5.2.1.1 General
This section deals with the structure of the meaning of words, or, strictly, meaning of lexical units, including idioms as well as words. It is assumed that lexical units often have several senses, semantic units of the kind identified by successive numbers in
SOED entries, for example. These are senses of words as used in utterances – lexical senses. As noted in the previous chapter, those senses are distinct from sublexical senses, which are treated in a following section. A word with both descriptive and emotive meaning (for example) is treated as having a combination of two senses, since the different meaning types have different types of structure.
Grammatical meanings are not discussed here; they do not have constituent structure, and are here distinguished from senses. The syntagmatic relations between word senses, which structure groups, will be dealt with in the later chapters on figure and group structure, rather than being treated here as constituents of word senses.
That goes with treating hierarchic structure separately from network structure, for the sake of clear exposition.
5.2.1.2 Problems in identifying senses
I define a sense as the content (descriptive or nondescriptive) by which people repre- sent the meaning of the word; that is, the image schemas, concepts and so on which come to people’s minds as the meaning of a word they encounter. (That is based on Barsalou’s definition (1992b, p. 31) of concepts.) But as we begin to think closely about senses, difficulties arise. It is commonly assumed, in the literature on polysemy for example, that senses, like words, are quite distinct – we can count them, after all.
If they are distinct, senses are units, and there must be fairly well defined boundaries between them.
However, the nature of their unity is elusive It is natural to think of it as arising from the things they denote in the real world: one thing, one sense. If we look for the unity by taking the sense as a single concept, we face the problem that we want to define the concept, but that entails other concepts, and so on indefinitely – or in a circle. If we try to work by linguistic senses instead of cognitive concepts – on the principle of one sense to one linguistic referent, for example – we have the problem illustrated by serial verbs. Compare “let go” and “release”, “let fall” and “drop”, as to whether each pair has one sense or two; and in many languages, asking your dog to
“fetch” the newspaper would be expressed as asking him to “go pick-up carry bring give” the newspaper: the apparent unity of the real event denoted by a sense is arbi- trary. However, we can see unity in the singleness of a sense’s syntagmatic function, and in its unifying central node (see below, §5.2.3).
There is also a problem with boundaries. Elements or features cannot be in two senses at once, since that would require copies of the feature (a copy in one sense and another copy in the other sense), which would have to be identical, which would require verification from an original, and so on indefinitely. The spatial metaphor entailed by “contain”, “in”, and so on is clearly unsatisfactory. See Cruse (2000) for further discussion of the issues.
We must accept that sense elements make a network, without boundary lines, and without senses defined by a unique combination of elements. That network view has psycholinguistic support from Fortescue (2009, 2010), for example, based on work by Pulvermüller in particular. Senses are represented in the brain by linked bundles
of neurons, which form columns in the cortex. The minimum linguistic unit is repre- sented by a group of physical columns, with multiple links to other columns, so that each minimal unit can be accessed by many others. A minimal sense is “represented”
by a mini-network of those minimal cortical units, with a “word column” as the focal centre of the pattern. Activation of a linked unit is not automatic; the stimulation must reach a threshold strength, and certain forms of stimulus will inhibit the con- nection, not activate it. The sense, then, is a gestalt–like pattern of activation. It resem- bles the figures displayed by a digital watch. The different segments of the display can be activated in different combinations, to make “1”, “2” and so on; it is the pattern of activation, not the segments themselves, that constitutes the meaningful figure.
From the “unity” and “boundaries” point of view, a sense is like an island which is a peak in a submerged mountain range; the peak makes a distinct unit, and the island seems isolated, but the base overlaps those of other mountains. Accordingly, it is appropriate to use SOED’s numbered senses as starting points for discussion, since their sharpness of distinction from other numbered senses has some theoretical jus- tification, and their variety of subsenses separated by commas and semicolons gives a useful indication of their radiating network links.
5.2.2 Word senses as a structure of meaning types 5.2.2.1 Synchronic structure
As indicated in the chapter on elements of meaning, words commonly have senses of different meaning types. For example, brave as adjective is given in SOED as follows.
(I have adjusted graphology and abbreviations a little.)
<1> “Courageous, daring; intrepid; able to face and withstand danger or pain. L15”
<2> “Splendid; spectacular; showy; handsome. Now literary. M16”
<3> “Of excellent quality, admirable; fine, highly pleasing. Frequently an exclama- tion of approval. Now archaic or dialectal. L16”
Sense <1> has descriptive meaning, with concepts such as COURAGEOUS, WITHSTAND, DANGER, and PAIN. Sense <2> has a semantically more complex structure: the attitudi- nal meaning of approval and the emotive meaning of pleasure (implicit in the choice of synonyms); the social meaning of being literary (stated explicitly); and descriptive meaning, with the concept STRIKING IN APPEARANCE, which I take to be common to the various affective synonyms. Sense <3> has the same attitudinal meaning and emo- tive meaning as sense <2>; it has the social meanings of being dialectical in some uses, and archaic in other use. I believe that it now has no descriptive meaning, although the synonyms given by SOED suggest that it does; its lack of descriptive meaning is clear from the Oxford English Dictionary (“OED”), which notes that in this sense the word is used “loosely, as a general epithet of admiration or praise”, as in Shakespeare’s
“O that’s a braue man, hee writes braue verses, speakes braue words” (As You Like It 1623). SOED’s note, “Frequently an exclamation of approval” shows that sense <3>
also often has grammatical meaning, consisting of the instruction to the hearer that the one word is to be taken as a complete utterance, a “speech act” of exclamation.
The word, with its three senses, is semantically a network: there are many dif- ferent elements of meaning, related more or less closely (in various ways, not in a continuum), and with some overlap, as in the occurrence of EXCELLENCE, explicit- ly or implicitly, in all three senses; the different meaning types constitute different dimensions of the network. The broad structure of that network – its structure of meaning types – can be set out usefully as a semantic map (Haspelmath 2003), as in Diagram 1.1 It sets out what is sometimes called “conceptual space”.
The map is to be understood as follows. Each box represents one sense of the word, with its contents representing the constituent meaning types; the sense is iden- tified in italics just above the box. Thus, brave <1> is shown as having only descriptive meaning; brave <2> shares descriptive meaning with brave <1>, and shares attitudi- nal, emotive and social meaning with brave <3>, but excludes grammatical meaning.
(Other features of semantic maps will appear in later sections.) For sense <3>, the third subsense is represented – the exclamation of approval – and labelled with “(iii)”.
(The social meanings are simplified in the diagram.) The network nature could be emphasised by linking all the senses with lines; but that has not been done, to keep the diagram clear.
1. To Haspelmath, semantic maps are only of grammatical meaning, not of content.
brave <3> (iii) brave <2>
brave <1>
Descriptive meaning (‘high quality’)
Attitudinal meaning (approval)
Social meaning (archaism) Grammatical
meaning (exclamation)
Emotive meaning (pleasure)
Diagram 1. Meaning types in the three senses of brave
5.2.2.2 Diachronic development of the structure
There is no obvious reason why sense <1> of brave is only descriptive, while sens- es <2> and <3> have affective and social meaning as well. That apparently arbitrary connection between meaning types can be understood through history – as can the sequence of the three senses in SOED.
The most common way in which word meaning types have developed is that a sense begins with descriptive meaning and later acquires other types of meaning.
Speakers associate with the word the feeling they have when using it, or the feeling they have for the referent; that association becomes conventionalised as standard for that use, and the feeling becomes established as a sense. The same process applies to attitudinal meaning, which often develops at about the same time as emotive mean- ing. Social meaning evidently is also acquired by association, but the association is with the situation in which we hear the word, rather than with our own use of it.
‘Admirable’ is an archaic or dialectal element of brave <3> because we meet it only in old literature or as an element in certain dialects.
I will illustrate the historical development of word meaning a little further (ex- panding the analysis in Chapter 4, §4.4.1), to highlight the importance and semi- autonomy of the different types of meaning, and to demonstrate the relevance of the analysis of semantic structure given in this book to the understanding of word history.
The etymology of brave given by SOED begins with the Greek barbaros, which meant
‘non-Greek, foreign’. Latin took it over, then Italian, and Spanish, in which language it meant ‘savage’, and by extension ‘courageous’. Then French “borrowed” it, and then English, in the 15th century, in sense <1>. By the 16th century it was being used in- creasingly with its quality dimension changed by loss of ‘courage’ and addition of ‘in appearance’ (“showy, handsome” in SOED’s words); those changes shifted it towards the general end of that dimension. There was evidently a social consensus on approv- ing what looks good, so approving attitude became a standard association with the descriptive meaning. Brave had thus acquired sense <2>, which I will paraphrase as
‘of good quality in appearance’, while sense <1> remained in use (‘of good quality in courage’). By the 17th century, that new sense was often used with the ‘good’ element shifted towards the vague end of the specificity dimension: the descriptive meaning was simply ‘of good quality’. The attitudinal meaning of approval had strengthened in this use, which added the affective meaning of pleasure (“highly pleasing” in SOED);
the word was thus often used in the Expressive function (“Frequently an exclama- tion”). Brave now had sense <3>. In the centuries since then, this sense has dropped out of standard use, remaining current only in certain dialects; therefore, it has the social meaning of being either archaic or dialectal. (Uses as noun and adverb devel- oped from the adjective quickly, in the 16th century; the verb was borrowed directly from French.)
Those developments are represented in Diagram 2. It uses the conventions of the diagrams in Chapter 4, but adds arrows to highlight historical change, and adds text between the boxes to explain the changes.
The pattern of semantic development is shown in Diagram 3, which is a dia- chronic version of the semantic map given in Diagram 1. It has the same entries, in the same pattern; but there are no boxes, because the map is not illustrating different senses. The lines link entries that are related historically, as well as being similar, and the arrowheads indicate the direction of change.
5.2.2.3 Independence of the meaning types
Words’ affective, attitudinal and social meanings are relatively independent of the de- scriptive meaning, which is evident in several ways. The types occur on their own:
brave <1> has descriptive meaning alone; brave <3> as exclamation consists of af- fective meaning alone, as do frightful, horrible, lovely, nice and so on; hello and please have social meaning alone. When types do combine, the link between them is linguis- tically arbitrary; ‘intelligent’, as descriptive meaning, combines with favourable attitu- dinal meaning in quick-witted and shrewd, but combines with unfavourable meaning in crafty and cunning.
There is one way, however, in which a meaning type may be dependent on anoth- er, seen in the historical development of word senses. Attitudinal meanings develop
Of good quality, in courage
Approving attitude is added.
Feeling is added
to attitude [Pleasure]
[Approval]
Descriptive meaning weakens and disappears
descriptive[No elements.]
[Approval]
Of good quality, in appearance
‘In courage’ is replaced by ‘in appearance’.
<1>
“Courageous, daring.”
15th C. Early 16th C.[?]. Late 16th C.
<2>
“Showy, handsome”
<3>
“Highly pleasing”
Diagram 2. Historical changes in types of meaning, in brave Descriptive
meaning Attitudinal
meaning Emotive
meaning
Social meaning
Grammatical meaning Diagram 3. Historical development of meaning types
from abstract senses, not from concrete ones, because it is only about abstract quali- ties that there is enough social consensus to establish an attitude as a standard part of a word sense. Similarly, affective meaning seems to depend on a previous attitudinal meaning.
5.2.2.4 Discussion: Word meanings as a structure of meaning types
The word meanings given in dictionaries such as SOED seem to present no more than the descriptive meaning, since they are nearly all worded in the same abstract descriptive way. (I am putting aside here the distinction between linguistic meaning and cognitive or “encyclopaedic” meaning.) However, the definitions do testify to the analysis of meaning I have been giving. First, synonyms are given carefully and sys- tematically, evidently to provide words with the same descriptive meaning but differ- ent affective or attitudinal meaning; that shared descriptive meaning (never identified directly, unfortunately) represents the necessary meaning (Chapter 4, §4.3.9). Social meaning is not given descriptively as meaning, but is sometimes indicated by tags, such as “coarse slang”. (SOED gives very few; COBUILD (2001) gives many more.) Grammatical meaning is given as part of speech, which does not give the nature of that meaning clearly, but at least it is done regularly. Lexicographical practice thus confirms the reality and importance of what I have called meaning types, although it accounts for them differently. One implication of the argument of this book is that lexicographers should change their treatment of meaning.
The distinction between descriptive, emotive and attitudinal meanings as struc- tural elements is supported neurolinguistically by Fortescue (2010, pp. 195–197). As noted in §5.2.1.2, words are embodied in the brain by bundles of neurons forming columns which run vertically in the cortex. The neurons have branches or “affor- dances” running horizontally from different levels of the column; the neurons for descriptive meaning run from the mid-level to lower and middle layers of the cortex, and those for grammatical meanings run from the top of the column to the upper lay- er of the cortex and to greater distances. At the bottom of the column are affordances that lead downwards to the limbic system of the brain, serving emotive and attitudinal functions. (Fortescue does not use my terms for meaning types.)
5.2.2.5 Conclusion: Word meanings as a structure of meaning types
The senses of different types – descriptive, affective, attitudinal and social – are par- tially interdependent; but the nature of the relationships varies, and their occurrence in particular lexical meanings is partly arbitrary. This part of the network is quite loosely structured.
5.2.3 Structure within descriptive senses 5.2.3.1 Basic structure of descriptive meaning
In discussing the types of meaning in Chapter 4, I described descriptive meanings as being structured in three parts, with a central node, percepts linked to it, and concepts linking them to wider meaning. Schreuder and Flores d’Arcais (1989) explain that structure as follows, using the meaning of coffee as their example. The meaning ‘coffee’
has, first, a minimal node, which by itself has no content. That content is provided by
“perceptual elements”, such as colour, smell, and taste (using the concrete concepts BLACK, LIQUID, etc.), and by “functional elements” (using abstract concepts giving such information as how coffee is produced and where it is grown). Those elements of the network are activated together or independently in different general situations, such as those of experience, thought, and speech, and in different speech situations, such those of referential and descriptive use (Chapter 4, §4.7.3). What Schreuder and Flores d’Arcais call the central “node” can also be thought of as a link, since it has function, not content of its own.
This structure explains much of the difference between the different senses of a word. For example, fetter <1> is “A chain… for the feet…” and fetter <2> is “Anything that confines…”; the word is linked to the perceptual elements ‘thing’ and ‘chain’, and to the functional elements ‘confine the feet’; sense <1> activates them all, but sense
<2> activates only the functional ones.
5.2.3.2 Structure of descriptive elements
To Schreuder and Flores d’Arcais (1989), the minimal central node of a sense is the access point for the whole sense in comprehension, and for thinking of the word out of context; it links the word form with meaning, and thence with knowledge. When it is activated out of context, all of the various perceptual and functional elements become accessible. When it is activated in context, their activation is selective, not automatic. In descriptive use (e.g. “It ain’t got the real coffee taste” – BNC), many of the elements will be accessed, and brought into the hearer meaning. In referential use, when a word is used simply as a name for the referent (e.g. “Pass me the coffee”), all of the descriptive and linguistic nodes are bypassed – the outer nodes – with the referent in cognition being accessed instead, rather as pronouns and gesturing make us bypass descriptive meaning, accessing the referent in the previous sentence or the physical situation.
Accessing the sense in production – that is, when we find a word to express the meaning being formulated – is quite different. Psycholinguistic studies of priming (e.g. Mcnamara 2003) show that word senses can be accessed from a related meaning of any type, and common experience confirms that. We can bring a word and its sense to mind in searching for an antonym or synonym, for a more or less general word, or for other technical senses. (Access can also be through nondescriptive elements, such as slang value or affective meaning.)