The last section dealt with structure within an individual word sense; this one deals with relations among two or more such senses – often referred to as “sense relations”
or “lexical relations”. The relations dealt with here are the paradigmatic ones of synon- ymy, antonymy and so on; syntagmatic relations such as modification are dealt with in Chapters 7 and 8, on hierarchic structure, since the relation of modifiers to head is by its nature hierarchic.
I begin indirectly by recalling the scale between cognitive and linguistic domi- nance, set out in Chapter 4 (§4.2.1), on areas of meaning. To Gentner and Boroditsky (2001), the scale is a developmental one. Children’s word learning is at first dominated by cognition, as they learn words that name things and concepts they already know – cat, cup, bed, perhaps; relational terms such as aunt are more difficult to learn, relying on knowledge of both language and the world; grammatical meanings come last, re- lying almost entirely on grasping language. The scale is also relevant synchronically, to semantic structure. Applying the scale to senses rather than words, and restricting it for the moment to concrete senses, the scale has purely referential senses such as those denoting people and places at the cognitive extreme. Entity senses such as ‘dog’
and ‘spoon’ come next, being strongly cognitive but carrying descriptive elements.
In the middle come event senses such as ‘skate’ and ‘enter’, and property senses like
‘red’ and ‘big’. Many spatial senses are strongly linguistic, since they rely on linguistic distinctions between cognitively similar situations, as with ‘over’ / ‘above’, and ‘in’ /
‘within’. At the linguistic extreme come senses such as those of pronouns, of and and
the, and grammatical auxiliary verbs; their senses may be defined by their place in the linguistic paradigm.
The relation between words that are wholly cognitive such as London and Bir- mingham is that they denote different places; they have no relationship in linguistic semantics. Similarly, the “synonymy” of ‘ice’ and ‘steam’, the “hyponymy” of ‘reptile’
and ‘dinosaur’, and the “antonymy” of ‘positive’ and ‘negative’ (of electric charges) are all cognitive; they will not be studied here, though they are often treated as if they were linguistic.
There are other problems in studying “synonymy” and so on. The terms are usu- ally applied to words; but words have so many senses, and such varied ones, that to generalise about one word’s relations is often impossible. Next, senses vary in different contexts, according to the frame or scenario into which they fit, so that a sense which is usually treated as synonymous with another may be contrasted with it in another use; for example, snake <2> and serpent <2b> both denote a treacherous or deceitful person, so can be synonymous; but serpent can evoke ‘malicious’ or allusion to the Biblical story of Adam and Eve, thereby contrasting with snake.
Antonyms, synonyms, hyponyms and so on thus form indefinite classes, not categories. (That view of sense relations is supported by Geeraerts (2010, p. 89) and Murphy (2003, p. 11), for example, and it fits the treatment given by Cruse (2011).) I will not study them as categories, since my concern is with the nature of semantic structure, not with the nature of particular items or classes of item; and I use the terms to relate my discussion to other writers’ discussions, not as technical terms.
5.3.2 Sense relations: Synonymy Introduction
Synonymy will be understood here as follows. Two senses are synonymous if their similarities are more important than the differences, for the purpose in hand. It is assumed that there are no pairs of words that are identical in meaning, and that even the sense of a single word is never the same in another use. (Even if repeated imme- diately, and therefore in the same context, it will change: if the answer “No,” becomes
“No, no”, the second occurrence of no will be either stronger or weaker.) Synonymy is structured by identity and difference in the areas, types, and dimensions of meaning, as set out in Chapter 4.
Areas of meaning
Synonymy varies according to the area concerned, cognitive or linguistic. Cognitive- ly, the words water, ice, steam, and H2O are synonymous in referential use; ice and steam (and in some contexts, water) may have extra concepts defining their state.
Linguistically, they are distinct: H2O is technical in social meaning, ice and steam are semi-technical, and water is general.
Frames and scenarios – structures of cognitive meaning – also structure synony- my. For example, bill is an approximate synonym for invoice in the scenario for dining in a restaurant, but not in other scenarios. Light has different synonyms according to its various frames: in the soil frame, it means ‘friable’ (sense <6>); in the clouds frame, it means ‘fleecy’ (also <6>); it occurs in other frames as ‘digestible’, ‘graceful’, and ‘unstressed’.
Types of meaning
If we take synonyms for woman, such as lady, dame, bitch and so on, we see that they share their descriptive meaning, but differ in the type of meaning they combine with it, giving speakers a very wide choice of what they can convey. The speaker can invoke derogatory affective meaning with lady, female or broad, or jocular meaning with dame, wench or female. (The ratings of type are all from SOED.) The speaker can also use lady to express a courteous, honorific, respectful, or ironically respectful attitude.
Speakers can identify themselves as North American with broad or dame, as Scots with lass, or Australasian with sheila. Readers will identify the writer as belonging to the past if they meet wench, or some uses of maid, or (once again) lady. Speakers can set an informal tone with hen if they are Scots, or (yet once more) with lady; they can make the tone poetic with maid, dame or lass, or slangy with sheila, dame or broad.
(I have not used slut as an example, because SOED gives it no rating at all for usage or level, although to me it is the most powerful of all affective words for ‘woman’.) Dimensions of meaning
Synonyms are also distinguished by their position on the various dimensions of the descriptive meaning type.
On the generality dimension, many of the words for ‘woman’ discussed in the last paragraph can be used in a general sense, denoting the concept ‘woman’, no more.
Doll in its sense <3b> is more particular: ‘a woman who is pretty but unintelligent or frivolous’. Matron <1>, however, is extremely particular: ‘a married woman, especially characterised by dignity, staid discreet behaviour, and plump, motherly appearance’.
Cool and frigid differ on the intensity dimension; boiling and hot, applied to water, differ on the vagueness dimension. Bread and loaf denote the same food, but differ on the boundedness dimension.
Salience distinguishes even the closest synonyms. For example, brave, bold, cou- rageous, valiant and intrepid are used by SOED to give the meaning of the others; they share most of their descriptive sense elements. But bold is unique in having ‘enterpris- ing’ as an element, which is accordingly more salient in it than the other elements;
intrepid is distinct in featuring ‘fearless’; valiant has ‘on the field of battle’ as salient.
Courageous seems to be different from the others in having no unique feature.
Expectedness is important also, as with leaden, sombre and dull. Obviously, dull has the conceptual meaning ‘dull’, as a necessary element. Some senses of leaden in- voke it as an expected element, because its sense <3c> is “lacking animation…”, which
equates with “…not lively…”, from the relevant sense of dull. Sombre can evoke ‘dull’
as a possible element, in sense <2>, “…oppressively solemn or sober”.
Discussion: Synonymy
The distinctions within the sense of synonyms provide a powerful source of expres- siveness in English. Consider example (3).
(3) “In the war led by Margaret Thatcher’s government – against the left, the trade unions, the post-war consensus – her side was crushingly, devastatingly, humiliatingly victorious.” (New Zealand Herald, April 19th 2013, page A25) The second and third of the underlined words are clearly intended to add extra mean- ing, so the reader must find something extra in their minor meanings, although all three mean ‘overwhelmingly’. Devastatingly must invoke its elements, ‘lay waste’, ‘rav- age’ and ‘make wretched’. Humiliatingly is normally a weaker word than the others;
but, because of the climactic structure, must be assigned a stronger meaning through intensification of its affective value. (It then has the interesting implicature that, for Thatcher’s enemies, being humiliated was worse than being crushed and devastated.
We may note also, in anticipation of later discussion, that the elements that supply the force of the last phrase are evoked partly by the rhythm and the sound symbolism.) Conclusion: Synonymy
Discussions of synonyms usually present them as being related only in their descrip- tive elements. But from this discussion we conclude that synonyms are related also through the areas they call on (cognitive or linguistic), and the types and the dimen- sions of meaning they use. The areas, types and dimensions constitute layers of the semantic network.
5.3.3 Other sense relations Antonymy
The term “antonymy” is more unsatisfactory than “synonymy”, for its range of mean- ing and the overextension of its use. See Cruse (2011, Chapter 9) for a useful discus- sion. Cruse rightly insists that we need a more general term, “incompatible,” to cover not only antonymous senses, but also senses denoting things on the same level of a taxonomy, senses denoting parts of the same whole, and hyponymous senses. Howev- er, since Cruse (2011) and Lyons (1977) provide a good coverage, and the issues that concern us here are unaffected, I will keep to the familiar concept of antonymy.
As argued in §5.3.1, many relationships usually thought of as antonymous are cognitive rather than linguistic; they will not be discussed here. There are, however, some linguistic antonyms, distinct from cognitive ones.
The linguistic distinctions between synonyms also apply to antonyms. Well wa- tered and fruitful are antonyms for arid when it means ‘parched’ or ‘barren’, of ground
(sense <2> of arid); but interesting is an antonym for arid when it means ‘uninterest- ing’ (sense <3>). (For dry and wet in their referential senses, the opposition lies in the conceptual area shared by language and knowledge.)
The words execute, lynch, slay, kill, murder and so on have synonymous senses, but they are distinguished from each other by antonymous descriptive elements. Each means ‘deprive of life’; but execute includes ‘by process of law’ whereas lynch is oppo- site in including ‘without process of law’. Similarly, some of them are distinguished by antonymous meaning types: slay in its usual sense is literary, archaic or jocular, in social meaning; in that respect, it is opposite to murder, which is a standard word.
(Slay also has some distinct minor sense elements.) Hyponymy, and so on
Cruse (2011) discusses several other relationships. Hyponymy, for example, is well- known; its hierarchies include taxonomies – from tableware down to teaspoon – and
“meronomies” of parts – from body down to fingernail; see §5.4.3 below for further discussion. “Chains” include haze, mist, fog, peasouper. “Grids” includes words relat- ed in two ways at once, as with sex and species, in man / woman // ram / ewe. Some authors discuss “semantic” or “lexical” “fields”, such as those of colour and minerals.
I will not discuss them, since the distinctively linguistic elements involved are dealt with above.
5.3.4 Variation in sense structure 5.3.4.1 Introduction
Section 5.3.2.4 above showed variation in the salience of different sense elements, and in selection among the expected and possible sense elements. That leads to the issue of what controls the variation in the structure of senses, when words are used in utter- ances. That variation and what controls it is the subject of this section.
The issue is a substantial one, for the following reasons. We assume that English words are markedly polysemous, and we know that those numerous senses vary in context. Moreover, it is intuitively clear that speakers do not vary the meaning with deliberate thought, and that the contextual variation is not signalled in the utterance.
There is therefore a great deal of variation, but no clear explanation of it.
Several things make the issues to be discussed here important, as well as extensive.
First, the issues affect the long-running and unresolved controversies about polysemy and monosemy. Second, they relate to the complexity of the account of sense elements given in dictionaries. In SOED, set verb1 has branches I to XI, with 79 numbered senses, and 54 lettered subsenses, which are further divided into subsenses separat- ed by semicolons, and those are divided into sense elements separated by commas.
Moreover, that complexity illustrates some fundamental features of semantic struc- ture: the hierarchic nature of the sense variations enters into the network, since each of the 79 numbered senses and of the multiple subsenses has its own set of antonyms,
synonyms and so on; and remarkably, all that variation of elements can occur without the core of the sense being destroyed, since all of the senses of set verb1 are variations of ‘put in place’. (Compare Ruhl (2002), arguing that the verb break is monosemous, and the emphasis made by Sign-Based Linguistics (e.g. Reid and others, 2002) that every sign has an invariant meaning.) Finally, from the hearer’s point of view, the var- iation is totally controlled by linguistic and situational context: the meaning exists in context; any meaning stated without context is an unreal abstraction.
The previous paragraph describes senses from the system aspect; from the other aspects, the situation is very different. It seems extremely unlikely that hearers who could understand any of the SOED’s variations on set have all of the variations in- dexed in their mind, explicitly and hierarchically, as learned units. Rather, they will have a much more limited set, which they use to infer the relevant exact meaning in context. Similarly, it is intuitively clear that speakers do not have such exact meanings in mind consciously, and presumably even their unconscious formulation is com- monly what linguists would regard as underspecified. Hence the need for a semantic system by which hearers can cope with sense variation.
The section will treat variation in sense structure as having three forms: select- ing among existing possible elements (subsection .2), specifying vague elements (subsection .3), and changing existing elements to something different in nature (subsection .4).
5.3.4.2 Selecting among possible elements The expectedness dimension in use
Expected elements are those which the hearer expects, from his knowledge of the sense, to be invoked when the word is used. Speakers, however, having a meaning of their own which they want to invoke, may not follow hearers’ expectations. I will il- lustrate the extent to which those expectations are fulfilled, and how the expectedness dimension operates as words are used, from flare as an intransitive verb, using the literal sense, given as example (4).
(4) “Burn with a spreading, unsteady flame; blaze or glow (as) with flame”.
(SOED, <4>) I take the two phrases, “Burn with… flame” and “Blaze… with flame”, as defining a single sense, united by the synonyms “burn”, “blaze” and “glow”. (The variation in wording seems intended to suggest the variety of possible elements; “blaze” suggests
‘up’ and ‘high’.) Taken that way, the sense structure on the expectedness dimension seems to be as follows. (In the words below for the sense elements (in single quotation marks), differences in part of speech and semantic class are not significant; mostly, the distinction between linguistic and cognitive meaning is ignored also.)
– Necessary elements, shared by the two phrases for flare <4>: ‘burning’ and ‘flame’
(stated explicitly); ‘shining’ (implied by ‘burning’ and ‘flame’).
– Expected elements: ‘spreading’, ‘unsteady’ (mentioned explicitly, but the dif- ference between the two alternatives shows them to be not necessary); ‘bright’
(seemingly necessary, but inhibited by “glow”).
– Likely elements: ‘brief’ (implied by “blaze”).
– Possible elements: ‘suddenness’ and ‘speed’ (implied by “blaze”); ‘upward move- ment’ (suggested by “blaze”); ‘not very bright’ (implied by “glow”).
That distribution of elements within the descriptive meaning of flare is illustrated in examples (5a–f). They are taken from COCA; they were the first half dozen instances of “flared” in this sense, when the corpus was consulted on 10th March, 2014.
(5) a. “…a flamb[é] dessert that flared up nicely when Vern … touched a match to it.”
b. “The fire beneath the burners flared high for an instant.”
c. “Lanterns and small fires flared.” (Description of an outdoor scene.) d. “Way out there in space that speck of light flared one more time and went
out.”
e. “At that very second, the lights flared on. We all blinked in the onslaught of brightness.”
f. “A light flared to life in the other room.”
The distribution of the elements is analysed in Table 1; the columns represent the ex- amples, (a) to (f); the rows represent the sense elements; “Y” stands for “Yes”, meaning
“Yes, this sense element is present.” Thus the first row below the headings indicates that the necessary element ‘shine’ occurs in all of the examples, (a) to (f). The mor- phological form given (e.g. “shine” or “shining”) is fairly arbitrary; as will be argued later, these fundamental meaning elements are not equivalent to words. The table shows each element’s level of expectedness. ‘Shine’, ‘burn’ and ‘flame’ are necessary Table 1. Sense elements of flare, in the sample sentences
Level of expectedness Sense element (a) (b) (c) (d) (e) (f)
Necessary ‘shine’ Y Y Y Y Y Y
‘burn’ Y Y Y (Y) (Y) (Y)
‘flame’ Y Y Y (Y) (Y) (Y)
Expected ‘spreading’ Y
‘bright’ Y Y Y Y Y Y
‘unsteady’ Y
Likely ‘brief’ Y Y Y Y
Possible ‘not bright’
‘sudden’ Y Y
‘speed’ Y
‘upward’ Y Y
elements in flare, and occur in all uses. ‘Spreading’, ‘bright’, and ‘unsteady’ are expect- ed, since they distinguish flare from synonyms like shine and burn.
The column for example (c) thus represents the analysis that in “Lanterns…
flared”, flared has the necessary elements ‘shine’, ‘burn’ and ‘flame’, and the expected element ‘bright’, and the likely element ‘brief’; and it indicates that the other elements are not invoked. Thus “Lanterns… flared,” means that the lanterns shone as if burning briefly with a bright flame. (‘Not bright’ is not invoked by any of the uses in this ran- dom selection of quotations.)
The rows for ‘burning’ and ‘flame’ show the “Y” for “Present” in brackets, for examples (d), (e) and (f). That is because those examples refer to lights, which do not burn or have flames – the sense varies between the lights frame and the flames frame;
in the real situation depicted, those elements were not present, but they are present in the linguistic meaning by metaphor, which is indicated in the definition by “as” in
“glow (as) with flame”.
Flare also illustrates again the odd fact that words can have incompatible ele- ments: flare as ‘blaze’ and ‘burn’ entails the expected element, ‘bright’; but as ‘glow’, it entails the merely possible element, ‘not bright’.
The possible sense elements of flare form a network, which can be diagrammed as a semantic map, as in Diagrams 7 and 8. Diagram 7 shows the elements with some of their relationships; for example, the top of the map represents the relationships that FLAMES BURN, and that they SHINE as they do so, and the left-hand side of the map indicates that the FLAME is often SPREADING and UNSTEADY. (In users’ mental net- work, there are other links, of course, as between BRIGHT and SUDDEN, and SUDDEN and SPEED, for example.)
Diagram 8 is a semantic map showing how the uses in (5) variously invoke the elements of flare’s semantic network. Thus the box labelled “(a)” at the top left corner represents the sense of flare in (5a), “…a flamb[é] dessert that flared up nicely…”; it adds the conceptual elements SPREADING, UNSTEADY, SUDDEN, and UPWARD to the basic elements of BURNING BRIEFLY with a BRIGHT FLAME. Note that the combination of elements for the uses does not always correspond to a subsense of the word as
Necessary elements
SPREADING UNSTEADY
NOT BRIGHT SUDDEN SPEED
BRIEF BRIGHT
BURN SHINE
FLAME
UPWARD Expected
elements Likely elements Possible elements
Diagram 7. Flare, as network of sense elements