1 Syntagmatic aspects: how to divide a sentence into lexical units

82



LANGUAGE, MEANING AND SENSE



meaning of the sentence. The second sentence, however, is different. John, to be sure, has the same semantic significance as

in the previous example (and Bill, too). So does the past tense affix. But what about pull? Obviously it does not mean the

same as it does in pull the rope. Does it, in fact, have any individual meaning? Once this question is posed, the answer is plain

—it does not, neither does leg, nor even -’s. Only a rather peculiar unit consisting of these three elements together, pull—’s

leg, can be said to carry a recognisable meaning. This unit cannot be broken down semantically any further: it is an

elementary lexical unit which means something like ‘tease’. Expressions of this sort, which are syntactically complex, but

semantically simple, are called ‘idioms’. Other examples are: to paint the town red, to be up the creek, to get cold feet, to cook

someone’s goose. Certain consequences follow naturally from the fact that the apparent constituents of idioms do not have

independent meanings. For instance, since the leg in to pull someone’s leg is not a ‘normal’ leg, it cannot be qualified as

right, or left, or wooden or injured; nor can it be topicalised (i.e. made into the topic of the sentence) or even pluralised. Thus

none of the following sentences has an idiomatic interpretation:

John pulled Bill’s left leg.

It was Bill’s leg that John pulled (not his arm).

As for Bill’s leg, John pulled it.

John pulled Bill’s legs.

Somewhat like idioms, but significantly different none the less, are ‘frozen metaphors’. These are relatively fixed expressions

(although usually not so rigidly as idioms) whose non-literal meanings can still be seen to be related to their literal meanings.

Consider She painted a glowing picture of her life in Wolverhampton. Presumably the person in question did not actually put

brush to canvas, but merely described her life in favourable terms. But the connection is obvious, unlike that between the

literal and non-literal meanings of We painted the town red. Compare also He has one foot in the grave, which in one of its

meanings is a frozen metaphor, and He got cold feet, which has an idiomatic interpretation. Frozen metaphors often translate

fairly intelligibly into another language; literal translations of idioms hardly ever make any sense. An idiom is a single lexical

item, at least from the semantic point of view. Frozen metaphors are less clear. They have a certain unity and integrity; on the

other hand, their global meanings are related at least to some extent to the meanings of their parts. Which of these

considerations is seen as the most significant will obviously vary with circumstances.

4.2

Lexical forms with more than one meaning

Many lexical forms clearly have more than one meaning: boot (‘luggage space of car’, ‘item of footwear’); rake (‘garden

implement’, ‘dissolute fellow’); tip (‘place for disposal of household refuse’, ‘piece of useful information or advice’, ‘pointed

end of object’). Other lexical forms can be found, however, which although they can be used to refer to different types of

thing, would not normally be felt to have multiple meanings. Take, for example, book (think of a paperback novel, a

dictionary, a textbook); horse (a mare, a stallion, a foal, a race-horse, a cart-horse); or red (scarlet, crimson, maroon). We

shall say that lexical forms like boot, rake, and tip are ambiguous, whereas book, horse and red are not ambiguous (at least

with respect to the interpretations mentioned) but general or non-specific. An important difference between an ambiguous

lexical form and non-ambiguous one is that in using the former a speaker will usually (unless he is indulging in deliberate

word-play) have one specific interpretation in mind, and the hearer will normally be expected to identify which it is. In

ordinary conversation this usually presents no problems because of the abundance of contextual clues: I’lljust nip along to the

bank to cash a cheque, John can’t row for toffee—we kept bumping into the bank. But if the hearer cannot identify the

appropriate sense of an ambiguous item he cannot be said to have fully understood the utterance in which it appears (this

could be the case with, for instance, We finally reached the bank). In the case of a word which is non-specific with regard to a

particular distinction of meaning, neither the speaker’s intentions nor the hearer’s understanding is required to be more

specific. If someone says I understand my brother’s new car is red, he or she may not even know what shade of red it is.

Generally speaking, at least for the quite unrelated choices of meaning illustrated so far, there is no inclusive meaning

attributable to an ambiguous form of which the separate interpretations are more specific manifestations. There is no meaning

of rake, for instance, which covers both its possible interpretations in the way that fruit subsumes apple and banana. A general

word, however, has only the inclusive meaning, although particular contexts may narrow this down. In Our teacher is

expecting a baby, for instance, although the word teacher applies indifferently to men and women, the teacher referred to is

obviously a woman.

There are various ways of testing a word to see whether it is ambiguous or not. A full discussion of ambiguity tests is not

possible here (for more detail see Cruse 1986:54–66), but it is perhaps worth looking briefly at one example —the so-called

‘identity test’. The way this works can be illustrated using the following two sentences:



AN ENCYCLOPAEDIA OF LANGUAGE



83



John is wearing a light jacket; so is Bill

John is sponsoring a child; so is Bill

Consider the first sentence: light has two possible interpretations here—‘light in colour’ or ‘light in weight’. But notice that

while we have a choice of readings for light in the first part of the sentence, in the second part (which of course has to be

interpreted ‘Bill is also wearing a light jacket’) we must stay with the reading selected in the first part. That is to say, either

John and Bill both have light-weight jackets, or they both have light-coloured jackets. The so-called ‘crossed interpretation’,

where John has a light-weight jacket, and Bill a light-coloured one (or vice versa) is not available (except in word play). In

contrast to this, there is no embargo on a crossed interpretation in the second sentence, at least in respect of the ‘boy’/‘girl’

distinction: the two children referred to may be of the same sex, or of different sexes. The possibility of a crossed

interpretation indicates generality; a restriction to identical interpretations in such circumstances indicates ambiguity.

When a word can be interpreted differently in different contexts (or even in the same context) it is useful to be able to

classify the differences as either contextual variation, polysemy or homonymy. Contextual variation is variation within a

single meaning, as with the two readings of cousin in My cousin has just had a baby and John’s cousin is married to my sister.

In such cases there is no ambiguity; in general, when ambiguity is present, we are dealing with either polysemy or homonymy.

We would not expect a dictionary, even a very large one, to list the contextual variants for words—since these are not limited

in number, that would be impossible anyway—but we would expect polysemous and homonymous variants to be listed, since

they involve different meanings.

The primary unit of lexicography is the lexeme: generally speaking, each lexeme has a separate main entry in a dictionary.

A lexeme is an abstract lexical unit which may be realised in a variety of forms. Thus, walk, walks, walked, and walking we

all variant realisations of the lexeme walk. The variant forms of a single lexeme may differ only in respect of inflectional

affixes. So, for instance, obey and disobey are different lexemes, as are kind and kindness, because they differ in respect of

derivational affixes. (The distinction between inflection and derivation is discussed in Chapter 3.) A lexeme is said to be

polysemous (or to exhibit polysemy) if it has more than one distinct meaning. A word-form is said to be homonymous (or to

exhibit homonymy) if it functions as the phonetic realisation of more than one lexeme.

How do we know whether a word-form with several meanings is a case of polysemy or homonymy? In principle this

depends on how closely related the meanings are: polysemous variants of a lexeme are relatively closely related; the meanings

of homonyms are unrelated. But herein lies a problem: the scale of semantic closeness is a continuous one, so where do we

draw the line between homonymy and polysemy? Extreme cases, of course, present no difficulty. Obviously the two senses—

i.e. distinct meanings—of lion (lion v. tiger/ lion v. lioness) are polysemous variants, and while we would expect the

distinction to be recognised in any reasonably comprehensive dictionary, we would not expect two main entries. In this case

we have evidence that the relationship between the senses is a close one, namely, the fact that a parallel relationship exists

between the senses of dog, tiger, and others. It is equally obvious that the word-form boot, with its quite unrelated meanings

‘item of footwear’ and ‘luggage space of car’, is homonymous, and represents (at least) two lexemes. In this case two main

dictionary entries would be expected. But what about expire (‘die’) and expire (‘come to the end of a period of validity’)? Is

this a matter of polysemy, or homonymy? Most lexicographers treat metaphorically related senses as polysemous variants and

do not give them separate main entries, but there is much less agreement among linguists. There is simply no clear answer.

5.

SENSE RELATIONS I: PARADIGMATIC

5.1

Introductory

At first sight the vocabulary of a language may appear to be fairly unstructured—some areas of meaning are more copiously

supplied with lexical items than others, but there seems to be a more or less random scatter overall. However, there is more

structure than initially meets the eye: certain meaning relationships between words crop up again and again, binding the

words into recurrent semantic patterns. In this section and the next we shall be looking at the most important of these

systematic relationships (a much more detailed treatment of sense relations will be found in Cruse 1986).

There are two fundamental kinds of semantic relationship between words. First, there are those which hold between words

which can be chosen at particular structure points in a sentence: Sarah stroked the cat/kitten/dog/squirrel/ horse/animal/

cushion. These are called paradigmatic relations; they represent systems of choices of conceptual categories provided by the

language. Second, there are meaning relations which hold between the words of a particular sentence—between, for example,

kick and ball, kick and foot, and left and foot in John kicked the ball with his left foot. Notice the oddness which results when

these relations are not in order: ?John kicked the ball with his left shoulder, ?John kicked the smoke with his left foot, ?John



84



LANGUAGE, MEANING AND SENSE



kicked the ball with his left hand. Semantic relations of this sort are known as syntagmatic relations; they add redundancy to

messages and serve textual cohesion. Syntagmatic sense relations will be examined in the next section. First let us look at the

paradigmatic variety in greater detail.

5.2

Synonymy

SYNONYMY amd OPPOSITENESS are the only sense-relations likely to be familiar to non-specialists. The study of both

has a long history, but until the appearance of Lyons’s Structural Semantics (1963) there was nothing of a systematic nature.

Writings on synonymy, for instance, tended to give lists of permissible differences between synonyms, but these were not

underpinned by a sound definition of the concept and the lists of differences were somewhat ad hoc. Superficially, the notion

of synonymy is a straightforward one: two words are synonyms if they have the same meanings. However there is an

inconsistency in the general use of expressions like ‘mean the same as’ which points to the existence of (at least) two

competing conceptions. On the one hand, it is extremely difficult to find a pair of words that an average group of educated

people would unanimously accept as being identical in meaning. Indeed, it is often laid down as a cardinal principle that

perfect synonyms do not exist in real languages, and therefore a writer or a translator is always faced with real choices—it

always makes some difference which word is chosen. On the other hand, we do find dictionaries of synonyms; and a child, or

foreign learner, who asks what repair, conceal or commence means will be told unblushingly that it means the same as mend,

hide or begin, as the case may be. The fact seems to be that there is an intuitively clear sense in which begin and commence,

say, do not have the same meaning, and an equally clear sense in which they do.

Those who maintain that natural languages do not have perfect synonyms are applying very strict criteria, and their

conception of synonymy is probably close to what will here be called absolute synonymy. This can be defined as follows: two

words are absolutely synonymous if and only if they are interchangeable in all contexts without any change in the degree of

semantic normality. Thus, the fact that Daddy, can you mend Teddy’s arm? is more normal than Daddy, can you repair Teddy’s

arm?, and that The Company undertakes to reimburse the cost of repairing damaged items is more normal than The Company

undertakes to reimburse the cost of mending damaged items, is enough to show that repair and mend are not absolute

synonyms. This definition does not rule out the possibility of such synonyms, but anyone who tries to find examples will soon

become convinced that absolute synonyms are, at the very least, few and far between.

Synonymy can be defined in another way, and for this we shall adopt the expression ‘descriptive synonymy’. As its name

implies, descriptive synonymy means sameness of descriptive meaning. Descriptive meaning, as we have seen, is what

determines the truth-conditions of sentences, so descriptive synonymy can be defined as follows: two lexical items are

descriptive synonyms if and only if they can be substituted for one another in all declarative sentences without affecting their

truth-conditions. By this definition, begin and commence are descriptive synonyms. This means that, for instance, The recital

began ten minutes ago and The recital commenced ten minutes ago have identical truth-conditions, that is to say, in any

situation in which one of them is true, the other must also be true, and in any situation in which one is false, the other is

likewise necessarily false.

Since descriptive synonyms are required to be equivalent only in respect of descriptive meaning, and since few, if any, are

absolute synonyms, it follows that in the vast majority of cases descriptive synonyms differ in meaning— perhaps in

expressive meaning, or evocative meaning, or both. The members of a pair of descriptive synonyms therefore are likely to

have different conditions of appropriacy, and must be used with care. Dictionaries commonly use descriptive synonymy (or

something even weaker) as a criterion of equivalence, which is why a dictionary can be a dangerous weapon in the hands of a

foreign learner. Multiplicity of synonyms is most striking in the so-called taboo areas: think of the range of descriptive

equivalents, polite and not-so-polite, technical or colloquial, for urinate or make love, for example.

A glance through any dictionary of synonyms suggests that the compilers have not felt constrained by the strict definition

of descriptive synonymy. In fact, probably only a minority of entries conform to this definition. Most are what might be

termed ‘near-synonyms’. Typical examples are brave:fearless, strong:powerful and murder:kill. A strict definition of this

relation is not at present available—it may not, in fact, be well-defined. It will be enough for present purposes to say that nearsynonyms are identical in the central aspects of their meanings, but differ in relatively peripheral aspects. So, for instance,

murder and kill are both centrally concerned with the causing of death; murder, however, has additional features of illegality

and intent.



AN ENCYCLOPAEDIA OF LANGUAGE



85



5.3

Relations of inclusion and exclusion

There are two parallel but distinct sorts of exclusion and inclusion relation; one type is to do with classification, and the subdivision of larger categories into smaller ones; the other type is to do with dividing things into their constituent parts. Let us

begin with the classificatory type.

Consider the relationship of inclusion between tree and willow. This can be viewed in two complementary ways. Either we

can say that the class of trees includes the class of willows as a sub-class, so that any member of the class of willows is

automatically and necessarily also a member of the class of trees (but not vice versa); or we can say that the meaning ‘willow’

incorporates the meaning ‘tree’. In the former case we would be viewing the relationship from the standpoint of reference,

while in the latter case we would be viewing it from the standpoint of sense. This very important semantic relation is called

HYPONYMY: in a pair of words so related, the more specific term is called the hyponym, and the more general term the

superordinate. Hence, in the pair tree:willow, tree is the superordinate, and willow the hyponym.

A useful test for hyponymy is based on the one-way entailment which holds between parallel sentences containing

hyponym and superordinate. So, for instance, the relationship of hyponymy between willow and tree is reflected in the fact

that It’s a willow entails, but is not entailed by, It’s a tree. (That is to say, if It’s a willow is true, then the truth of It’s a tree

necessarily follows; but if it is known only that It’s a tree is true, then nothing can be inferred about the truth of It’s a willow.)

Other examples of hyponymy are: trout:fish, lion:animal, carnation:flower, lorry:vehicle, hammer: tool, cathedral:building,

iron:metal, petrol:liquid.

The relation of inclusion just described is partnered by an equally important relation of exclusion. Consider how trout and

pike, or willow and sycamore are related. It may at first be thought that this is just difference of meaning. But there is more to

it than that. Plumber and golfer are also different in meaning, but there is no reason why someone should not be both. In the

case of willow and sycamore, however, while it is perfectly possible for a tree to be neither (if it is, say, a larch), it is not

possible for it to be both. This relationship is called INCOMPATIBILITY. We can recognise it by the following entailment

relations between sentences: It’s a willow entails but is not entailed by It’s not a sycamore; likewise, It’s a trout entails but is

not entailed by It’s not a pike. Incompatibles frequently come in largish sets, all hyponyms of a single superordinate: trout,

salmon, pike, carp, cod, hake, etc. (all hyponyms of fish). In this way, hyponymy and incompatibility are essential to the building

up of classificatory schemes for objects and phenomena in the world.

Relations of inclusion and exclusion of the second type hold between the names of things and the names of their parts. The

inclusion and exclusion are thus, at least in the central cases, spatial in nature. The name of this semantic relationship is

MERONYMY; the term designating the whole is called the HOLONYM, and the term designating the part is called the

MERONYM. Examples of meronymy are: finger:hand, spoke:wheel, petal:flower, lens:telescope.

The meronymic relation of exclusion holds between the different parts of a particular whole: properly constituted parts do

not overlap. It will be clear that meronymy, like hyponymy, is an important cognitive structuring relationship. It is part of the

human desire for order and intelligibility that names of objects, processes and so on are related in the ways detailed in this

section; well-constructed naming schemes are powerful conceptual tools.

5.4

Opposites

Opposites have intrigued laymen and scholars alike from antiquity. They still present a nest of problems of analysis and

description. The basic idea of oppositeness is apprehended at a very early age, so it must be cognitively simple, but a fully

satisfactory characterisation of the notion has yet to be put forward. It seems likely, however, that at the heart of any

opposition there are two component features: one is binarity—a quintessential two-ness; the other is some idea of

confrontation—in the most elementary cases a purely spatial one, but in many cases an analogical extension of this. A number

of distinct types of opposite can be defined, each with its own peculiar properties. Here we shall illustrate only the major

varieties: complementaries, antonyms, reversives and converses (for the most part, the terminology established by Lyons

(1963) will be used).

5.4.1

Complementaries

Examples of this type are: dead:alive, true:false, open:shut. The picture presented by a pair of complementaries is of a

conceptual area exhaustively divided into two compartments in such a way that anything that does not fall under one of the

terms must necessarily, under pain of unintelligibility, fall under the other term. Hence the denial of one term is tantamount to

the assertion of the other. In other words, It’s not dead, for example, entails and is entailed by It’s alive (in the same way, It’s



86



LANGUAGE, MEANING AND SENSE



not alive entails and is entailed by It’s dead). Notice that dead and alive are also incompatibles (like, for example, red and

blue) by our previous definition. However, in the case of ‘mere’ incompatibles, the denial of one term is not logically

equivalent to the assertion of the other: It’s not red does not entail It’s blue.

5.4.2

Antonyms

(This term is often used to refer to opposites of any type; here it has a more specific meaning.) Typical members of this

groups are: large:small, long:short, fast:slow, strong:weak, heavy:tight, hard:soft, difficult:easy. Antonyms differ from

complementaries in two fundamental ways. Firstly, denying one term is not tantamount to asserting the other: just because

something is said to be ‘not short’, it does not follow that it is ‘long’. Secondly, members of an antonym pair denote degrees

of some variable property which are greater or lesser than some reference value, which itself varies from context to context.

It’s long and It’s short, for instance, do not designate any particular lengths in absolute terms—think of the sort of lengths

indicated by long fingernails and long rivers, for instance; rather, they designate values on the scale of length which are

greater or less than some average or expected value. Antonyms exemplify to perfection the linguistic property of gradability:

they can be freely modified, that is, by intensifying expressions such as very, slightly, extremely, rather, and so on, and they

occur readily in the comparative and superlative degrees, whether these be marked inflectionally, as in longer or slowest, or

formed periphrastically, as in more intelligent, most polite.

5.4.3

Reversives

Under this heading we find pairs such as enter:leave, rise:fall, advance:retreat, ascend: descend, appear:disappear, tie:untie,

dress:undress. All of them denote changes of state of some sort, which are, in principle at least, reversible, and each verb in a

pair refers to a process which, in principle, could undo the process indicated by its partner and restore the original state. Thus,

if someone advances two steps, he can regain his original position by retreating two steps; if a suitcase gets unpacked, the

original disposition of contents can be achieved by packing it again, and so on. Interestingly, although reversives refer to

processes, it is not necessary for the processes denoted by the members of a pair to be strict reversals of one another: what is

necessary is simply a reversal of the direction of change between terminal states. An example will make this clear. Take the

pair fasten:unfasten: the action of fastening something may not be exactly the same as the action of unfastening executed

backwards. But this does not affect the reversive relationship. What is important for this is simply that the object undergoing

the process of fastening should start out unfastened and end up fastened, while something undergoing the process of

unfastening should start out fastened and end up unfastened.

5.4.4

Converses

The pairs above:below, before:after, precede:follow, buy:sell, and lend:borrow exemplify this category. Converses are

sometimes called relational opposites. They all express relationships between two (or more) people or things. Both members

of a pair express what is essentially the same relationship, but viewed from the vantage point of different participants. Take,

for instance, buy and sell: Brian sold the car to Michael and Michael bought the car from Brian both indicate that a particular

transaction has taken place. But the first sentence highlights Brian’s role in the proceedings, while the second focuses on

Michael. These sentences express a relationship between three participants: Brian, Michael and the car (tacitly, there is a

fourth, namely, a sum of money); for some converse pairs there are only two participants: Brian preceded Michael, Michael

followed Brian. It is worth noting that the comparative forms of antonym pairs express a converse relationship: Brian is taller

than Michael, Michael is shorter than Brian; so do the active and passive forms of transitive verbs: England beat Austria,

Austria was beaten by England.

5.5

Markedness and polarity

Lexical opposites exhibit two very interesting and not yet fully understood properties, which are interlinked in complex ways:

these are polarity, and markedness. We shall begin with polarity. There is a strong intuitive feeling in the case of most lexical

oppositions that there is a positive term and a negative term, even when no negative prefix is present:

positive terms: long, heavy, good, clean, alive, true, build, light



AN ENCYCLOPAEDIA OF LANGUAGE



87



negative terms: short, light, bad, dirty, dead, false, demolish, dark

It turns out to be extremely difficult to define exactly what is meant by ‘positive’ and ‘negative’: different criteria seem to operate

in different instances. However, it can be said that positiveness is generally associated with a relatively high (as opposed to

low) value of some scaled property such as length, speed or weight: long:(short), fast:(slow), heavy:(light); with a favourable

(rather than unfavourable) evaluative attitude: good:(bad), clean:(dirty), kind:(cruel); and with constructive or ordering

activity (as opposed to destructive or disordering activity): build:(demolish), pack:(unpack), arrange:(disarrange). Negative

terms, in addition to being opposed to positive terms in the ways just mentioned, sometimes produce a reversal of polarity

when applied to themselves, which does not occur with positive terms:

It’s false that it’s false entails It’s true

It’s true that it’s true does not entail It’s false

(This is analogous to the arithmetical (−1)×(−1)=(+1).) Another characteristic of negative terms is that they usually do not

accept negative prefixes: compare untrue:(?unfalse), unkind:(?uncruel), unclean:(?undirty).

Markedness also shows itself in a number of different ways. The most interesting manifestations appear with antonyms.

The terms of an antonym pair are typically strongly asymmetrical in their properties. For instance, only one term from a pair

will yield a neutral question of the form How X is Y?: someone who asks How tall is John? betrays no preconceptions

concerning John’s height; on the other hand, the somewhat less probable question How short is John? would indicate that the

questioner was assuming that John was on the short side. The ability to form a neutral question is a characteristic of unmarked

terms, so tall is the unmarked term and short the marked term of this particular opposition. The same criterion indicates that

big, much, many, fast, wide, deep and far are the unmarked terms of their respective oppositions. Only one term of a pair of

antonyms (again the unmarked one) is capable of yielding a name for the underlying dimension: we say Its length is two

metres rather than ?Its shortness is two metres, and Its width is two metres rather than ?Its narrowness is two metres. We

describe a person as fifty years old, and not, except jocularly, as ?fifty years young, and as two metres tall rather than two

metres short. As a final instance of asymmetry, we may note that unmarked terms are much more readily modified by such

expressions as twice as and half as: compared twice/half as many, twice/half as long, twice/half as big, which are all normal,

with twice/half as few, twice/half as short, and twice/half as little, which most people find a trifle odd.

There is a close association between markedness and polarity. The relationship is, in fact, quite complex, but briefly, there

is a strong tendency for the positive term to be also the unmarked term.

6.

SENSE RELATIONS II: SYNTAGMATIC

6.1

Semantic normality and abnormality

It is obvious that some combinations of words are normal in that they constitute a fully acceptable phrase or sentence for

which it is easy to think of an occasion of use; other combinations are not normal in this sense. There are two principal

dimensions of normality. To begin with, words must fit together grammatically: When shall we leave? is a grammatical

sentence, *When the from why? is not. But the rules of grammar do not guarantee a good semantic fit: ?The beans were too

melodious for cream punishment is perfectly grammatical, but is semantically very odd. (It is true that it is virtually

impossible to construct a grammmatical sentence that defies all attempts at interpretation, but this is a reflection of the

malleability of meaning, not of the sapience of syntax.) Our concern here is with the semantic dimension of normality.

We have seen that the patterns of normal and abnormal occurrence of a word are a reflection of its meaning. In a sense,

there is only one kind of normality, but there are several types of abnormality, and this fact makes abnormality potentially

more informative than normality as an indicator of a word’s meaning. We may first of all divide semantic oddness into two

broad types. On the one hand there is semantic clash—?a radiant scowl, ?a whopping great terminological inexactitude, ?

melodious beans—and on the other hand, semantic redundancy—?female aunts, ?Bite it with your teeth, ?a bad disaster.

Intuitively this division is fairly clear. It is a little more difficult, but still worthwhile, to distinguish varieties, or at least

degrees, of semantic clash. The least serious degree of clash may be termed inappropriateness; here a wrong synonym or nearsynonym is involved: ?I was sorry to hear that your father had kicked the bucket. Among more serious clashes, we can

distinguish contradiction—?senile youngster, ?John blushed green—and incongruity—?melodious beans, ?hand-knitted

scent. The difference between these is mainly one of degree. Contradiction involves a clash of values along a particular

dimension: for instance, blushing involves a change of colour, but green is the wrong colour. Incongruity arises when the



88



LANGUAGE, MEANING AND SENSE



incompatibility of meaning reaches up to the highest levels of abstraction, i.e. when whole dimensions are inappropriate. For

instance, not only is melodiousness not a familiar property of beans, but beans do not make any characteristic noise at all, so

the anomalousness of melodious beans is not merely a matter of choosing the wrong type of noise, but is more radical.

6.2

Sectional restrictions, collocational restrictions and presuppositions

Semantic oddness arises because of the semantic constraints lexical items impose on their companions in sentences. These

constraints, called CO-OCCURRENCE RESTRICTIONS, are a necessary consequence of having meaning: we would be

justified in describing a word as meaningless that showed no preferences at all with regard to its companions. Co-occurrence

restrictions involving descriptive meaning are dependent on grammatical structure. For instance, a transitive verb will

typically impose different constraints on its subject and object; hear, for instance, demands an animate subject, but not an

animate object: The cat/?chair heard the mouse/bell. But the object of hear is required to be a sound (John heard the scream),

a sound-producer (John heard the cat) or something abstract that can be embodied in the form of sound (John heard the

story).

In principle, co-occurrence restrictions are multilateral; but in any pair of words which interact with one another

semantically it is usually possible to identify a ‘selector’ and a ‘selectee’—that is to say, the constraints operating in one

direction are more easily definable and more stringent than those operating in the other direction. Consider the relationship

between John and drank in John drank the potion. It is difficult to make any useful generalisation about the meanings of

verbs which will accept John as subject, but it is relatively easy to say what sort of subjects drank requires; drank is thus the

selector and John the selectee. The selector in drank the potion is again drank: little meaningful generalisation is possible

concerning what sort of things can be done to potions; on the other hand, we know that the objects of drank must normally

refer to liquids.

Two sorts of co-occurrence restriction can be distinguished—SELECTIONAL RESTRICTIONS and COLLOCATIONAL

RESTRICTIONS. The difference lies in whether the restrictions are an inescapable consequence of the meaning of a word

(selectional restrictions), or whether they are, as it were, extraneous to the core meaning of the word (collocational

restrictions). Consider, first, an example of selectional restrictions. It is obvious that to be normal the direct object of the verb

kill must be capable of referring to something that is animate and alive at the time the action is carried out. It makes no sense

to kill something that is already dead, or that is inanimate, like a table or a chair: hence both ?John killed the chair and ?John

killed the corpse of the pirate violate the selectional restrictions of kill Contrast this with the oddness of The Vice-chancellor’s

wages are hardly enough to make ends meet (as opposed to The Vice-chancellor’s salary is…). The restriction which is

violated here does not arise logically from the concept of ‘earnings’, which in any case is common to both wages and salary;

rather, it is attached as a kind of peripheral extra. This is an example of collocational restriction. Roughly speaking, if a semantic

oddness can be cured by replacing one or more items by their synonyms (or near-synonyms), then we are dealing with

collocational restrictions. (Put another way: violating collocational restrictions leads to inappropriateness; violating

selectional restrictions leads to contradiction or incongruity.)

Selectional restrictions are usually explained in terms of lexical presuppositions. These are a semantic property of selectors.

Normally, when we use a word which exerts selectional restrictions, we presuppose, or take for granted, that those restrictions

are satisfied; but this is not normally part of what is being stated, questioned, commanded or whatever in the utterance. It is

necessary at this point to distinguish presupposition from entailment. Suppose someone says John killed the curdgeon. Even

if we had never heard of a curdgeon before (and who has?), we would be justified in inferring both that a curdgeon was a

living thing, and that the curdgeon in question was now dead. But these two inferences have different statuses. The fact that

the curdgeon is dead is a logical entailment of John killed the curdgeon; it is part of what is being stated. The fact that a curdgeon

is a living thing is presupposed by the use of kill but is not actually stated. Notice that neither John didn’t kill the curdgeon

nor Did John kill the curdgeon? entails The curdgeon is dead, but they both presuppose that a curdgeon is a living thing, and

will engender an inference to that effect in a hearer. Constancy under negation and interrogation can be taken as diagnostic of

presupposition as opposed to entailment. (The reader should be warned that the topic of presupposition is a complex and

contentious one and should refer to Chapter 6; a fairly naïve position has been taken here, which should, however, be

adequate to identify the phenomenon.)



AN ENCYCLOPAEDIA OF LANGUAGE



89



7.

LEXICAL FIELDS

7.1

Saussure and structuralism

The idea of a lexical field—a structured group of words with related meanings that perhaps has some sort of distinctive life of

its own—has been around since the early days of modern linguistics. It still shows some vitality, although there are, and have

been, a number of different versions of it. The development of what is now known as field-theory was strongly influenced by

Saussurean structuralism, a few of the relevant concepts of which we now need to examine.

A convenient starting-point is Meillet’s dictum that a language is a system ‘où tout se tient’—that is to say, a relational

structure in which everything hangs together with everything else. This principle applies throughout language, from

phonology to semantics. For Saussure, a language system consisted, at every level, of sets of paradigmatic choices, arranged

along the syntagmatic axis according to definite principles of combination. (Halliday succinctly summarised this picture as a

union of ‘chain and choice’.) There is a certain obviousness about this, at least once it has been pointed out. But Saussure’s

conception is more radical than it might appear at first sight. He held that linguistic units (of whatever sort) did not possess

inherent significance in isolation, but acquired their ‘valeur’—their linguistic value—only by virtue of their relationships,

paradigmatic and syntagmatic, with other units in the system. Applied specifically to lexical semantics this principle means

that it is useless to inquire into the meaning of warm, for instance, without at the same time examining its relations with hot,

cool, cold, scorching, freezing, and so on. This is because the meaning of warm is basically a point in a network of contrasts.

The same applies to concrete nouns like cat: we cannot say what cat means in isolation, because it, too, is a point in a system

of contrasts, along with such other items as animal, kitten and dog. (The connection between these ideas and the definition

given earlier of sense will be obvious.) As Lyons has pointed out, a consequence of this interdependence of word meanings is

that it is impossible for a child (or anyone else) to learn the meaning of a single item out of a structured set without at the

same time mastering the other members of the set: one does not know the meaning of dog until one can refrain from applying

the label to foxes and wolves, for instance.

Each language, according to Saussure, is in principle a unique system, different from all others, and should be analysed in

its own terms. This, too, has relevance for lexical semantics: an important consequence of a language’s uniqueness is that

there will inevitably be some lack of congruence between any two languages (and not just at the semantic level). Lexical

incongruence— the lack of ‘semantic fit’ between words from two different languages—will be acutely obvious to anyone

who has tried to translate a passage word-by-word from one language to another. The fact is that each language ‘packages’ its

meanings differently, both syntagmatically (notice the different distribution of the elements of meaning in John ran up the

stairs and Jean monta l’escalier en courant) and paradigmatically (both watch and look at are normally rendered into French

as the unambiguous regarder; English does not distinguish between fleuve and rivière, and the distinction between rivière and

ruisseaudoes not correspond exactly to that between river and stream).

7.2

Trier and the ‘mosaic’ model of a lexical field

For a long time the study of lexical semantics consisted mainly in tracing the history of the meanings of individual words—a

study by no means without interest, but not lending itself to systematisation or abstraction. It was the work of Trier and his

followers that initiated what has been called ‘a new phase in the history of semantics’ (Ullmann 1962:7). Since Trier, a large

body of descriptive and theoretical work has been carried out under the banner of field-theory (for a review, see Geckeler

1971). We shall illustrate the general approach by reference to Trier’s work.

Trier’s innovation was to adopt an out-and-out structuralist approach to word meaning. He viewed the entire vocabulary of

a language as an integrated system, in which each lexical item was defined and delimited by its relations with its fellows in

the system. He pictured the vocabulary of a language as forming a kind a mosaic, with individual words playing the part of

tessellae. Individual words were grouped into larger units called lexical fields, which themselves formed mosaic-like elements

of the whole vocabulary. These primary and intermediate units of the mosaic were held to cover completely, without gaps, all

areas of meaning covered by the vocabulary. Trier saw the whole system as being in a state of flux. Hence, if a change

occurred in any item, at whatever level, consequential changes in neighbouring items would inevitably follow.

This model of semantic structure led to a new way of studying word meanings. Tracing the changes occurring in the

meanings of individual words, said Trier, was of limited interest; what was much more revealing was to look at changes in a

whole system over time. Needless to say, the whole vocabulary of a language could not be studied in this way; what was

feasible, however, was to study the more limited lexical fields, that is to say, sets of words covering a particular conceptual

area. Trier’s most famous example concerned the conceptual field of knowledge and understanding in Medieval German.