6
6 Semantics: Word and sentence meaning
6.1
6.2
6.3
6.4
Branches and boundaries of semantics
Types and facets of meaning
Structural semantics: Meaning structures in the vocabulary
Cognitive semantics: Prototypes and metaphors
Semantics (Greek semain- = to mean) is the only branch of linguistics
which is exclusively concerned with meaning. Semantics studies the
meaning or meaning potential of various kinds of expressions: words,
phrases, and sentences. This chapter is mainly confined to the study of
word meaning (lexical semantics; lexicology). Research in lexical semantics addresses the following questions:
■ How can the concept of meaning be elucidated, including the relation
between meaning and external reality?
■ What are appropriate tools for analysing and describing meanings?
■ What kinds of semantic structures exist within the vocabulary (or:
lexicon) of a language?
focus on word
meaning
(lexicology)
These semantic structures are uncovered by describing recurrent semantic relations between the words, more exactly the lexemes, of a language
(e. g. relations such as near-equivalence or contrasts in meaning). Lexical
semantics proceeds from the assumption that words are symbols, i. e.
signs expressing an arbitrary relation between a form and its meaning(s).
This relation is considered to be exclusively a matter of convention (see
chapter 1 on the model of the linguistic sign proposed by Ferdinand de
Saussure).
6.1 | Branches and boundaries of semantics
Semasiology (form → meaning): Studies in semantics usually start out
from a given form and ask for its meaning, i. e. move from signifier (signifiant) to signified (signifié). This direction of research is also most relevant to non-linguists: Whenever we consult a dictionary, we are looking
for an answer to the question “What is the meaning of X?”. The branch of
semantics which adopts this approach is called semasiology (science of
meanings), a concept which originally covered all of semantics. It was
only in the 20th century that the term semantics (introduced by Michel
Bréal) replaced the term semasiology.
Onomasiology (meaning → form): The opposite way of studying meaning is called onomasiology (science of names; from Greek onomaz = to
name). It proceeds from a given meaning to the forms that express it.
Whenever we consult a dictionary of synonyms (thesaurus) such as Roget’s Thesaurus of English Words and Phrases, we are adopting an onomaJ.B. Metzler © Springer-Verlag GmbH Deutschland, ein Teil von Springer Nature, 2020
B. Kortmann, English Linguistics, https://doi.org/10.1007/978-3-476-05678-8_6
dictionary
thesaurus
143
6
Semantics: Word and sentence meaning
siological approach: We want to find out which word(s) can be used to
express a given concept. Take, for example, the concept – or lexical field
(see section 6.3) – of killing (German töten): kill, murder, slay, slaughter,
butcher, massacre, and assassinate are words which can be used to translate the concept expressed by töten.
Paradigmatic vs. syntagmatic semantics: The lexical (or: semantic)
field of killing serves as a useful example for illustrating another crucial
distinction, which ultimately derives from an important dichotomy proposed by Ferdinand de Saussure: The contrast between two distinct kinds
of relations contracted by every element in a language: (paradigmatic)
relations of choice, and (syntagmatic) relations of combination (see chapter 1.3.1). These relations are also relevant to semantics.
Semantic relations between lexical alternatives are the focus of paradigmatic semantics. This includes how members of a lexical field can be
replaced by other members of the field, especially if there are extensive
similarities between their meanings (i. e. if these words are synonymous).
Syntagmatic semantics, on the other hand, is concerned with questions
such as the following: Which of the above-mentioned lexical alternatives
is appropriate in a given sentence (e. g. kill as opposed to murder or assassinate)? Kill is the most general term, murder implies the intentional
killing of a human being, assassinate relates to the killing of an important
person (usually a politician). For this reason, only (1) is acceptable, while
(2) is odd:
(1)
(2)
President X was assassinated last night.
?Many innocent villagers were assassinated last night.
Selection restrictions: In this comparison of verbs of killing, there are
even more semantic restrictions when we look at the direct objects which
can be combined with each verb: kill has fewer restrictions than murder,
which in turn has fewer restrictions than assassinate. Such restrictions on
possible combinations of meanings, so-called selection restrictions, are
sometimes very wide-ranging. In some cases, a given lexeme can only be
combined with very few other lexemes. Extreme examples are provided
by many words which are rarely used: It is often possible to predict with
which other lexemes such rare words are likely to occur in a sentence.
Collocations: A popular example of such typical combinations of
words, so-called collocations, are the various expressions for groups of
animals in (3):
(3)
a
Fregean Principle
144
Ï
Ô
Ô
Ì
Ô
Ô
Ó
flock
gaggle
pack
pride
shoal
¸
Ô
Ô
˝
Ô
Ô
˛
of
Ï sheep / goats / birds
Ô
geese
Ô
wolves / hounds
Ì
Ô
lions
Ô
fish
Ó
¸
Ô
Ô
˝
Ô
Ô
˛
Compositionality: Syntagmatic semantics is not only concerned with possible combinations of particular words (such as those discussed in (1) to
(3)), it also deals with the meaning of complex linguistic expressions,
Types and facets of meaning
including sentences. The crucial principle that determines the meaning of
complex expressions is the principle of compositionality, which stipulates
that the meaning of a complex expression in natural language depends on
(and can be reconstructed from) the meaning of its parts and the syntactic relations holding between these parts. This important principle of sentence semantics is often called Frege’s or the Fregean Principle, since it is
commonly attributed to the German philosopher and mathematician
Gottlob Frege (1848–1925). The principle of compositionality is held to
ensure that we can understand the countless sentences we encounter
every day, even though we have never heard them before.
There are limits to compositionality, however. Consider, for example,
idioms such as to kick the bucket or German den Löffel abgeben. Their
meaning (here ‘to die’) cannot – or can only in part – be reconstructed
from the meanings of their component parts; thus, for idioms like these,
the connection between form and meaning tends to be just as arbitrary
and conventional as it is for most single words. But even where the principle of compositionality does apply, it does not guarantee that one really
understands what the speaker or author means with a particular utterance, at least in those cases where the intended meaning goes beyond
what is literally said.
Semantics vs. pragmatics: The distinction between what is said and
what is meant, i. e. between (literal) sentence meaning and (intended)
utterance meaning, is closely linked to the distinction between semantics
and pragmatics, even though the correlation is not perfect (see the detailed discussion in chapter 7). Pragmatics studies language use (parole),
focusing on both the linguistic and the non-linguistic context of utterances, as well as speakers’ utterance-related intentions. Thus, a central
– for many the central – aspect of pragmatics is its concern with principles that allow us to infer what is meant from what is said in a particular
context. At the heart of pragmatics are questions such as ‘What does the
speaker mean by uttering X?’ and ‘Why are hearers usually able to recognize speakers’ intention(s) without great difficulty?’.
In semantics, on the other hand, context is almost completely ignored,
and speaker intention is entirely left out of consideration. Thus, the division of tasks between semantics and pragmatics may roughly be characterised as follows: semantics deals with the meanings or the meaning
potential of expressions out of context (i. e. context-invariant, speakerindependent meaning), whereas pragmatics deals with the meanings of
expressions (mainly utterances) in a particular context (i. e. contextsensitive, speaker-dependent meaning).
6.2
idioms
pragmatics: meaning in context
semantics: meaning out of context
6.2 | Types and facets of meaning
What we typically have in mind when talking about word meanings are
the meanings of lexemes that belong to one of the four lexical word
classes (nouns, verbs, adjectives, adverbs).
Lexical vs. grammatical meaning: Lexical meaning contrasts with the
grammatical
meaning =
abstract
145
6
far fewer grammatical meanings
lexical meaning
expressive and
social meaning
146
Semantics: Word and sentence meaning
grammatical meaning of function words (e. g. pronouns, prepositions,
conjunctions; see chapter 3.1 on auto- and synsemantic words). Grammatical meaning also includes the meaning of inflectional affixes and the
semantic roles (e. g. agent, patient) associated with grammatical relations. The differences between grammatical and lexical meaning are only
gradual. Grammatical meaning in general is abstract; just think of the
meanings of case or tense morphemes, or of the marking of (in-)definiteness with the help of a and the. Lexical meaning, on the other hand, is
frequently far more concrete. In this respect, grammatical meaning contrasts particularly strongly with the lexical meaning of those nouns that
denote concrete countable entities. Note, however, that the (lexical)
meanings of abstract nouns such as condition, cause, or concession are no
less abstract than the (grammatical) meanings of adverbial subordinators
such as if, because, or although. The above examples of grammatical and
lexical meanings can thus be located at opposite ends of a continuum
from abstract to concrete concepts. The meanings of personal pronouns
and spatial prepositions tend to be located even further towards the middle of such a continuum (and thus towards the transitional area between
grammatical and lexical meaning).
Another essential difference between lexical and grammatical meaning
relates to the fact that the number of grammatical meanings encoded in
languages is comparatively small and – even from a cross-linguistic point
of view – probably also finite, whereas there are an infinite number of
potential lexical meanings. It is therefore much easier to provide an overview of the domain of grammatical meanings. Not surprisingly, regular
processes of meaning change – both within a single language and across
languages – have been identified primarily in the domain of grammatical
meanings. Meaning changes in lexical words, on the other hand, are
clearly more idiosyncratic, and cannot be captured with the help of a
relatively small number of general principles of the type discovered for
function words. Issues of historical (or: diachronic) semantics will be
discussed in chapter 9.
Descriptive vs. expressive vs. social meaning: In what follows, we will
largely focus on lexical meaning, more precisely on the descriptive (or:
cognitive) meaning of lexical words. Thus, special emphasis will be
placed on the representative function of language (cf. the various functions of language described in chapter 1), i. e. on those aspects of meaning that allow us to describe the world. Expressive and social meanings
will not be dealt with in greater detail. The following examples have to
suffice:
■ the exclusively expressive meaning of gosh!, and the differences between father and daddy, policeman and cop(per), or very and jolly with
regard to their expressive meaning;
■ the exclusively social meaning of welcome and farewell words such as
hello and goodbye, forms of address like sir and madam with their
social meaning component, and the differences between forms of address like pal, mate, and love (as used in grocer’s shops in England:
What can I do for you, love?) with regard to their (expressive and) social meaning.
Types and facets of meaning
Generally speaking, the interpersonal function of language is relegated to
the research periphery in semantics: Semanticists rarely devote particular
attention to those aspects of meaning that enable us to express feelings,
points of view, and speaker judgments (i. e. expressive meaning), or
which signal and establish social relationships (i. e. social meaning).
Descriptive meaning: But what exactly is the descriptive meaning of a
lexical word? To answer this question, we will turn to three central pairs
of concepts used in semantics which more or less overlap: sense – reference, intension – extension, and connotation – denotation. The firstmentioned terms in these three pairs (sense, intension, and connotation)
relate to the conceptual side of meaning and to (language-internal) definitions of meaning.
Reference: By contrast, the three contrasting terms (reference, extension, denotation) relate to extra-linguistic reality, i. e. to the relation between language and the world. The term reference, for example, designates the relation between entities in the external world and the words
which are used to refer to these entities (e. g. people, objects, events,
places, points in time, etc.). The referent is the entity referred to (“picked
out”) by an expression in a particular context.
(4)
a.
b.
c.
6.2
extra-linguistic
reality
Take the bottle and put it in the dustbin.
She took a bottle and put it in a dustbin.
A bottle is not a dustbin.
Both (4a) and (4b) deal with a particular bottle and a particular dustbin.
The only difference is that in (4a) the referents of the bottle and the dustbin are accessible to the hearer. In both cases, the referents of the/a bottle
and the/a dustbin vary from utterance to utterance. Matters are different
in the case of (4c): here, a bottle does not refer to a particular bottle, nor
does a dustbin refer to a particular dustbin; both noun phrases are thus
used in a non-referring sense. But even though both noun phrases in (4c)
lack a referent, they still have an extension.
Extension – denotation: The term extension designates the class of objects to which a linguistic expression can be applied, i. e. the class of its
potential referents (in (4c) the class of all bottles and the class of all dustbins). A referent of a linguistic expression is always a member (or subset)
of the class of objects that constitutes the word’s extension.
The term denotation is frequently used synonymously with extension.
Both terms are sometimes understood in a broader sense, covering not
only the relation between nouns or noun phrases and groups of individuals or objects, but also the link between words belonging to other word
classes and the phenomena they relate to. Thus verbs denote situations,
adjectives denote properties of individuals and objects, and adverbs denote properties of situations.
Sense (vs. reference): The sense of an expression is its descriptive
meaning, more exactly meaning we know in virtue of our knowledge of
language, which – in contrast to reference – is independent of a particular
utterance and the situational context in which the utterance was made.
The distinction between sense and reference was introduced by the Ger-
147
6
propositional
content
148
Semantics: Word and sentence meaning
man philosopher Gottlob Frege. It is not difficult to see why such a distinction is useful. For one thing, linguistic expressions with different
meanings (“senses”) may very well have the same referent(s). Just think
of the noun phrases the Leader of the Conservative Party and the Prime
Minister of Great Britain, which differ in sense, but not necessarily in
reference: The phrase the Leader of the Conservative Party may refer to the
same person as the phrase the Prime Minister of Great Britain.
Intension (vs. extension): Similar observations apply to the capital of
Prussia, the capital of the Third Reich, and the capital of Germany; all of
these phrases refer to Berlin. This example also illustrates that the referent of a linguistic expression may change, while its meaning remains the
same: in 1992 Bonn was still the capital of Germany, today the capital of
Germany is Berlin. The relevance of the sense-reference distinction is also
brought home by words which lack a referent, but do have a sense. Cases
in point are unicorn and dragon. The sense of a linguistic expression
essentially consists of characteristic features, so-called semantic features,
which determine the class of entities it may be used to refer to, i. e. its
extension. These features are (typically binary) traits that describe essential aspects of the meaning of a word, and a bundle of such semantic
features, e. g. [+ human, – adult, + female] for girl, can only describe
the intension of a linguistic expression (more on this in sections 6.3.1 and
6.4). Note that so-called connotations are not part of the intension.
Connotation (vs. denotation): Connotations are typically secondary
meanings which can vary according to culture, region, social class, etc.
and which are often restricted to particular contexts. This does not mean,
however, that connotations are completely subjective associations which
different speakers connect with expressions on the basis of entirely different personal experiences. Connotations can be generalized to a certain
extent, they are part of the encyclopaedic meaning of a lexeme, i. e. meaning known in virtue of our knowledge of the world (as opposed to its
dictionary meaning, i. e. its descriptive meaning, the much more rigid
definition we find in dictionaries).
Sentence meaning – truth conditions: This chapter is concerned with
lexical semantics. For presenting the complete picture, though, it should
be mentioned that we can also speak of the descriptive meaning, denotation, or extension of phrases and sentences. The descriptive meaning of a
sentence is called its propositional content and is a mental concept covering all situations the relevant sentence can potentially refer to. Sentences
can be true or false in a given situation, and the conditions under which
they are true are called the truth conditions of a sentence. They specify
the conditions under which a given sentence can correctly be used to refer to a certain (kind or set of) situation(s) in the world.
The following sections deal with the two major approaches to the analysis of lexical meaning: the first approach investigates recurrent semantic
structures in the vocabulary (structural semantics; section 6.3), the second examines the relations holding between word meanings and our conceptual system (cognitive semantics; section 6.4).
Structural semantics: Meaning structures in the vocabulary
6.3
6.3 | Structural semantics: Meaning structures in the
vocabulary
A network of semantic relations: Even today, lexical semantics is still
committed to classical structuralist assumptions to a considerable extent
(see chapter 1). One of the principles that has proved particularly influential is the idea of language as a complex system of relations: Every
linguistic element is integrated into the system (langue) through a network of paradigmatic and syntagmatic relations; nothing happens outside
of the system. Applied to semantics, this view implies that word meaning
is to be treated as something relative, as a purely language-internal phenomenon. A word’s meaning, its sense, constitutes a node in a network
of semantic relations.
More precisely, the meaning of an expression is defined in part by
what it has in common with other expressions, but above all by what
distinguishes it from them (de Saussure speaks of the signe différentiel:
the meaning of a word is what it is not). Thus, if we want to grasp the full
meaning of a verb like march, we have to know how the manner of walking described by this expression differs from the manner of walking described by similar verbs like pace and stride. All of these terms are part of
an extensive network of motion verbs. Some verbs that belong to this
network, such as amble, saunter, and stroll, constitute a subclass of expressions which stand in a relation of oppositeness to march, pace, and
stride. The latter verbs denote a quick, determined manner of walking,
whereas the verbs in the former group denote a slow, aimless manner of
walking.
In sum, the principal goal of structural semantics is to show that the
vocabulary of a language is a structured whole in which nothing happens
in isolation and where various recurrent semantic structures can be identified. The two most important types of such structures (or: networks) are
lexical fields and lexical (or: sense) relations.
structuralism:
meaning as something entirely
language-internal
signe différentiel
major goal
6.3.1 | Lexical fields
Lexical (or: semantic) fields are groups of words which cover different or
partly overlapping areas within the same extralinguistic domain. Above
we already encountered three examples of lexical fields:
■ verbs of asking (ask, inquire, interrogate, question, wonder, etc.),
■ verbs of walking (walk, march, pace, amble, stroll, prance, sneak,
stagger, swagger, etc.), and
■ verbs of killing (kill, murder, assassinate, etc.).
Further examples include:
■ colour adjectives,
■ adjectives relating to mental abilities (intelligent, clever, smart, bright,
brilliant, brainy, stupid, dumb, silly, thick, dense, etc.),
■ different types of footwear (shoe, moccasin, clog, slipper, sandal,
trainer, boot, etc.),
149
6
Semantics: Word and sentence meaning
■
■
■
Jost Trier
criticism of lexical
fields as mosaics
badly spread
bread-and-butter
bundles of semantic features
150
legwear (trousers, dungarees, socks, stockings, tights, leggings, etc.),
teaching and research staff at universities (professor, reader, lecturer,
fellow, etc.), or
temporal conjunctions (when, as, while, after, since, etc.).
There is an infinite number of such lexical fields. The crucial idea behind
grouping lexemes by semantic similarity is the assumption that the meaning of a field member can only be fully determined and delimited with
reference to its semantic neighbours. From a diachronic point of view,
this means that any semantic change within a lexical field may affect all
members of the lexical field plus the intricate network of semantic relations holding between them. Such potential changes in lexical fields include the addition of a new word, the loss of a word, and a change in the
meaning of one or more of their members.
Traditional vs. modern conceptions of lexical fields: In fact, the theory of
lexical fields (‘Wortfeldtheorie’), which was developed by the linguist Jost
Trier in the 1930s, has its roots in the study of semantic change, and hence
in diachronic (or: historical) semantics. However, it did not take long for
the study of lexical fields to occupy a central place in synchronic word
semantics, even if some aspects of Trier’s account are in need of revision.
Pertinent criticism has been levelled, for example, at his conception of
lexical fields as mosaics, whose boundaries can be clearly delimited and
which do not have any gaps or overlaps. This ideal hardly exists:
■ category boundaries are often fuzzy (see also section 6.4); as a result
it may be difficult to determine which lexical field a word belongs to;
■ there are many examples of gaps in lexical fields (e. g. in English or
German, adjectives are missing which – in analogy to blind, deaf/taub,
or mute/stumm – denote the absence of the ability to smell or taste);
■ and, finally, in many cases there are more or less conspicuous meaning similarities (and thus overlaps) within a lexical field (e. g. intelligent, clever, smart).
The suggestion to compare a lexical field to a piece of bread which is
unevenly buttered (thicker in some places, thinner or not at all in other
places) thus seems to be much more useful than the mosaic comparison.
Much as the latter, however, this conception of lexical fields neglects the
fact that field members are related to one another along more than merely
two dimensions in most cases. For instance, pace may differ from stroll
with regard to the dimensions ‘speed’ and ‘purposefulness’; but these
dimensions are no longer sufficient if we want to describe the difference
between these two verbs and other members of the same lexical field
(e. g. stagger or trudge). Here, we need further dimensions such as ‘degree of body control’ or ‘degree of effort’.
Componential analysis: Componential analysis (or: feature analysis,
semantic decomposition) has proved to be a very useful tool for describing semantic similarities and differences between members of a lexical
field. In analogy to the conception of phonemes as bundles of distinctive
features (e. g. the phoneme /p/ as [+ consonant, – voiced, – nasal,
+ occlusion, + plosive]; see chapter 2.2.1), the meaning of a word is
Structural semantics: Meaning structures in the vocabulary
conceived of as a bundle of (ideally binary) semantic features or semes.
Consider, for example, girl [+ human, – adult, + female] in contrast to
boy [+ human, – adult, – female], or pace [+ quick, + purposeful] in contrast to stroll [– quick, – purposeful]. These features can roughly be
equated with the dimensions that structure a lexical field, but wherever
possible they should be chosen so as to allow a ‘yes (+) / no (–)’ characterization. The choice of relevant semantic features is to some extent
arbitrary, of course, and singling out useful features is more difficult for
some semantic areas than it is for others.
In general, the method of semantic decomposition becomes more and
more difficult to handle the more fine-grained the semantic analyses are
supposed to be (just think of the lexical field of motion verbs). In addition, it is open to debate whether there really is a limited, universally
valid inventory of semantic features relevant to the analysis of word
meanings. It also remains unclear what role semantic features play in
human categorization, i. e. whether semantic features are cognitively real
(more on this in section 6.4).
Structure of the mental lexicon: By contrast, the psychological reality
of lexical fields is indisputable. It can be shown that they are more than
simply a convenient theoretical construct, and that they do indeed fulfil
an important function in structuring the information stored in our mental
lexicon. Word association tests with ordinary people and especially with
people suffering from aphasia have clearly shown that there is a much
stronger psychological link between members of the same lexical field
than between members of different lexical fields.
This is hardly surprising: within a lexical field there is a much stronger
network of lexical (or: sense) relations. Such relations represent another
important principle structuring our mental lexicon. Of particular importance are the kinds of relations holding between red – blue – green – yellow, or Monday – Tuesday – Wednesday, etc. (relations of semantic incompatibility), and those relations that hold between word pairs such as
man – woman, husband – wife, hot – cold, buy – sell (relations of oppositeness of meaning) (see section 6.3.2). Apart from these paradigmatic
lexical relations and lexical fields, collocations also play an important role
in processing, storing, and retrieving lexical information. Examples of
collocations include syntagmatic lexical relations which exist, for example, between adjectives like blond, auburn, curly, wavy, or unruly, on the
one hand, and hair, on the other hand (see also section 6.1 above).
6.3
problems
lexical fields = psychologically real
part of the mental
lexicon
6.3.2 | Sense relations
Sense (or: lexical) relations) are specific semantic relations between
words. The five relations discussed in this section have in common that
they systematically occur in an infinite number of word pairs or word
groups (especially with lexical fields), and that they are all paradigmatic
relations, thus represent possibilities of choice between lexical alternatives (e. g. leggings in contrast to trousers, tights, or stockings).
Synonymy: The concept of synonymy is used to describe semantic
5 paradigmatic
sense relations
descriptive/cognitive synonyms
151
6
Semantics: Word and sentence meaning
equivalence or rather extensive semantic similarity between two or more
lexemes. The term is typically used in reference to the descriptive meaning of words (hence the term descriptive or cognitive synonymy). Synonyms thus have the same semantic features. However, most synonyms
differ with regard to their conditions of use: Descriptive synonyms may
be interchangeable in many, but not all, contexts.
In (5a), for example, it is impossible to replace deep by its synonym
profound; matters are different in (5b):
(5)
a.
b.
This river is very deep.
The incident made a deep impression on me.
Descriptive synonyms may differ with regard to their connotations (dog
– mongrel, cock – rooster, worker – employee, baby – neonate), with regard to stylistic level or register (begin – commence, buy – purchase, intoxicated – drunk – pissed), with regard to regional or social variety (e. g.
differences between American and British English), or with regard to
their collocations (e. g. a big/large house, but Big/?Large Brother is watching you).
total synonymy
Cases of total synonymy, i. e. of interchangeability in all contexts (e. g.
Apfelsine – Orange in German), are very rare. It is not difficult to see why.
A linguistic system which has (many) total synonyms
synonymy
would be uneconomic. Why should a language have
two (or more) lexemes with absolutely identical usage
descriptive
total
conditions? In fact, total synonymy between two words
is always only temporary: either one of the synonymous lexemes is lost, or the two items will be semantically differentiated,
Figure 6.1:
developing different usage conditions.
Types of synonymy
Opposites: Synonymy contrasts with antonymy, a term covering various types of semantic opposites (oppositeness).
either-or
Complementary antonymy: We speak of complementary or binary anrelationship
tonyms (or: complementaries) if there is an either-or relationship between
the two terms of a pair of semantic opposites, i. e. if the two antonyms
exhaust all possible options in a particular conceptual domain (e. g.
asleep – awake, dead – alive, live – die, pass – fail). In these cases, the
meaning of one lexeme is equivalent to the negation of the other lexeme.
narrow vs. wide
(Gradable) antonymy: Complementary antonymy is commonly conantonymy
trasted with gradable antonymy, where the two expressions involved
merely constitute opposite poles of a continuum. Alternative terms for
gradable antonyms include contraries, or simply antonyms. (Note, however, that the term antonymy can also be used in the wider sense of ‘oppositeness’.) Examples of gradable antonyms are hot – cold (notice the
various intermediate stages like warm – tepid – cool), broad – narrow,
large – small, and old – young (cf. also pairs of nouns like beginning –
end, war – peace). The great majority of gradable antonyms are pairs of
adjectives. Some of these pairs display a certain asymmetry in the sense
that one of the two contrasting lexemes can appear in more contexts than
the other. Thus, if we want to know a person’s age (How ___ are you?) or
the length of an object (How ___ is it?), we use old or long, respectively,
differences between descriptive
synonyms
152
Structural semantics: Meaning structures in the vocabulary
rather than young or short. The members of pairs like old – young, long
– short differ in markedness: the term with the wider range of uses is
called unmarked (old, long), the one with a more limited range marked
(young, short).
Relational opposites: A further type of antonyms are relational opposites (or: converses). They describe the same situation from different perspectives (e. g. teacher – pupil in sentences like John is Mary’s teacher vs.
Mary is John’s pupil). Further examples include pairs of deverbal nouns
in -er and -ee (e. g. employer – employee, examiner – examinee, interviewer – interviewee), comparative forms of adjectives (older – younger,
longer – shorter), pairs of verbs like give – take, buy – sell, rent – let, or
pairs of prepositions like above – below.
Directional oppositeness: The fourth type of antonymy, directional oppositeness (directional opposites or reverses), does not involve different
perspectives on the same situation, but rather a change of direction
(especially motion in different directions). Examples include open – shut,
push – pull, rise – fall, come – go, leave – return, (turn) right – (turn) left,
tie – untie, and button – unbutton.
Let us next turn to sense relations which involve hierarchies in the
vocabulary, i. e. super- and subordination.
Hyponymy: The term hyponym refers to words like rose, tulip, daisy,
and lily, which stand in a relationship of subordination to a more general
expression like flower. Conversely, the generic term flower is the superordinate or hyperonym (or: hypernym) of rose, tulip, daisy, and lily. Hyponyms have all semantic features of the hyperonym plus some additional
ones, which distinguish them from the hyperonym, on the one hand, and
from other hyponyms situated on the same hierarchical level, on the
other hand (consider, for example, the features distinguishing rose from
daisy, or daisy from lily). Hyponyms relating to the same hierarchical
level are called co-hyponyms or heteronyms.
Interestingly, one of the oldest methods of defining meanings is based
on the concept of hyponymy: According to this approach, we should first
identify the superordinate category (the so-called genus proximum, i. e.
the hyperonym), and then single out the specific properties (differentia
specifica) which distinguish the lexeme from its hyperonym (e. g. daisy ‘a
flower which is very common, small, and white with a yellow centre’). It
follows from the relationship of inclusion between the intension of the
hyponym and that of the hyperonym (i. e. the intension of the former including the intension of the latter), that there is a relationship of inclusion
on the level of extension as well: the extension of the hyperonym includes
the extension of the hyponym (the set of roses is a subset of the set of
flowers.)
Heteronymy (or: incompatibility): Alternative terms for co-hyponymy,
the relationship between hyponyms situated at the same hierarchical
level, are heteronymy and incompatibility. This captures the fact that in
most cases co-hyponyms/heteronyms are semantically incompatible in a
given context (either This is a rose is true in a particular context, or This
is a tulip is true, but not both). Heteronyms are not always incompatible,
however: e. g. novel and paperback are hyponyms of book and hetero-
6.3
a matter of
perspective
hierarchies
hyponym <->
hyperonym
153
6
Figure 6.2:
Hierarchical sense
relations
Semantics: Word and sentence meaning
nyms of each other, but they are not incompatible (This is a novel and
This is a paperback may both be true descriptions of the same object).
Sometimes incompatibility is also described as a fifth type of antonymy.
Hierarchical sense relations
hyponymy
hyperonym
plant
hyponyms
hyperonyms
hyponyms
meronymy
vs. hyponomy
Figure 6.3:
Sense relations –
an overview
meronymy
holonym
flower tree
bush
…
rose
daisy
…
tulip
face
meronym
holonym
meronyms
mouth nose
eyes
lips
tongue …
teeth
Meronymy refers to part-whole relationships in the vocabulary (e. g. cockpit – airplane, spoke – wheel, finger – hand, toe – foot, mouth/nose/eye
– face, door/window/roof – house). Such meronymic relationships hold
between words on different hierarchical levels. (Caution: the term for the
type of hierarchy involving such part-whole relationships is meronomy,
but this has nothing to do with linguistics.) Thus, door is a meronym of
house (the holonym), but the word also has its own meronyms (e. g. handle and lock). Meronymy, as opposed to hyponymy, is not necessarily a
transitive relationship. If A is a hyponym of B, and B a hyponym of C,
then A is always a hyponym of C (e. g. for A = bobtail, B = dog, and
C = animal). In contrast, meronymic relations need not be transitive
(e. g. for A = hole, B = button, and C = shirt), though there do exist examples of transitive meronymic relationships (e. g. A = lips, B = mouth,
and C = face). Meronymy and hyponymy involve completely different
types of hierarchies. Hyponymy involves a relationship of inclusion between classes: the extension of the hyponym is included in that of the
hyperonym. The hierarchical relationships involved in meronymies are of
a completely different type, relating to individual referents of meronymic
terms (a finger is part of a hand, a hand part of an arm, etc.). This has
nothing to do with a relationship between different classes.
^ĞŶƐĞ ƌĞůĂƚŝŽŶƐ
ᬿᭀᬿ᭘ᬹᬺᬶᭂᬱᭂ᬴ᬹᬺ᬴ᬱᬽ
ᬹᬺᬶᭂᬱᭂ᬴ᬹᬺ᬴ᬱᬽ
ᭃᭌᬿᭀᬿᭌᬾᭌ
ᭌᬾᭌ
ᭀᭁᭁᭀᭃᬺ᭄ᬺᭀᬿᭃ
ᬹᭌᭁᭀᬿᭌᬾᭌ
᬴ᭀᬾᭁᬽᬶᬾᬶᬿ᭄ᬱᭂᭌ
ᬱᬿ᭄ᭀᬿᭌᬾᭌ
᬴ᭀᬿ᭄ᭂᬱᭂᬺᬿᬶᭃᭃ
;Őᭂᬱᬵᬱᬳᬽᬶ
ᬱᬿ᭄ᭀᬿᭌᬾᭌͿ
᬴ᭀᬿᭊᬶᭂᭃᬶᬿᬶᭃᭃ ᬵᬺᭂᬶ᬴᭄ᬺᭀᬿᬱᬽ
ᭀᭁᭁᭀᭃᬺ᭄ᬺᭀᬿ
154
…
ᬾᬶᭂᭀᬿᭌᬾᭌ
ᬾᬶᭂᭀᬿᭌᬾ
ᬹᬶ᭄ᬶᭂᭀᬿᭌᬾᭌ
Structural semantics: Meaning structures in the vocabulary
6.3
6.3.3 | Lexical ambiguity: Polysemy and homonymy
Polysemy vs. homonymy: Lexemes with only one descriptive meaning are
called monosemous. Many lexemes, however, have several descriptive
meanings and are thus (a) members of more than one lexical field, and
(b) nodes in a network of sense relations that is even more complex than
the network of semantic relations contracted by monosemous lexemes.
Such ambiguous words can be divided into two major types: polysemous
and homonymous items (homonyms). The different meanings of polysemous lexemes are commonly felt to be related. Typically, one of these
senses has developed from the other sense via metaphorical or metonymical processes (e. g. mouth ‘mouth (part of the body) / river mouth / cave
entry’, or wing ‘wing of a bird / building / car / airplane / political
party’). For homonyms, by contrast, it is neither synchronically nor, in
many cases, diachronically possible to establish a connection between
the different meanings (e. g. race ‘a sports event’ / ‘a human race’, or
mole ‘animal’ / ‘dark spot on a person’s skin’). In the case of polysemy,
we can speak of a single lexeme having several meanings, whereas in the
case of homonymy we speak of different lexemes that happen to have the
same form. Dictionaries often reflect this distinction: a polysemous word
has only one entry (with various meanings that are numbered consecutively), whereas a homonym has several entries (e. g. mole1, mole2, mole3).
Types of homonymy: Homonyms can be more precisely differentiated
with the help of two criteria: (a) medium-independent vs. mediumdependent formal identity, and (b) complete identity vs. differences in
grammatical properties. In some cases, homonyms are identical in both
spelling and pronunciation, and thus qualify as ‘true’ homonyms, or
homonyms in the narrow sense. In many others, however, they are identical in spelling only, but differ in pronunciation, or vice versa. Homophones are lexemes which are identical in pronunciation, but differ in
spelling (see – sea, sight – site, flower – flour), while homographs are
identical in spelling, but differ in pronunciation (lead /led/ ‘kind of metal’
vs. /liːd / ‘piece of leather attached to dogs’ collars’, bass /beɪs/ ‘man
with a deep singing voice’ vs. /bæs/ ‘type of fish’). The second criterion
for distinguishing different types of homonyms applies equally to true
homonyms, homophones, and homographs. It can be formulated as follows: Are the homonyms under consideration identical with regard to
two types of
ambiguity
homonyms,
homophones,
homographs
Homonymy
in general
homonymy
total
partial
medium-independent
pronunciation:
homophony
orthography:
homography
total
total
partial
partial
Figure 6.4:
Homonymy
155
6
cognitive
advantages
Semantics: Word and sentence meaning
their grammatical properties (in particular concerning their word class
and inflectional morphology)? If yes, we are dealing with total homonymy, total homophony, and total homography, respectively (see all examples above); if no, we are dealing with partial homonymy (bear N –
bear V), partial homophony (rite N – write V), or partial homography
(tear N – tear V).
Does etymology help? It is frequently impossible to give a clear answer
to whether an ambiguous word is an example of polysemy or homonymy
– which once again illustrates the fact that there are no sharp dividing
lines in language. Even if we consult etymological information (e. g. with
the help of the Oxford English Dictionary, short: OED), which normally
should be avoided in synchronic analyses of meaning, it remains unclear
how far back we should go in the history of a word, and of what use this
method really is. Take, for example, the two senses of pupil (‘student’ and
‘part of the eye’). Both meanings derive from the same Latin origin:
pupilla = ‘orphan, ward’ and pupula = ‘pupil, eye’ are both derived
from pupa = ‘little girl’. So this could count as evidence of polysemy.
However, the two senses are so far apart in Present-Day English that we
tend to classify pupil ‘student’ and pupil ‘part of the eye’ as homonyms.
“Maximizing” polysemy: In general, polysemy is considerably more
frequent than homonymy. This is not surprising from a psychological and
economical perspective. Polysemy is a product of our metaphorical and
metonymical creativity and allows us to describe, in a motivated way,
something new with the help of something already known. In this way,
polysemy adds to the flexibility and adaptability of the vocabulary of a
language without increasing the number of lexemes. A language which
makes extensive use of polysemy keeps the memory load to a minimum,
because fewer words have to be stored in our minds than would be the
case if we had to learn a separate word for every concept. Wherever possible, ambiguous words will be classified as cases of polysemy rather than
homonymy. This tendency is particularly pronounced in cognitive semantics (see section 6.4).
Ambiguity in puns: Polysemous and homonymous terms have one conspicuous feature in common: A given context usually forces us to select
one particular meaning of these words. An exception to this rule are
puns, which are based on the fact that two meanings of a word or wordform are activated at the same time. Examples include the newspaper
heading Wait watchers (which alludes to the organization Weight Watchers), the announcement in (6a) informing the local citizens that a shoe
shop will be opening soon, or the panda joke in (6b):
(6)
156
a.
b.
Soon we’ll take the wait off your feet.
A panda walks into a bar, sits down and orders a sandwich.
He eats the sandwich, pulls out a gun and shoots the waiter
dead. As the panda stands up to go, the bartender shouts,
“Hey! Where are you going? You just shot my waiter and you
didn’t pay for your sandwich! Who do you think you are?”
The panda yells back at the bartender, “Hey man, I’m a
PANDA! Look it up!”
Structural semantics: Meaning structures in the vocabulary
6.3
The bartender opens his dictionary and reads the following
definition of panda:
“A tree-dwelling marsupial of Asian origin, characterized by
distinct black and white colouring. Eats shoots and leaves.”
Ambiguity vs. vagueness: Usually, however, only one particular meaning
of an ambiguous word fits a given context. Ambiguous words are disambiguated by contextual selection of one of their (descriptive) meanings. A
common test of ambiguity are cases where two different contexts are
relevant to the interpretation of a word. These contexts require the activation of different meanings of the word and therefore lead to a bizarre or
unacceptable sentence meaning (in rhetoric, the term zeugma is used for
cases such as (7a)). A second test, the so-called identity test in (7b), also
shows very clearly that expire is ambiguous:
(7)
a.
b.
?John
?John
and his driving license expired last week.
expired last week; so did his driving license.
Such tests, then, allow us to determine whether a word is ambiguous or
merely vague. Vague terms are unspecified for certain semantic features
(e. g. monarch is unspecified for sex: Her father / His mother was a monarch). For this reason, they display a certain flexibility in their use. What
is crucial, however, is that this flexibility does not lead to the assignment
of more than one meaning. It is therefore characteristic of vagueness that
we are not compelled by a particular context to decide between two or
more meanings (in other words, there is no need for disambiguation by
contextual selection).
At most, a particular context leads to a more precise specification of
the word’s meaning (e. g. by emphasizing or suppressing a feature). This
type of specification is called contextual modulation of meanings, as illustrated for window in (8a–c):
(8)
a.
b.
c.
d.
tests
contextual
modulation
Joan opened / shut / repaired the window. (neutral)
Joan painted the window. (frame)
Joan cleaned / broke / looked through the window. (glass
panel)
While painting the window, Joan broke it.
As shown in (8d), such modulations can be combined in a sentence without leading to a zeugma. In sum, the distinction between vagueness and
ambiguity is of great importance when it comes to determining whether
a word has merely one meaning or whether it has several meanings: A
word which is vague has one meaning only, while words that are ambiguous have several meanings.
157
6
Semantics: Word and sentence meaning
6.4 | Cognitive semantics: Prototypes and metaphors
language inseparable from cognition
perceiving/
constructing
similarities
Categorization: Cognitive semantics developed in the 1980s on the basis
of findings in cognitive psychology. Scholars working in this area of research have challenged many time-honoured assumptions familiar from
structuralist semantics. The chief difference between the two approaches
is that structural semantics defines and analyses meaning from a purely
language-internal perspective (i. e. on the basis of semantic networks
connecting lexemes), whereas cognitive semantics explains meaning primarily in terms of categorization (i. e. the grouping of similar phenomena
into one class). In cognitive semantics, meaning is considered to be inextricably linked to human cognition, to the way we perceive the world and
group phenomena into conceptual categories. Language and cognition
are taken to be inseparable: the structure of linguistic categories is held to
reflect the structure of conceptual categories (e. g. in the sense that the
meaning of a word is the cognitive category connected with it).
Categorization essentially involves the perception or construction of
similarities between otherwise different entities. Prototypes and metaphors play a central role in this process: Prototypes are considered to
serve as reference points for categorization. The concept of metaphor, on
the other hand, brings up the following questions: Which processes allow
us to perceive or construct similarities? Are these similarities objectively
given or subjectively created?
6.4.1 | Prototypes
knowledge
organization
the traditional,
feature-based
view
flexibility and
fuzziness
158
Prototype semantics vs. feature semantics: When categorizing a given animal as a duck (rather than a goose), a given drinking vessel as a cup
(rather than a mug), a given activity as running (rather than walking), or
a given car as new (rather than used), we access what we know about
poultry, cups, running, and old and new cars, respectively. This raises an
important question: How is this knowledge organized?
According to the traditional view, categorization is achieved by means
of “necessary and sufficient conditions”: An object counts as an X only if
it possesses all the features which define an X. Necessary conditions (criteria) are features which are indispensable for an entity to belong to a
given category. The term “sufficient conditions” is used when these features are jointly sufficient for assigning the entity to a certain category
(i. e. when all of the various necessary criteria apply). This model of categorization implies that categories have clear-cut boundaries. The traditional account of categorization lies at the heart of feature (or: componential) semantics (see section 6.3.1), which is sometimes rather dismissively referred to as checklist semantics. Componential semantics can be
traced back as far as Aristotle’s theory of concepts.
Prototypes as cognitive reference points: Cognitive semantics, more
precisely prototype semantics, rejects the classical view of categorization,
at least for the majority of concepts: categorization in everyday life – less
so in the domain of science – is much more flexible and fuzzy than is
Cognitive semantics: Prototypes and metaphors
suggested by traditional componential semantics. Some ducks have no
wings, others cannot quack, even though they have all other properties
associated with ducks and are therefore spontaneously categorized as
ducks. Why? Because they do, after all, come very close to our idea of an
‘ideal’ (or: prototypical) duck, or at least correspond much more to the
prototype of a duck than to the prototype of a rivalling category (e. g. a
goose). Thus, we can assign an entity to a category if it shares at least
some central features with the category prototype, and (in most cases)
looks physically similar to it.
Dictionary vs. encyclopaedic knowledge: Prototype semantics assumes
that all knowledge which is accessed in a particular situation is relevant
to the process of categorization. For this reason, we cannot strictly separate ‘dictionary’ knowledge (‘knowledge of what is essential, pertaining
to what speakers know in virtue of their command of a language’) from
encyclopaedic knowledge (‘additional knowledge pertaining to what
speakers know in virtue of their acquaintance with the world’). For example, there may be situations in which what is crucial to categorizing an
animal as a duck is the encyclopaedic knowledge that during their search
for food ducks hold their head under water and raise their tail in the air
(cf. the German nursery rhyme All my little ducklings).
Internal heterogeneity of categories: The above examples show that
categories have an internal structure, which implies that they are not
homogeneous: not all category members are equally good representatives
of the category; rather there are different degrees of representativeness.
Categories have a core consisting of the best representatives (the prototypes), which serve as reference points in the process of categorization
and which are surrounded by increasingly peripheral members that are
more and more different from the prototype(s).
Fuzziness of category boundaries: Yet another important difference
about the conception of categorization differentiates prototype semantics
from the view espoused by traditional (including componential) semantics. Cognitive semantics emphasizes that category boundaries are often
not clear-cut (fuzziness of category boundaries). Therefore, it is frequently
not possible to give a clear answer to the question whether or not an
entity belongs to a category. There are grey areas of transition between
neighbouring categories where we are incapable of unambiguously assigning an entity to one category rather than another. Different speakers
may thus assign the same entity to different categories, and even individual speakers may classify the same entity differently on different occasions.
Family resemblances: Many findings associated with the psychological
theory of prototypes, which underlies prototype semantics, involve the
notion of family resemblances. This concept was developed by the philosopher Ludwig Wittgenstein. Wittgenstein used the example of games to
show that a category can be held together by nothing more than a complex web of overlapping and crisscrossing similarities among its members, comparable to the various similarities displayed by different members of a family. The analogy between categories and families turns on
the fact that family members usually resemble each other with respect to
6.4
Ludwig
Wittgenstein
159
6
hierarchy of
categories
‘more-or-less’
semantics vs.
‘all-or-nothing’
semantics
160
Semantics: Word and sentence meaning
various “crisscrossing” similarities: Some members have a similar nose,
others the same skin, yet others the same eyes, etc. Similar observations
can be made with respect to the individual members of the category
GAME. Some games are amusing, some involve winning and losing, yet
others require particular skills, etc. In such cases, establishing necessary
conditions is difficult if not impossible. There need not even be a single
feature which is shared by all category members. As a consequence, prototype theory discards the idea that category membership is always determined by necessary conditions. It also rejects the assumption that categorization should necessarily be construed as a comparison between the
entity to be categorized and the prototypes (cognitive reference points) of
a category. This does not mean, however, that the concept ‘prototype’ is
given up; prototypes of a category are characterized by a high degree of
family resemblance.
Basic-level categories: Some scholars suggest that the family resemblance model is particularly suitable for explaining superordinate categories, such as Wittgenstein’s GAME, or ANIMAL, PLANT, FURNITURE,
and CONTAINER, while the classical prototype model is particularly illuminating for those kinds of categories that are situated at the psychologically most basic level (so-called basic-level categories like DUCK, DOG,
CAT, FLOWER, TABLE, BAG). These categories are psychologically basic
in the sense that they contain the most information in relation to the cognitive cost of storing them. Their basicness is reflected in quite a number
of facts: Basic-level categories are acquired very early by children, they
are rapidly recognized and represent the default choice in spontaneous
categorization (“Look, a ...!”), and they represent the highest level of
classification at which a single image can represent the entire category.
Significance of prototype semantics for lexical semantics: The connection between lexical semantics and what has been said above about categories and categorization is natural. According to prototype semantics, the
meaning of a word like duck is the cognitive category that is associated
with it. As a consequence, word meanings contain all of the above-mentioned properties of cognitive categories: we can distinguish central and
more peripheral meanings of a lexeme, and word meanings are not rigid.
There are often gradual transitions between word meanings (recall the
notion of contextual modulation discussed above; e. g. the different uses
of window in He painted the window and He smashed the window). Prototype semantics is thus a ‘more-or-less semantics’, which – due to its integrative approach that rejects the traditional distinctions between dictionary and encyclopaedic knowledge, and between meaning and cognitive
categories – is much closer to psychological reality than traditional feature
semantics (or ‘all-or-nothing’ semantics) in structuralist lexicology. This
does not, however, detract from the usefulness of feature semantics for the
description and comparison of word meanings, especially for identifying
semantic structures like lexical fields and sense relations.
Prototype and structural semantics complement each other: We do not
even have to abandon the feature approach as a theory of how meanings
are mentally represented: Neither the ‘standard version’ of prototype theory nor the more recent family resemblance model can do without a fea-
Cognitive semantics: Prototypes and metaphors
6.4
ture-based classification. It is just that the features relevant for categorization are those belonging to the prototypes of a category. Moreover,
there is no list of necessary features that needs to be checked for successfully assigning entities to a particular category. Ultimately, prototype and
feature semantics complement each other, in the sense that feature semantics receives a sounder psychological basis.
6.4.2 | Metaphors
Vehicle – tenor – tertium comparationis: The term metaphor (Greek
metaphero = ‘carry somewhere else’, with the noun metaphora already
in its modern meaning) traditionally refers to a figure of speech which is
based on a relationship of similarity or analogy between two terms from
different cognitive domains. This similarity, which may be objectively
given or merely subjective, is typically held to enable metaphors to ‘transport’ one or more properties of a (usually relatively concrete) source domain (or: vehicle) to a target domain (or: tenor), which is typically more
abstract. The similarities involved in metaphorical mappings are often
called the tertium comparationis (or ground).
Examples of metaphors: Typical examples are:
■ animal metaphors (Smith is a pig / fox / rat / ass / stallion),
■ synaesthetic metaphors (extensions from one field of sensory perception to another, e. g. in loud colours, soft / warm / sharp voice), and
■ so-called anthropomorphic metaphors (transfers from the human
domain, especially human body parts, to all sorts of non-human
domains, e. g. leg of a table, arm of a river, face / hands of a clock, foot
of a mountain, mouth of a river).
Metaphors are traditionally neglected in lexical semantics, though they do
play a role in historical semantics and in syntagmatic semantics. Historical semanticists view metaphor as an important cause of semantic change
(see chapter 9).
Metaphors in syntagmatic vs. cognitive semantics: In syntagmatic semantics, metaphors have been explained in terms of selection restrictions. For example, in sentences like Smith was a rat or He picked one hole
after the other in my argument, selection restrictions are violated (semantic incongruence): in the first example [+ human] clashes with [– human],
in the second [+ concrete] (pick a hole) with [– concrete] (into an argument). In cognitive semantics, metaphors are seen in a completely different light: Metaphor is not considered as a purely linguistic phenomenon,
but as a fundamental cognitive process which enables us to grasp the
world and organize our knowledge. Metaphors pervade everyday language and are crucial to human thought processes, they are not simply
dispensable ornamental accessories. Many metaphors are likely to go unnoticed by ordinary speakers. This is not surprising, though. We are often
no longer aware of many metaphors simply because they are firmly anchored in human cognition and have become part and parcel of ordinary
language.
selection restrictions vs. grounding
in cognition
161
6
relevance for
categorization
focus on everyday
metaphors
metaphorical
concepts vs.
metaphorical
expressions
Semantics: Word and sentence meaning
Metaphor as a basic cognitive process: How do cognitive semanticists
arrive at this conception of metaphors? This question can be answered by
having a closer look at the process of categorization, i. e. of comparing
new things to already familiar ones. At the heart of this process lies the
search for similarities or analogies. It is easier to understand and describe
the world if we can grasp new concepts with the help of existing categories. In some cases, this may involve extending these categories. However, such a strategy of understanding unknown concepts in terms of familiar ones has the advantage that (a) the categories we need for grasping
the world are not unnecessarily multiplied, and (b) that classifications are
not arbitrary, but motivated by similarities between those entities that are
new and those that are already familiar.
Such similarities do not have to be objectively given; (some) similarities underlying (some) metaphors are predominantly constructed by
speakers. It is language users themselves who determine the ground of
comparison (tertium comparationis). Some metaphors strike us as novel
and original even after we have encountered them many times. Cases in
point are ‘poetic metaphors’ found in classical rhetoric and literary works,
e. g. My life had stood – a loaded gun in corners... (Emily Dickinson).
Cognitive semantics is not primarily concerned with this type of metaphor but focuses for the most part on ‘everyday metaphors’, i. e. conventional metaphors which are not isolated but rather part of entire systems
of metaphors. It is commonly assumed that these metaphorical systems
allow us to structure particular areas of experience. Let us take a look at
some examples of relevant metaphors (9) and systems of metaphors (10).
Asymmetry / unidirectionality: The arrows in these two sets of examples represent the link between source and target domains, highlighting
one of the fundamental properties of metaphors, namely their asymmetry
or unidirectionality (at the most general level: concrete → abstract, spatial
→ non-spatial). The emphasis cognitive linguistics places on the conceptual nature of metaphors is reflected in the distinction between metaphorical concepts and metaphorical expressions. According to cognitive linguists, metaphorical concepts such as ARGUMENT IS WAR take priority
over concrete metaphorical expressions like attack (a claim) or shoot
down (an argument). Every metaphorical expression can be subsumed
under one or several metaphorical concepts. In fact, we can use such
metaphorical expressions only because the corresponding metaphorical
concepts are part of our conceptual system. Conceptual metaphors are
usually indicated by capital letters.
(9)
a.
b.
c.
162
LIGHT → THOUGHTS/KNOWLEDGE/INTELLECT
illuminating/obscure ideas, a murky discussion, a bright person, a clear argument, make ideas transparent, I see ‘I understand’
WAR/PHYSICAL ARGUMENT → VERBAL ARGUMENT
his criticisms were right on target, shoot down an argument,
attack a weak point in someone’s argument
MONEY → LANGUAGE
coin new words, owe someone an answer, richness in expressions
Cognitive semantics: Prototypes and metaphors
6.4
(10) UP – DOWN → PERSONAL WELL-BEING (e. g. HAPPINESS,
HEALTH, POWER, STATUS)
a. HAPPINESS/GOOD IS UP, BAD (LUCK) IS DOWN
feel up/down, be in high/low spirits, fall into a depression
b. HEALTH IS UP, ILLNESS/DEATH IS DOWN
be in top shape, be at the peak of health, fall ill, drop dead
c. CONTROL/INFLUENCE IS UP, LACK OF CONTROL/INFLUENCE IS DOWN
be in high command, at the height of power, on top of the situation, fall from power, be under control
d. HIGH STATUS IS UP, LOW STATUS IS DOWN
rise to the top, be at the peak of your career, be at the bottom
of the social hierarchy, fall in status
It is a basic assumption in cognitive semantics that such metaphors are
more or less constantly used for structuring abstract concepts in terms of
concrete (especially spatial) ones.
Metonymy: Like metaphor, metonymy (gr. metonymia = renaming) is
a classical figure of speech which has been assigned a completely new
status in cognitive semantics. Consider the examples in (11) and (12):
(11) a.
PRODUCER FOR THE PRODUCT: She owns a Picasso and two
Frida Kahlos.
b. OBJECT/INSTRUMENT FOR OBJECT/USER OF INSTRUMENT: The buses are on strike.
c. PLACE FOR INSTITUTION: The White House is planning to
attack Iran.
d. INSTITUTION FOR THE PEOPLE IN CHARGE: The university will reject this proposal.
e. PLACE FOR RESPONSIBLE PEOPLE: Table 10 want their bill.
(12) PART FOR THE WHOLE (pars pro toto)
a. He’s a good hand at gardening.
b. There are not enough good heads in this company.
c. I don’t see any new faces – nothing seems to have changed.
Metonymy, too, is considered to be a central cognitive process which
enables us to ‘get a better grasp’ on the world. The main difference between metaphor and metonymy is this: Metonymies do not involve a
transfer from one cognitive domain to another. They are rather based on
an existing objective connection between two “contiguous” phenomena,
such that one phenomenon stands for the other. Thus, metonymies are
not based on a relationship of similarity, but of contiguity: The phenomena or entities concerned are part of the same situation or, more generally, part of the same conceptual structure. Picasso does not resemble his
pictures, buses do not resemble bus drivers, and table 10 does not resemble the restaurant guests that sit at that table. But there is certainly a direct connection between painters and their paintings, bus drivers and the
buses they drive, or the plate and the dish that is served on it (just compare the standard encouragement for finishing off one’s meal used in
key difference
from metaphor:
contiguity
163
6
polysemous
expressions as prototype categories
Semantics: Word and sentence meaning
German especially when addressing children: Jetzt iss schön den Teller
auf! lit. ‘eat up the plate’). Various types of such connections are illustrated in (11) and (12). Those in (12) form a separate group which in
classical rhetoric is called synecdoche, a term which covers part-whole
and whole-part relations (as in German Zünd doch mal bitte den
Weihnachtsbaum an! lit. ‘Please light the Christmas tree’).
Prototypes, metaphors and polysemy: The two central concepts of cognitive semantics – prototypes and metaphors – are both relevant to investigating polysemy at the level of word meaning. It is not difficult to see
why prototypes are crucial to explaining polysemy: Polysemous expressions can be described as prototype categories in that they can have one
or more central meanings (the prototypes), each of which can have increasingly peripheral sub-senses, and all of which are connected by different family resemblances. This can be illustrated with the help of prepositions, a word class which is notoriously polysemous. For instance,
there are countless ways in which over can be used as a preposition, but
we can single out three central meanings: place (‘above’) in (13), place
(‘above’) in connection with path (‘across’) in (14), and a covering sense
in (15):
(13) a.
b.
(14) a.
b.
c.
(15) a.
b.
c.
relevant also for
semantic change
motivation vs.
arbitrariness
164
The lamp hangs over the table.
The painting is over the mantelpiece.
The plane flew over the house.
John walked over the hill.
John lives over the hill.
The board is over the hole.
The guards were posted all over the hill.
There was a veil over her face.
Each of these three central meanings has a prototypical core (the one in
the (a)-examples) and other meanings that can be systematically derived
from the central meaning (e. g. over in (15c), which involves a vertical
rather than a horizontal axis). One of the above three (groups of) prototypical senses, notably the one in (14a), has a more central position than
the others, and is thus even “more prototypical” than the other two prototypes in (13) and (15). In many cases, the connections between the
different meanings are established by metaphors.
Metaphorical transfer both accounts for the synchronic relations between different senses of a word and offers a diachronic explanation how
one sense develops from another. There is thus a close connection between polysemy and metaphor as a central cognitive mechanism for
grasping and classifying new entities with the help of familiar ones, and
abstract things with the help of concrete ones. In this context, polysemy
is deliberately construed in a wide sense, i. e. a polysemous word may
have senses belonging to different word classes (e. g. over as a preposition, adverb, and part of a compound).
Further differences between cognitive and structural semantics: It is
not only for ambiguous expressions that cognitive semanticists have
shown that the link between signifié and signifiant is more motivated
Cognitive semantics: Prototypes and metaphors
than suggested by traditional structural semantics. The assumption that
this link is far less arbitrary than is commonly conceded has led to further
insights that put two central structuralist ideas into perspective: (a) linguistic categories provide speakers with the cognitive categories that enable them to grasp the world (this position is called extreme determinism); (b) therefore, the conceptual system of a language has to be analysed on its own terms. The second of these structuralist assumptions is
inextricably linked to the hypothesis that languages are autonomous systems and cut up the same conceptual domain in different ways. For this
reason, every language is held to create its own view of the world, a position known as extreme relativism.
Sapir-Whorf hypothesis: Cognitive semanticists challenge the latter position (also known as the linguistic relativity principle or the Sapir-Whorf
hypothesis), arguing that the basic categorization and metaphorization
processes are the same, or at least very similar, for all people – at least
among the members of the same cultural community, but in many cases
also across cultures. Consequently, the differences concerning the ways
in which members of different speech communities categorize the world
are limited. A well-known example illustrating this fact are colour terms.
Comparative analyses of basic colour terms in different languages have
shown that languages may indeed differ as to where they set the boundaries between neighbouring categories (here primary colours). Crucially,
however, speakers of different languages agree on what constitutes the
centre of the respective colour categories, i. e. on what constitutes the
‘best’ red, green, blue, etc. The perception and processing of reality is
thus not primarily a matter of the native language one happens to speak.
Linguistic categories do not determine our cognitive categories. Quite to
the contrary, they reflect the structure of our conceptual system.
6.4
challenging extreme relativism
Checklist Semantics – key terms and concepts
ambiguity ↔ vagueness
antonymy (complementary
antonymy; contrariness;
converseness; directional
opposition; heteronymy)
arbitrariness
asymmetry
basic-level category
categorization
cognitive semantics
collocations
componential analysis
conditions of use
connotation ↔ denotation
contextual modulation
contextual selection
conventionality
degrees of representativeness
descriptive / cognitive ↔
expressive ↔ social meaning
determinism ↔ relativism
disambiguation
encyclopaedic meaning of a
lexeme ↔ dictionary meaning
family resemblance
fuzziness of category boundaries
heterogeneity of categories
heteronymy / incompatibility /
co-hyponymy
hierarchical sense relations
(hyponymy; meronymy)
historical / diachronic semantics
165
6
Semantics: Word and sentence meaning
holonym
homonymy (total ↔ partial;
homography, homophony)
hyponymy ↔ hyperonymy
idiom
intension ↔ extension
lexical ↔ grammatical
meaning
lexical relations
lexical semantics / lexicology
lexicology ↔ lexicography
markedness
meaning of utterance
mental lexicon
meronymy
metaphor
metaphorical extension
motivation
necessary ↔ sufficient condition
paradigmatic ↔ syntagmatic
semantics
polysemy
proposition
prototype
prototype semantics ↔ feature
semantics
Sapir-Whorf hypothesis/
principle of linguistic relativity
selection restrictions
semantic feature / seme
semantics ↔ pragmatics
semasiology ↔ onomasiology
sense ↔ reference
sentence meaning
sentence semantics
signe différentiel
source domain, vehicle ↔ target
domain, tenor
structural semantics
synonymy (descriptive/
cognitive ↔ total)
semantic field
truth conditions
Exercises
1. Which of the following uses of mean are relevant in a discussion of
what semantics is concerned with and how it differs from pragmatics?
a) This face means trouble.
b) What does soliloquy mean?
c) If you’re not there by six, I’ll be gone. And I mean it.
d) You’re meant to take off your shoes in a mosque.
e) Sorry, I don’t quite understand. What exactly do you mean?
f) Do you mean to say you can’t come?
g) His work means everything to him.
h) I never meant her to read this letter.
i) Smoke means fire.
2. Fill in the chart below with ‘+’ or ‘–’ as appropriate:
a) establishes a link between language and the world
b) independent of a particular utterance
c) involves a set of possible referents
d) to be found in a dictionary definition
e) lists defining properties
166
Exercises
sense
6.4
reference intension extension denotation connotation
a.
b.
c.
d.
e.
3.
a) What are the semantic relations between see and the other lexemes in the following groups that represent different lexical fields?
see – hear – feel, see – know – understand, see – look at – watch,
see – visit – meet, see – imagine, see – sea
b) Under which conditions can a lexeme belong to more than one
lexical field?
4. Identify the lexical relations holding between the following pairs of
words:
frame – window, expand – contract, mole – spy, fill – empty, (go) in
– (go) out, fail – succeed, hyponym – hypernym, picture – painting,
zero – love, semantics – linguistics, freedom – liberty, after – before,
book – index
5.
a) What is funny about the headline “Where’s the party?” (Subtitle:
“How to get young people to vote for their politicians”).
b) Explain the linguistic basis of the panda joke in example (6b).
6. Explain the role of context in drawing a distinction between
(a) semantics and pragmatics, (b) vagueness and ambiguity, (c) total
and cognitive synonymy.
7.
a) What is structural about structural semantics?
b) What are the major differences between structural and cognitive
semantics?
8. Which of the following statements are true and which are false?
a) Compiling a semantic field and identifying the sense relations
among the field members are both instances of adopting an onomasiological procedure.
b) Semantic fields are two-dimensional and have neither gaps nor
words with overlapping or identical senses.
c) Polysemous lexemes cannot belong to more than one semantic
field.
d) Homonymy and semantic change are two sides of the same coin.
e) Oppositeness plays an important role in the organization of our
mental lexicon.
167
6
Semantics: Word and sentence meaning
f) Hyponymy involves the inclusion of semantic features of the
higher categories.
g) Semantics is exclusively concerned with the descriptive meaning
of content words.
h) Prototype categories (e. g. bird, dog, cup, toy) always have fuzzy
boundaries.
i) Componential analysis and prototype theory do not exclude each
other.
j) Categorization always involves metaphor.
Advanced
9. Absolute synonyms are rare. High and tall are considered near synonyms in English. Try to answer the following questions based on the
examples below, which illustrate typical uses of the two lexemes.
a) In which of the examples are high and tall interchangeable? In
which contexts is the choice restricted to just one of these items?
b) Based on your answers to a), try to give an outline of semantic
similarities and semantic differences between high and tall.
Some authentic examples, mostly from the BNC:
high
1. Good health is not just about providing efficient high quality
medical services. (BNC:A0 J 1358)
2. Due to the high level of burnout common in such chefs, few are
offered jobs. (BNC:A0C 1377)
3. People chose to spend a high proportion of their disposable income on buying and running a car because car ownership enhances their lives. (BNC:A2L 103)
4. Mrs Thatcher has been advised that a complete ban on strikes is
not a practical proposition and may entail a high degree of political risk. (BNC:a2T 112)
5. But availability of coal resource has never been the industry’s
problem – the essential difficulties are lack of demand and high
cost of production.
6. VW has high hopes for the Polo in this country. (BNC:A6 W 431)
7. I didn’t expect nothing like this. It’s got this great high roof and
loads of trains. It’s real smoky and that. (BNC:A74)
8. “Come out!” The voice echoed in the high arches of the church.
(BNC:HU0)
9. The room was large and square with high ceiling and two tall
curtainless windows. (BNC:BN1)
10. High Skies and Fat Horses (Title of a novel by William J. Wallisch)
tall
11. The effect is similar to dressing a tall man in a pinstripe suit – it
simply accentuates the length! (BNC:A0G 1442)
12. Suddenly, there stood beside me a very tall figure, six foot six or
more, bearded and misty-white in appearance. (BNC:B2G 265)
168
Exercises
6.4
13. From Middenheim’s many tall towers it is possible to look out
over the Great Forest to the south and the Drakwald to the west ...
(BNC:CN1 386)
14. It was still night and the tall trees stood silently against the stars.
(BNC:ACE 3258)
15. I could think of nothing except going to London and finding my
way among its tall buildings studded with lights. (BNC:A0U
1374)
16. They knew the owner, the well-to-do, the grandees back from Jamaica and Bengal who sat here now behind the tall walls and
drew their rents. (BNC:A0N 448)
17. Don’t allow your personal feeling to cloud your judgement in
finances or joint arrangements – a tall order because you seem to
be emotionally involved, too. (BNC:CB8 2837)
18. The judge obviously thought that A had told a tall story. (BNC:H81
70)
10. The possessive in English is a highly interesting construction, given
that it can express a wide range of semantic relationships. For example, the possessed can be ...
i)
something owned by the possessor (his book)
ii)
one of the possessor’s relatives (her brother)
iii)
one of the possessor’s body parts (his arm)
iv)
an “unowned possession” (the child’s schoolbooks)
v)
an individual somehow related to the possessor (our dean)
vi)
a physical quality (his weight)
vii) a mental quality (her intelligence)
viii) a permanent location (their neighborhood)
ix)
a transient location (my spot)
x)
a situation (my predicament)
xi)
an action carried out (Oswald’s assassination)
xii) an action undergone (Kennedy’s assassination)
a) What types of linguistic mechanism(s) might motivate the different uses of the possessive illustrated above?
b) Is there a common semantic feature characterizing all the different
types of “possession” which can be marked by the possessive? At
first sight, it might be suggested that the common denominator
underlying all kinds of possession marked by the possessive is
“association”. On second thoughts, however, this explanation does
not account for the asymmetries illustrated in (i) and (ii) below.
Try to explain why the expressions in (i) are possible, while the (ii)
cases are usually problematic. It may be useful to think of contexts
in which some of the otherwise odd-sounding cases in (ii) would
become acceptable.
(i)
the girl’s doll; the man’s car, the dog’s paw; the horse’s ticks;
the boy’s aunt;
(ii)
*the doll’s girl; *the car’s man; ?the paw’s dog; ?the ticks’
horse; *the aunt’s boy (‘her nephew’)
169
6
Semantics: Word and sentence meaning
11.
a) Compare the following sentences from English and Spanish, focusing on potential differences in how the two languages “construe”
motion events. Which aspects of the events are asserted (explicitly
stated) and which aspects are only implied?
(1a) English: The
boy climbed
the
tree.
(1b) Spanish: El
nino está subido en
el árbol
‘The boy is
climb-PART en [‘in’, ‘on’] the tree’
[= the boy is in a state of having climbed the tree]
(2a) English: The boy put (threw) the ball down into a container
(2b) Spanish: El nino metió la pelota en el recipiente que había
abajo
‘The boy put the ball en (can mean ‘in’ or ‘on’) the
container that was below’
b) Do you think these differences have an impact on the way we
think? Try to outline how exactly this influence may be reflected in
our thought patterns.
12. The expression single (as opposed to married) is easily defined in
terms of necessary and sufficient conditions. Construct such a definition and discuss whether this type of definition accurately reflects the
way we conceptualize “single” persons and fully captures our use of
the expression.
Sources and further reading
Aitchison, Jean. 20124. Words in the mind: An introduction to the mental lexicon.
Oxford: Blackwell.
Allan, Keith (ed.). 2009. Concise encyclopedia of semantics. Oxford: Elsevier.
Benson, Morton/Evelyn Benson/Robert Ilson (eds.). 2010³. The BBI dictionary of
English word combinations. Amsterdam/Philadelphia: Benjamins.
Croft, William/David Alan Cruse. 2004. Cognitive linguistics. Cambridge: Cambridge University Press.
Cruse, Alan. 2006. A glossary of semantics and pragmatics. Edinburgh: Edinburgh
University Press.
Cruse, Alan et al., eds. 2002. Lexikologie/Lexicology. [HSK]. Berlin/New York: de
Gruyter.
Cruse, David A. 1986. Lexical semantics. Cambridge: Cambridge University Press.
Cruse, David A. 2011³. Meaning in language: An introduction to semantics and
pragmatics. Oxford: Oxford University Press.
Davidson, George W./Peter Mark Roget. 2004. Roget’s thesaurus of English words
and phrases. London: Penguin Books.
Davis, Steven et al., eds. 2004. Semantics: a reader. New York: Oxford University
Press.
Geeraerts, Dirk. 2010. Theories of lexical semantics. Oxford: Oxford University
Press.
Geeraerts, Dirk/Hubert Cuyckens (eds.). 2007. Handbook of cognitive linguistics.
Oxford: Oxford University Press.
Goatly, Andrew. 2011². The language of metaphors. London/New York: Routledge.
Jaszczolt, Katarzyna M. 2002. Semantics and pragmatics: Meaning in language
and discourse. London/Munich: Longman.
Lakoff, George. 1987. Women, fire, and dangerous things: What categories reveal
about the mind. Chicago: University of Chicago Press.
170
Sources and further reading
6.4
Lakoff, George/Mark Johnson. 1980, reprint 2003. Metaphors we live by. Chicago:
Chicago University Press.
Lappin, Shalom/Chris Fox (ed.). 2015². The handbook of contemporary semantic
theory. Malden, MA: Wiley-Blackwell.
Löbner, Sebastian. 2013². Understanding semantics. London: Arnold.
Lyons, John. 1977. Semantics. 2 vols. Cambridge: Cambridge University Press.
Murphy, M. Lynne. 2010. Lexical meaning. Cambridge: Cambridge University
Press.
Murphy, M. Lynne/Anu Koskela. 2010. Key terms in semantics. London: Continuum.
Oxford English Dictionary online: OED online. Oxford: Oxford University Press.
Riemer, Nick (ed.). 2016. The Routledge handbook of semantics. London: Routledge.
Saeed, John. 20164. Semantics. Oxford: Blackwell.
Taylor, John R. 20033. Linguistic categorization. Prototypes in linguistic theory.
Oxford: Clarendon Press.
Ungerer, Friedrich/Hans-Jörg Schmid. 20062. An introduction to cognitive linguistics. London: Longman.
171