or false.
No doubt this is one of the aspects of meaning that has to he considered what has variously been called 'cognitive', 'ideational', 'denotational', or 'propositional' meaning. But it is by no means the only kind of meaning and it is not even clear that it is the most important. Certainly we should not wish to say that the prime or only function of language is to provide information, to inform hearers or readers of 'facts' that they do not already know (though some linguists and philosophers have believed this). A great deal of our meaning is not 'ideational' at all, but is 'inter-personal' or 'social', relating ourselves to others. There are a number of ways (not all distinct) in which we can see that language is not simply a matter of providing factual information.
First, we do not merely make statements; we also ask questions and give orders. Indeed, the grammars of most, if not all, languages reflect these distinctions by having question forms and imperatives (though the grammatical function does not always correspond with the distinction between stating, asking and ordering - what is grammatically a statement can be semantically an order, e.g. You're coming tomorrow). It seems easy enough to handle questions in terms of information, since they are obviously requests for information; they can thus, in part, have an ideational meaning. But it is much less clear how we can handle orders in a similar way; they are concerned with getting action, not information.
Secondly, there are a variety of what today are called 'speech acts'. We persuade, we warn, we insinuate; we use language, that is to say, to influence other people in many different ways. This is the first aspect of language that a child teams - he discovers that by using his cries he can attract attention and then that the appropriate speech will manipulate adults into giving him food, playing with him, etc. This aspect of language has recently come to interest linguists, but its relation to ideational meaning is not at all clear yet (see 8.2).
Thirdly, much of what we say is not a statement of fact but an evaluation. Some semanticists have made a great play with the emotive difference between politician and statesman, hide and conceal, liberty and freedom, each implying approval or disapproval. The function of such words in language is, of course, to influence attitudes. There will be some further discussion of this in the section on synonyms (4.1).
Fourthly, language is often deeply concerned with a variety of social relations. We can be rude or polite, and the decision to be one or the other may depend upon the social relationship with the person to whom we are speaking. Thus we may ask for silence with Shut up, Be quiet, Would you please be quiet?, Would you keep your voice down a little please? The choice depends on whether we wish to be rude or not - and this relates to the status of the person addressed. Some parts of language are wholly social and carry no information (even if we include giving orders, etc., within information) at all. Examples are Good morning, How are you?, and all the Englishman's remarks about the weather. In most societies replies and questions are often about the family, but no real information is being sought - the speaker does not want to know about the health of the wife of the man he is talking to, but is simply making social contact. Even a great deal of 'small talk' at parties is really of the same kind. It is not intended to transmit information, but is simply part of the social activity. As W. S.Gilbert said (Patience):
The meaning doesn't matter
If it's only idle chatter
Of a transcendental kind.
Fifthly, as we have already noted, we need not 'mean what we say'. We can by the appropriate use of intonation be sarcastic, so that That's very clever means, 'That's not very clever'. We can also with the appropriate intonation imply what is not said. Thus I don't like coffee with a fall-rise intonation may well imply I like tea and She's very clever may suggest She's rather ugly. The moral of this is that semantics cannot fully succeed without an investigation of the prosodic and paralinguistic features of language' (see 1.3 and 8.2).
Sixthly, there is the kind of meaning found in the notorious When did you stop beating your wife? For this presupposes that you once beat her, though it nowhere states that you did. Similarly it has been argued that The King of France is bald presupposes that there is a King of France and that presupposing his existence does not assert it. Presupposition is thus distinct from assertion. The subject of presupposition will be dealt with in 8.4.
There may, of course, be other kinds of meaning. Notice also that if someone says There's a house over there, I may ask. What does that mean?, i.e. What am I to conclude from that? Although this is close to the first sense of meaning that we discussed in 1.1, it will be clear now that this kind of 'meaning' is surely outside semantics. It is concerned with the way we may use information, but in particular cases, and so is a matter essentially for utterance-meaning.
2.5. The word as a semantic unit.
It is normally assumed that dictionaries are concerned with words and that therefore the word is, in some sense at least, one of the basic units of semantics. Yet there are some difficulties.
First, not all words seem to have the same kind of meaning. A very familiar distinction is that made by the English grammarian. Henry Sweet between 'full' words and 'form' words. Examples of full words are tree, sing, blue, gently and of form words it, the, of, and. Only full words seem to have meaning of the kind we have been interested in so far. The form words seem to belong to grammar rather than to semantics; more strictly they belong to grammar rather than to lexicon (see 7.1). They can still be said to have meaning, but meaning of a grammatical kind. Yet this is not so much the meaning of the word itself but rather its meaning in relation to the other words, and perhaps to the whole sentence. We should not, for that reason, wish to look for the meaning of such words in isolation, but only within the sentence. (Dictionaries often attempt to define them, but with little success.)
Secondly, it is not at all clear that the word is a clearly defined unit, except as a conventional one resulting from the rules for writing that we all have learnt at school. Words as we know them are the written items between which we have learnt to put spaces. But we may well question whether this is necessarily an indication of a well-defined linguistic element. In Arabic the definite article is written as part of the word; in English it is not. There are no clear criteria for deciding which of these is the more appropriate. Or let us compare greenhouse with White House (in The White House). Apart from our conventions of spacing are there good reasons for saying that the former is one word, the latter two?
Bloomfield offered a solution by suggesting that the word is the 'minimum free form', the smallest form that may occur in isolation. But this all depends on what is meant by 'in isolation'. For we shall not normally say the, is, by in isolation. We might, of course, produce these 'words' in reply to a question such as What is the second word here? or Did you sav 'a' or 'the'? But this just begs the question. We learn to utter in isolation just those items that we have learnt to recognise as words. Bloomfield went on to identify an element smaller than the word, a unit of meaning - the MORPHEME: examples are -berry in blackberry or -y in Johnny. Later linguists were more interested in the status of such words as loved where they could identify the morphemes love- and d. Here the two elements seem clearly to have the distinct meanings of 'adore' and 'past'. But problems soon arose especially with words such as took, which appears to be both 'take' and 'past', yet cannot easily be divided. The purely grammatical status of such words is not our concern, but we must recognise that there are two independent 'bits' of meaning. The best way to handle this is not in terms of morphemes (i. e. parts of words), but rather by redefining the term word in a different, though not unfamiliar way. We have been using this term in the sense that love and loved are different words. But we could also say that they are forms of the same word - the verb 'to love' (which, oddly enough, we identify by using two words, to and love). A technical term for the word in this second sense is LEXEME. It is lexemes that usually provide dictionary headings. There will not be two entries for love and loved, but one only (and this may even include the noun love as well as the verb, though we may not wish to extend the term lexeme in a similar way). If we proceed on these lines we can talk about the meaning of words (i. e. lexemes), and independently of the meaning of grammatical elements such as plural or past tense. Instead of treating loved as the two morphemes love and d, we shall analyse it in terms of the lexeme love and the grammatical category of tense, This solution leaves us with the word (defined as the lexeme) as the unit for our dictionary. But we are still left with the meaning of the grammatical element. Sometimes it may seem to be fairly simple and independent, e. g. plural which means 'more than one' (but see 7.2). But often this is not so. Case in Latin largely marks relations within sentences - the subject, the object, etc. Gender, too, in Latin is only superficially concerned with the physical feature of sex, and its main function is to indicate grammatical relations - that a certain adjective modifies a certain noun, etc. (see 7.3). The status of these elements is thus often not very different from that of the form words we discussed earlier.
Thirdly, there is a problem with what have been called TRANSPARENT and OPAQUE words (Ullmann, 1962). Transparent words are those whose meaning can be determined from the meaning of their parts, opaque words those for which this is not possible. Thus chopper and doorman are transparent, but axe and porter are opaque. Comparison with other languages, German in particular, is interesting. In English thimble, glove and linguistics are opaque (the same is true of the equivalent French words, too); in German the corresponding words are all transparent - Fingerhut ('finger-hat'), Handschuh ('hand-shoe'), Sprachwissenschaft ('language-science'). This suggests not only that one word may be seen as consisting of several bits of meaning, but also that the number of 'bits' is arbitrary. (Do we look for elements of meaning in thimble because of the German word?) There are problems of detail too. If we decide that chopper is to be interpreted in terms of chop and -er (indicating the instrument), what do we say of hammer? Is this transparent? The -er shows that it is an instrument, but what is 'hamming'? There is, then, no precise way of determining the semantic elements within a word.
Fourthly, there are many words in English that are called PHONAESTHETIC, in which one part, often the initial cluster of consonants, gives an indication of meaning of a rather special kind. Thus many words beginning with si- are 'slippery' in some way - slide, slip, slither, slush, sluice, sludge, etc., or else they are merely pejorative - slattern, slut, slang, sly, sloppy, slovenly, etc. The sk- words refer to surfaces of superficiality - skate, skimp, skid, skim, skin, etc. The reader may consider also the meaning of words beginning with sn-, str-, sw-, tw-, etc. An arousing set is that which ends in -ump; almost all refer to some kind of roundish mass - plump, chump, rump, hump, stump, and even perhaps dump and mumps. But we cannot generalise too far. Not every word with these phonological characteristics will have the meaning suggested, and, moreover, we cannot separate this part and state the meaning of the remainder, e.g. the meaning of -ide in slide or -ate in skate.
Fifthly, semantic division seems to 'override' word division. Consider, for example, heavy smoker and good singer. Semantically these are not heavy + smoker (a smoker who is heavy) and good + singer (a singer who is good). The meaning rather is one who smokes heavily or sings well. We can divide, if we insist, but the first division has to be between heavy smok- and -er, good sing-and -er, if we want to retain the parallelism between the form and the meaning. Further amusing examples that have been suggested are artificial florist and criminal lawyer.
Sixthly, although we have ram /ewe, stallion/mare, we have no similar pairs for giraffe or elephant. We have to say male giraffe, female giraffe, or if we know the correct term elephant bull and elephant cow. Such considerations, together with the fact that we have the words cow and calf, may lead us to define bull as male adult bovine animal and to see this as an indication of four distinct elements of meaning in the same word.
Finally we have the problem of idioms. By an IDIOM is meant a sequence of words whose meaning cannot be predicted from the meanings of the words themselves. Familiar examples are kick the bucket, fly off the handle, spill the beans, red herring. The point is clear if we contrast kick the table, fly off the roof, spill the coffee, red fish. Semantically, idioms are single units. But they are not single grammatical units like words, for there is no past tense *kick the bucketed. There will be a more detailed discussion of idioms in 5.3.
All these considerations may lead us to abandon the idea that the word is a natural unit for semantics, however useful it may be for the dictionary maker. C.E.Bazell commented, 'To seek a semantic unit within the boundaries of the word simply because these boundaries are clearer than others is like looking for a lost ball on the lawn simply because the thicket provides poor ground for such a search.'
Yet we must be careful. We must not conclude from all this that we can simply ignore the words of the language and instead look for independent 'meanings', for semantic entities, that is to say, that are totally unrelated to words. (On such a view it is only when we have done the semantic analysis that we will attempt to relate our semantic units to those of the grammar, including words.) That this is quite unacceptable is shown by the discussion in this book. For throughout we have been establishing meaning by comparison of linguistic forms, almost always involving words. Idioms are notable only because they consist of several words but function like one, while in discussing morphemes, transparent words, phonaesthetics and all the rest we were nearly always comparing words with words We shall sometimes have semantic units larger than the word and often semantic units smaller than it (or at least allow that one word may have more than one 'bit' of meaning). But we shall never get away from these linguistic forms to some, 'disembodied' meanings. For consider bull again. There is nothing here to indicate size, weight, colour, speed, fitness, etc, yet these could all be regarded as relevant to its meaning. It is, rather, only those features that are brought out by contrast with cow and calf that are relevant to semantics, i.e only those that depend upon word contrasts
Chapter 4 will be called Lexical structure. This is to emphasise the point I have just been making. The alternative title Semantic structure might suggest that we can look for meanings and relations between them divorced from the forms of language.