crisp, p , heywood,j & steen, g metaphor identification and analysis, classification and quantification

Language and Literature Copyright © 2002 SAGE Publications
(London, Thousand Oaks, CA and New Delhi), Vol 11(1): 55–69
[0963–9470 (200202) 11:1; 55–69; 020844]

Metaphor identification and analysis, classification and

quantification

Peter Crisp, Chinese University of Hong Kong, Hong Kong

John Heywood, Lancaster University, UK

Gerard Steen, Free University Amsterdam, The Netherlands

Abstract

Identifying metaphorically used words in the way we have proposed in the other articles
in this special issue inevitably leads to the detection of recurring structural patterns of
metaphor usage. It is the aim of the present article to explore these patterns in a
systematic fashion and develop a taxomony of the propositional structure of metaphors.
As a preliminary step, a decision has to be made about the units of discourse within
which one may examine patterns of the propositional structure of metaphorical
language. This article has adopted the position that all non-downgraded clauses, in the
manner of Mann and Thompson’s Rhetorical Structure Theory (1988), would be the
most eligible candidate.

Inside the boundaries of these text units, we have found single as opposed to

multiple metaphor, simple as opposed to complex metaphor, pure as opposed to mixed
metaphor, and restricted as opposed to extended metaphor. Moreover, these four
oppositions may also be combined with each other. Of the 16 logically possible
combinations, 2 are ruled out, but the other 14 combinations present just as many
structural types of metaphor.

We then move on to discuss how such a taxonomy may be used in the quantitative

characterization of the metaphorical style of an author. We show how the taxonomy
may be applied at various levels of measurement, ranging from the word through the
proposition to the text unit itself. Another possibility is to perform these measurements
at the level of metaphorical mappings, which is the option we have chosen to apply to
two stretches of fiction, by Sara Maitland and Salman Rushdie. The provisional results
of this analysis are then finally presented and discussed.

Keywords: complex metaphor; extended metaphor; multiple metaphor; propositional
structure; pure metaphor; restricted metaphor; simple metaphor; single metaphor

1 Introduction

A major advantage of the detailed approach we have taken to identifying
metaphor in naturally occurring discourse is that a number of distinct patterns of
metaphorical language use, and of the propositional structures underlying these,
have forced themselves upon our attention. Some at least of these patterns would
almost certainly have been overlooked by a more casual approach. Consider for
instance the differences between the following expressions containing
metaphorically used words from the Maitland (1990) text analysed by Heywood
et al. (2002):

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

(1)

Rachel seemed able to absorb everything that Phoebe tried.

(2)

He had set his mark on them.

(3)

Phoebe … asked herself with a sudden rush of nostalgia.

(4)

That was the stuff of melodrama.

The first sentence contains only one metaphorically used word, while each of (2)
through (4) contains two. Moreover in both (2) and (3) the two metaphorically
used words seem to denote in the same source domain whereas this is clearly not
the case for (4). There is also a further difference between (2) and (3). Whereas in
(2) the two metaphorically used words express two elements of the underlying
main metaphorical proposition, the predicate concept SET and the argument
concept MARK, in (3) the two words of sudden rush express what is just a single
argument in the underlying main metaphorical proposition, but an argument
containing some kind of internal semantic complexity.

This article will be devoted to establishing a taxonomy for metaphor that will

enable the differences of metaphorical language patterning noted in the four
sentences above to be classified systematically. In establishing this taxonomy we
will work with the three analytic levels of the surface linguistic expression, the
proposition and the conceptual mapping. All three of these levels are essential.
The importance of the linguistic expression is evident, as, in the context of
contemporary metaphor theory, is that of the conceptual mapping. The necessity
of working also with the intermediate level of the proposition is a central
assumption of our approach, one that was argued for in the first papers of this
special issue (Crisp, 2002; Steen, 2002). Propositional analysis we will see is an
essential tool in constructing a useful taxonomy of metaphor. We shall see, for
instance, that it would be difficult to account within a conceptually oriented
approach for the differences in metaphorical patterning between (2) and (3) above
without employing propositional analysis.

There are several advantages to creating a taxonomy of metaphor types of the

kind that we will present, but an important one is the possibility of the
quantification of metaphorical expressions in discourse for the purpose of corpus-
linguistic research. Another possible benefit of such a taxonomy is that it may
help the analyst to identify metaphors in the first place. It is possible to create a
schema for analysis in which a handful of questions have to be answered about
any sentence containing possible metaphorical expressions. Incorporating
classification into the business of identification itself in this manner helps to
sensitize the analyst to the presence of a wide variety of different kinds of
metaphors, a number of which might otherwise be missed.

This article will present a detailed discussion of the different patterns of

metaphorical expression exemplified above and discussed in context by Heywood
et al. (2002). We shall then continue with the categorization of these into a
taxonomy, the translation of this taxonomy into a schematic decision procedure
for metaphor classification, and finally the quantitative application of this
taxonomy to text analysis.

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

2 Units of discourse

If all patterns of metaphorical language were as clearly distinct as in the previous
examples, life would not be so difficult for the linguist interested in metaphor. In
fact many types of discourse do exhibit many instances of metaphorical language
patterning that are differentiated just as clearly and straightforwardly as in the
previously shown examples. However, if one’s aim is to establish a taxonomy for
the structure of metaphorical language which will be applicable in principle to
any genre or instance of discourse, then examples from more challenging types of
discourse, such as literary prose and poetry, also have to be dealt with. Some less
straightforward illustrations are offered below, from the Maitland (1990) and the
Rushdie (1995) texts discussed in more detail by Heywood et al. (2002), in order
to show that we have to introduce some further assumptions before we can
actually develop such a taxonomy.

(5)

It had been … a time when her body and her mind fitted together … tidily
and wholly.

(6)

Only in her anger could she drown out the dark shadow that pity and guilt
cast over her.

(7)

Abraham in a feverish rage spent hours crawling across the floor in search
of magic.

(8)

Their women, far from being grateful, turned on them, snarling, in late night
conversations telling them to shut up.

(9)

She grew moody and inward and sat behind her dustlines, besieged within
her own fortifications.

Although (5) through (9) are all complete sentences each containing more than
just two metaphorically used words, there are clear differences between them. For
instance, the most obvious distinction is probably the one between (9), which
involves clausal co-ordination, and the others, which do not. Intuitively there is a
lot to be said for treating the two co-ordinated clauses of (9) as independent units
of discourse. If this is done then ‘She grew moody and inward’ will count as one
linguistic expression with two metaphorically used words, grew and inward, and
will thereby become comparable to either (2) or (4) above, depending on the
perceived semantic relation between grow and inward.

Fortunately there is nothing either arbitrary or controversial in our isolating a

potentially sub-sentential unit of discourse, for the basic distinction employed by
many discourse analysts is precisely that between the sentence and the text span
or text unit, as in Mann and Thompson (1988). We take a text or T-unit to be a
semi-independent clause, a category that covers main clauses, matrix clauses plus
their embedded clauses, non-restrictive relative clauses and most adverbial
clauses. Appositions can also function as T-units. Such T-units form semantic
wholes in that they present a state of affairs as relatively integrated and separated.
Thus, despite their complexity, (5) to (7) each count as one T-unit, whereas (8)
and (9) do not. (See as follows for a display of the T-unit structure of [8] and [9].)

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

(8), with its verbless clause and its two non-finite adverbial clauses, counts as
four; and (9), with its two co-ordinated clauses, the second of which itself
contains a non-finite adverbial clause, counts as three. (5), (6) and (7), however,
each count as only one T-unit, though they do contain dependent clauses. This is
because in the case of (5) and (6) these are restrictive relative clauses introduced
by a time when in (5) and the dark shadow that in (6). Such clauses do not
possess the semantic independence required to function as a T-unit, since they
help to identify the primary referents of the noun phrases in which they occur. In
the case of (7), there is an adverbial adjunct that presents essential information
about how the hours were spent by the ‘he’ in question. Restrictive relative
clauses and embedded clauses, whether functioning as subjects, objects or
complements, and some finite adverbial clauses, lack the requisite semantic
independence to function as T-units. The two non-finite adverbial clauses in (8),
‘snarling’ and ‘in late night conversations telling them to shut up’, however, are
presented independently, marked off by commas; they are thus T-units. Each of
them has one metaphorically used word and so is comparable to (1) earlier. The
non-finite adverbial clause in (9), ‘besieged within her own fortifications’, has
two metaphorically used words and so is comparable to (2), if as seems
reasonable the two words are taken to denote in the same source domain (cf.
Heywood et al., 2002).

We display the T-unit constituency of (8) and (9) below in order to make things

completely clear:

(8)

Their women, far from being grateful, turned on them, snarling, in late night
conversations telling them to shut up.
8a. Their women turned on them
8b. far from being grateful
8c. snarling
8d. in late night conversations telling them to shut up

(9)

She grew moody and inward and sat behind her dustlines, besieged within
her own fortifications.
9a. She grew moody and inward
9b. and sat behind her dustlines
9c. besieged within her own fortifications

We have now specified a surface linguistic unit of discourse, the T-unit, that

should allow a range of quantifiable comparisons to be made at the level of the
linguistic expression between different genres and instances of discourse
containing metaphorical language. The level of the linguistic expression is
however only one of our three analytic levels. Although one may for a whole
range of purposes want to produce quantifiable generalizations at this level, by
itself it is relatively opaque. It has to be related to its potential role in the
construction of metaphorical mappings if it is to be of real interest to the analyst
aware of the conceptual dimensions of metaphor. However, as we saw in the first

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

two articles of this issue, we cannot move straight from linguistic expressions to
metaphorical mappings. We also need to work with the intermediate level of the
proposition. This becomes immediately clear when we display T-unit structures as
we have done previously. For, although each of the T-units shown earlier
expresses at least one main proposition, when the clauses involved are elliptical or
reduced, as they often are, only propositional analysis can make the underlying
propositional structure of predicate and argument(s) explicit. To spell this out by
means of an example, here is the propositional analysis of (8):

(10) S8A P1

(TURN WOMEN)

(ON P1 MEN)

(POSSESS MEN WOMEN)

S8B P1

(FAR-FROM P2)

(MOD WOMEN GRATEFUL)

S8C P1

(SNARL WOMEN)

S8D P1

(TELL WOMEN P3)

(TO P1 MEN)

(HAVE-TO P4)

(SHUT-UP MEN)

(TIME P1 CONVERSATIONS)

(MOD CONVERSATIONS LATE-NIGHT)

We now have a list of T-units of equal value which will facilitate the examination
and comparison of metaphorical structures across T-units within and between
sentences, together with a propositional analysis of these T-units. Lists of T-units
could form the input for an analysis of overall text structure in the fashion of
Rhetorical Structure Theory (Mann and Thompson, 1988, 1992), so that local
propositional analysis could be complemented by global coherence relation
analysis. The latter may be of importance when one wants to examine series of
metaphors across extended stretches of text.

3 Patterns of metaphorical structure

Now that we have established which units of linguistic expression may be
compared to each other when it comes to metaphorical structure, we can look at
the distinct patterns exhibited by these units. There is of course considerable
formal linguistic variation across T-units, as can be seen from (8). This means that
there is also considerable variation in their propositional structure, as can be seen
from (10). What unifies this variation in formal and propositional structure is the
fact that each T-unit is, by virtue of its relative semantic integrity, a distinct unit of
discourse that can be separately analysed and classified with regard to its
metaphorical properties.

Let us now re-examine (1) and the instances from (5) through (9) that can be

compared to it:

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

(1)

Rachel seemed able to absorb everything that Phoebe tried.

(8)

Their women turned on them
far from being grateful
snarling

Three of the above four T-units contains a single metaphorically used word. (We
are ignoring the complexities raised by the potential metaphoricity of on in (8a)
and far in (8b) for ease of exposition.) Each of these words in turn expresses a
single element of propositional structure, either a predicate or an argument. (We
shall in future use the phrase ‘semantic item’ to refer neutrally to either a
propositional argument or predicate.) In fact in this case all three semantic items
are two-place predicates (see 10 shown earlier).

There is in principle no limit to the number of propositions a T-unit may

express, no limit, that is, to its propositional complexity. Propositional complexity
when present is due to the presence of various kinds of subordinate proposition.
Yet for all its grammatical and propositional complexity, (1) is from the point of
view of its metaphorical structure very simple. Its single metaphorical word
expresses a single semantic item, this item just happening to be part of an
embedded proposition. The same is true of (8a) and (8c). In each case a single
metaphorical word expresses a single semantic item in the underlying
propositional structure. The presence of a single metaphorical word signals the
presence of an underlying metaphorical proposition, which provides in turn the
basis for constructing a cross-domain mapping, as may be seen by following the
five-step procedure proposed by Steen (1999). The proposition is able to do this
because the semantic item that the metaphorical word expresses relates to the
source domain of the potential cross-domain mapping.

We can now go on to complicate this picture in order to arrive at a taxonomy

for metaphor that is based on propositional structure. We will deal first with two
aspects that we term multiple metaphor and complex metaphor. Examples (2) and
(9) below exemplify multiple metaphor, while (3) and (5) exemplify complex
metaphor.

(2)

He had set his mark on them.

(9)

She grew moody and inward and sat behind her dustlines, besieged within
her own fortifications.
She grew moody and inward
and sat behind her dustlines
besieged within her own fortifications

(3)

Phoebe … asked herself with a sudden rush of nostalgia.

(5)

It had been … a time when her body and her mind fitted together … tidily
and wholly.

(2), (9a), (9b) and (9c) each contain metaphorical words signalling two different
semantic items in the underlying metaphorical proposition.

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

(11)

(SET JIM MARK)

(ON P1 WOMEN)

(POSSESS JIM MARK)

S9b

(SIT-BEHIND FLORY DUSTLINES)

(POSSESS FLORY DUSTLINES)

The underlying metaphorical proposition has at least two semantic items that
relate to the potential metaphorical source domain. This is what makes multiple
metaphors multiple.

(3) and (5) above are obviously different from (2) and (9). Although they too

each contain more than one metaphorical word, the total reaching four in (5), in
both cases these metaphorical words involve two different propositions, one
superordinate and the other subordinate and dependent on a semantic item in the
superordinate proposition. In (3) this item is an argument and in (5) it is a
predicate: see S3 P2 and P3 and S5 P3, P4 and P5 in (12):

(12) S3

(ASK PHOEBE PHOEBE)

(WITH P1 RUSH)

(MOD RUSH SUDDEN)

(OF RUSH NOSTALGIA)

(TIME P2)

(FIT-TOGETHER BODY-AND-MIND)

(MOD P3 TIDILY)

(MOD P3 WHOLLY)

In terms of the linguistic surface, what is at issue is the modification of a head-
word, whether noun, verb or whatever, by one or more modifiers, whether
adjective, adverb or whatever.

We can now precisely define complex metaphor. (3) and (5) and other T-units

like them exhibit complex metaphor because they contain at least one semantic
item relating to a potential source domain occurring in a downgraded proposition
that is not co-referential with the semantic item on which that downgraded
proposition is dependent, and the semantic item on which the downgraded
proposition is dependent also itself relates to the potential source domain. Thus in
the qualifying downgraded proposition S3 P3 in (12) there are two metaphorical
arguments one of which is co-referential with the argument RUSH in the
proposition S3 P2, which is itself metaphorical. In the modifying downgraded
propositions S5 P4 and S5 P5 in (12) there is a single metaphorical argument in
each downgraded proposition, and the predicate of the superordinate proposition
S5 P3 on which S5 P4 and S5 P5 are dependent is also metaphorical. Thus, while
multiple metaphor extends the realization of a metaphor across a main
proposition, complex metaphor extends it within a proposition that is itself
dependent on a metaphorical semantic item within that main proposition.

The two remaining categories that we will incorporate into our taxonomy,

those of mixed and extended metaphor, are more familiar. It is however important

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

to show how they fit into our analytic system. A mixed metaphor in our definition
is present if within one metaphorical proposition, whether main, embedded or
downgraded, or in two metaphorical propositions one of which is superordinate to
the other that is itself downgraded and dependent on an item of the superordinate
proposition, there are two or more semantic items which relate to two or more
different potential source domains. This is what we have in (4), of which the
propositional analysis follows:

(13) S4

(REF THAT STUFF)

(OF STUFF MELODRAMA)

In S4 P2 the two arguments relate to the two different potential source domains of
‘physical substance’ and ‘drama’. Since the mixed metaphor is realized in a
downgraded proposition which is dependent on a co-referential argument in S4
P1, this metaphor is complex as well as mixed. In fact in the nature of the case a
mixed metaphor must always be either a complex or a multiple metaphor, for the
simplest kind of metaphor can with its single metaphorical word denote only in
one single potential source domain.

We can see from this that the analytic level of the proposition is vital for

establishing a taxonomy of metaphor. It enables us to see that the category of
mixed metaphor cross-cuts with other metaphorical categories, those namely of
multiple and complex metaphor. Yet of course at this stage the level of the cross-
domain mapping has also become crucial, for one can only judge mixed metaphor
to be present if one judges that there is more than one cross-domain mapping
present inside one unit.

The role of the cross-domain mapping also plays a necessary part in the

definition of extended metaphor. In our procedure it is a necessary condition of
extended metaphor being present that two or more successive metaphorical T-
units are judged to realize the same cross-domain mapping. This does not
however furnish a complete definition of extended metaphor since this condition
while being necessary is not sufficient, but space forbids a further explication of
our concept of extended metaphor.

4 Towards a taxonomy

To notice patterns of metaphorical structure is one thing, but to integrate them
into a proper taxonomy for metaphor is another. We have now provisionally
isolated a number of aspects of metaphor, but these do not seem to exclude one
other. They are potentially cross-cutting. A striking example is provided by (6),
which is repeated below for the sake of convenience.

(6)

Only in her anger could she drown out the dark shadow that pity and guilt
cast over her.

It is hopefully clear by now that the co-presence of drown out and shadow in one

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

proposition makes for both multiple and mixed metaphor. However, shadow is
also modified by dark, signalling the presence of a metaphorical downgraded
proposition dependent on the metaphorical argument expressed by shadow, which
entails complexity. (We shall temporarily leave aside the restrictive relative clause
following shadow, which contains yet a further metaphorical item, cast.) As a
result of this, this T-unit simultaneously exhibits mixed, multiple and complex
metaphor. We need to establish a form of taxonomy that recognizes our categories
of metaphor and allows them to cross-cut. We will in this section of our article
concentrate first upon the categories other than extended metaphor, then deal with
extended metaphor, and finally present our taxonomy in what is effectively a
decision procedure for metaphor classification.

The basic idea that we work with is that a T-unit consists of a main proposition

which may display various degrees of propositional complexity. Multiple
metaphor arises when a proposition contains more than one metaphorical
semantic item, as with (2) and (9a), (9b) and (9c) earlier. Complex metaphor
arises when a metaphorical item of a proposition is further developed by a
downgraded proposition which contains one or more novel metaphorical items, as
in (3) and (5). Mixed metaphor arises when multiple or complex metaphor
contains metaphorical items which may be derived from more than one source
domain, as in (4) and (6). The fact that one can have multiple and complex
metaphor at the same time is thus seen to be not at all surprising. It is simply a
consequence of the fact that if you have two or more metaphorical semantic items
in one proposition then one or more of these items may have a metaphorical
downgraded proposition dependent on it. To treat mixedness as a third variable,
one independent of the previous two variables as they are of each other, also
makes sense. It is a semantic variable that has to do with what source domains the
metaphorical items in the propositional analysis relate to, and this is an issue that
is independent of the structural properties of these items which make them either
multiple or complex metaphors. In other words, it is not redundant to say that the
combination of drown out and shadow is both multiple and mixed, for multiple
metaphor does not have to be mixed and mixed metaphor does not have to be
multiple – it may also be complex. Nor is it redundant to say that dark shadow is
complex and not mixed, for complex metaphor may be mixed and mixed
metaphor need not be complex – it may also be multiple.

The three variables of multiple, complex and mixed metaphor hence provide

the basis for three distinct types of observations about the conceptual structure of
metaphor when it is analysed by means of propositionalization. It is important to
notice that the variables are labelled by the marked terms of their measuring
scales: there is also metaphor which is not multiple, not complex and not mixed.
It is convenient to coin terms for these categories of metaphor, too, so that a
metaphor may always be described by a positive term. From now on, we will
speak of singular as opposed to multiple, simple as opposed to complex, and pure
as opposed to mixed metaphor.

All three categories dealt with earlier apply to the propositional structure of

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

metaphor within T-units. Some metaphors however may extend across T-units. A
candidate for extended metaphor would be ‘their women turned on them,
snarling’ in (8), where the metaphorical image of a dog turning on a human is
further elaborated in the next T-unit by means of ‘snarling’. When a metaphor is
not continued in the next T-unit, it may be called restricted.

A schematic representation of our taxonomy would be as follows (see Figure

1). For any T-unit containing one or more metaphorically used words, one may
decide whether the metaphorical mapping is continued in the next T-unit or not
(extended or restricted). Whatever the answer to this question, one may then
proceed to examine whether the main metaphorical proposition expressed by the
T-unit contains only one or more than one metaphorical semantic item (multiple
or singular). Whatever the answer to this question, one may then continue to ask

RISP ET AL

Language and Literature 2002 11(1)

Figure 1 Taxonomy of metaphorical mappings within T-units

Pure

Mixed

Pure

Mixed

Pure

Mixed

Pure

Mixed

Pure

Mixed

Pure

Mixed

Simple

Complex

Simple

Complex

Simple

Complex

Simple

Complex

Singular

Multiple

Singular

Multiple

Restricted

Extended

T-unit containing
metaphorically used
concept(s)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

whether the main metaphorical proposition has a metaphorical item with a
downgraded proposition dependent on it that itself contains at least one
metaphorical semantic item (complex or simple). Finally, if the answer to either
or both of questions 2 and 3 has been positive, then it has to be decided whether
the complex and/or multiple metaphorical mapping is pure or mixed. There are
only two out of 16 possible combinations of values of our four basic variables
which are logically impossible: one cannot have mixed metaphor in the case of
the combination of simple and singular metaphor, be it restricted or extended.

5 The unit of metaphor, or the problem of quantification

The phenomena of complex, multiple, mixed and extended metaphor may be
intuitively recognized and related to the structure of the propositional analysis of
T-units, but this does not yet offer us a precise enough instrument for the exact
description and measurement or quantification of metaphor in discourse. There
are several ways in which metaphors may be counted, and the four metaphor
properties that we have distinguished may be related to these various levels of
metaphor measurement with more or less felicity.

For instance, when we return to the complicated example of (6), there are

several problems to be noticed:

(6)

Only in her anger could she drown out the dark shadow that pity and guilt
cast over her.

T-unit (6) is not merely restricted, complex and mixed, but also:

(1)

its underlying main proposition contains two, and only two, metaphorically
used semantic items and that these are the predicate expressed by drown out
and the argument expressed by shadow (multiple metaphor property);

(2)

only one of its semantic items, the argument expressed by shadow, has a
downgraded metaphorical proposition dependent on it (complex metaphor
property);

(3)

shadow in fact has two, and only two, downgraded metaphorical
propositions dependent on it, that expressed by the adjective dark and that
expressed by the relative clause ‘that pity and guilt cast over her’ (complex
metaphor property);

(4)

the metaphorically used items are related to three source domains: the
predicate expressed by drown out is related to the source domain of liquids;
the complex structure of the argument expressed by shadow and the
downgraded propositions expressed by dark are related to the source
domain of light; and the term cast is related to the domain of movement
(mixed metaphor property).

However, it should be observed that for some research purposes it may be

convenient to ignore the additional details and label T-units for any incidence of

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

any of the marked values of the variables we have discussed until now. As a
result, (6) would be described as exhibiting restricted, complex, multiple and
mixed metaphor. This would be a relatively gross level of measurement if it is
compared with some form of counting metaphors based upon a full propositional
analysis, but it might be useful and give a speedy answer to some research
questions about the metaphorical nature of T-units as T-units in discourse.

Another alternative is suggested by the propositional procedure we have taken

to metaphor identification and analysis. This alternative approach would attempt
to tag propositions rather than T-units for the variables of multiple, complex,
mixed and extended metaphor. If this were successful, it would be a method for
classifying metaphorical ideas, operationally defined as propositions containing
metaphorically used concepts, according to their structure and content along the
lines suggested above. Thus, let us assume that the propositional structure of the
bulk of (6) is (14):

(14) P1

(POSSIBILITY P2)

(DROWN-OUT PHOEBE SHADOW)

(IN P2 ANGER)

(MOD P3 ONLY)

(POSSESS PHOEBE ANGER)

(MOD SHADOW DARK)

According to our previous analyses, P2 would present a metaphorical proposition
that is restricted (it is not continued in the next T-unit), complex because one of
the metaphorical items is further modified (SHADOW, by DARK in P6), multiple
(DROWN-OUT plus SHADOW) and mixed (idem). The analysis of P6 as an
independent metaphorical idea unit, however, would seem less easy to solve.
According to the semantic spirit of our procedure, it is not multiple itself,
although it does look as if it is. It is not complex itself, either, but it is a part of a
complex metaphor, because of its relation to P2. And it is possible to say that it is
not mixed but pure, even though it does not sit easily in the categories of complex
and multiple metaphor.

A similar story could be told about counting metaphors as words or their

associated concepts. One could then adopt the convention of speaking of
metaphorically used concepts which participate in complex, multiple or mixed
metaphors, however these are defined, as propositions, T-units, or in yet another
way. This operational definition has been used by Steen (submitted) in a study of
the relation between metaphor recognition and metaphor properties. Important
parts of the variance occurring in the underlining data of metaphorically used
words turned out to be predictable by a handful of structural variables, including
the participation of the metaphorically used words in multiple metaphor. Yet one
other way of defining the unit of metaphor will be described later, but the point of
the present paragraphs has been to emphasize that it is not the only and not ‘the’
correct way to quantify metaphor.

Apart from looking at metaphors as concepts, propositions or T-units, which

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

are three basic and general levels of the conceptual structure of any analysis of
discourse, we may also define metaphors as conceptual mappings that cut across
these levels of discourse organization. This is the abstract conceptual approach
that we have referred to in the second article of this special issue (Steen, 2002),
and it is the method used by most cognitive linguists, who follow Lakoff and his
associates. In metaphor as a conceptual mapping, one only needs to assume the
existence of two independent conceptual domains, and it is an empirical issue
how many of the abstract conceptual correspondences are realized in a discourse
by means of metaphorically used words.

Probably the grossest quantitative question that can be asked of a text with

regard to metaphor is how many metaphors it contains. As can be seen from the
above, there is no single automatic answer to this question. Is one going to count
metaphorical concepts, metaphorical propositions, metaphorical T-units or
metaphorical mappings? Clearly, the answer to this question will depend upon the
interests of the analyst on any given occasion. Perhaps also on some occasions the
nature of the text involved might make one choice seem particularly appropriate.
But there can be no single absolute answer to the question of what to count as the
unit of metaphor.

The option for the unit of metaphor to be counted in a text that we have

adopted for illustration in this article is the number of cross-domain mappings.
This can be determined as follows. Every metaphorical T-unit which is neither
mixed nor extended will signal one such mapping, and so will count as one. Every
metaphorical T-unit that is mixed will have assigned to it the number of the source
domains which it can potentially activate. Any sequence of T-units which is
extended will count as one, no matter how many members it may have.

6 Sketch for an application

There is no space here for an exhaustive application of our procedure to a text or,
more interestingly, to a set of texts. However, it seems worthwhile to give a
relatively informal account of its quantitative application to the two texts we have
worked on so far, the excerpts from Sara Maitland’s Three Times Table and
Salman Rushdie’s The Moor’s Last Sigh taken from the Lancaster University
Speech and Thought Presentation Corpus (see Heywood et al., 2002 for further
details). Since the Maitland text is a piece of popular fiction and the Rushdie a
piece of serious ‘literary’ fiction one could expect there to be differences between
the metaphorical patterning of the two texts, and in fact our procedure does
indeed indicate striking differences between them.

The Maitland text is approximately 2420 words long and, using our option of

defining the number of metaphors as the number of cross-domain mappings, it
contains 181 metaphors. The Rushdie is approximately 2156 words and contains
144 metaphors. This gives a metaphor/word ratio of 7.5 percent for the Maitland
and 6.7 percent for the Rushdie. A chi-square test shows that there is no

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

significant relation between the variables of number of words and number of
metaphorical mappings (X

(1)

= 0.85, n.s.). A mere counting of the numbers of

metaphors does not therefore reveal much in the way of difference between the
two texts.

When, however, we apply our taxonomy to the two texts, we obtain the

following findings. Although in both texts a large majority of the metaphors have
completely ‘negative’ values for our four categories, that is, they are restricted,
singular, simple and pure, Rushdie has a somewhat smaller proportion than
Maitland. Of Maitland’s metaphors, 70–75 percent are restricted, singular, simple
and pure, while the figure for Rushdie is 65.7 percent. However, this difference is
not statistically reliable (X

(1)

= 0.63, n.s.).

When one begins to look in detail at the nature of the metaphors in the two

texts that are multiple, complex or mixed, or some combination of these, one
remarkable difference is revealed. Of Rushdie’s metaphors, 17.4 percent are
multiple, while the figure for Maitland is 7.1 percent; this difference is
statistically reliable (X

(1)

= 7.09, p <0.01). It is probably also worth noting that

while all but one of Maitland’s multiple metaphors have only two metaphorical
semantic items, three of Rushdie’s have three. The ‘triple/double’ ratio for
Maitland is 8.3 percent while that for Rushdie is 13.6 percent. The differences
with regard to complex and mixed metaphor taken in isolation are less important,
though Rushdie does score higher on both these categories: 4.7 percent of
Maitland’s metaphors are complex and 7.7 percent are mixed, while 6.3 percent of
Rushdie’s are complex and 10.4 percent are mixed. However, there are no
statistically reliable relations here.

As already pointed out, one metaphorically used word may be classified from

more than one perspective. For instance, it is worth noting that we find in Rushdie
three instances of mixed metaphor combined with an additional independent
multiple or complex metaphor, while there is only one example of this, the
already cited (6), in Maitland. An example of Rushdie’s combination of mixed
metaphor with one or more independent complex or multiple metaphors where
there is a more obvious surface linguistic, as well as underlying semantic,
complexity is T-unit (15):

(15) which had managed by an act of will to wrench itself free of its fixed orbit

This is a non-restrictive relative clause modifying the noun body in ‘a heavenly
body’:

(16) P1

(MANAGE BODY P4)

(BY P1 ACT)

(OF ACT WILL)

(WRENCH BODY P5)

(MOD BODY FREE)

(OF FREE ORBIT)

(MOD ORBIT FIXED)

RISP ET AL

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

The combination of P1, P2 and P3 produces a complex metaphor with two
downgraded propositions. P4 gives rise to a multiple metaphor with two
metaphorical semantic items, the predicate WRENCH and the embedded
proposition P5. The combination of P5, P6 and P7 again gives rise to a complex
metaphor with two downgraded propositions. The T-unit as a whole is a mixed
metaphor drawing on three, and not just two, source domains, those of handed
action, concrete astronomical reality and abstract geometry, an orbit being an
abstract and not a concrete entity. There is nothing like this in the Maitland text. It
thus seems that our taxonomic model together with the method of propositional
analysis that underlies it is sensitive to differences of metaphorical patterning in
the two texts.

Acknowledgement

The third author wishes to thank the Netherlands Organization for Scientific Research and the British
Council for travel grant BR 30–539, which facilitated preparation of this article.

References

Crisp, P. (2002) ‘Metaphorical Propositions: A Rationale’, Language and Literature 11(1): 7–16.
Heywood, J., Semino, E. and Short, M. (2002) ‘Linguistic Metaphor Identification in Two Extracts

from Novels’, Language and Literature 11(1): 35–54.

Maitland, S. (1990) Three Times Table. London: Chatto and Windus.
Mann, W.C. and Thompson, S.A. (1988) ‘Rhetorical Structure Theory: Toward a Functional Theory of

Text Organization’, Text 8: 243–81.

Mann, W.C. and Thompson, S.A. (eds) (1992) Discourse Descriptions: Diverse Analyses of a Fund-

raising Text. Amsterdam: John Benjamins.

Rushdie, S. (1995) The Moor’s Last Sigh. London: Jonathan Cape.
Steen, G.J. (1999) ‘From Linguistic to Conceptual Metaphor in Five Steps’, in R.W. Gibbs Jr and G.J.

Steen (eds) Metaphor in Cognitive Linguistics, pp. 57–77. Amsterdam: John Benjamins.

Steen, G.J. (2002) ‘Towards a Procedure for Metaphor Identification’, Language and Literature 11(1):

17–33.

Steen, G.J. (submitted) ‘Metaphor Properties and Metaphor Recognition’, Journal of Pragmatics.

Addresses

Peter Crisp, Department of English, Chinese University of Hong Kong, Shatin, New Territories, Hong

Kong. [email: b276780@mailserv.cuhk.edu.hk]

John Heywood, Department of Linguistics and Modern English Language, Lancaster University,

Lancaster LA1 4YT, UK. [email: j.heywood@lancaster.ac.uk]

Gerard Steen, Department of English, Free University, P.O. Box 7161, 1007 MC Amsterdam, The

Netherlands. [email: gj.steen@let.vu.nl]

ETAPHOR IDENTIFICATION AND ANALYSIS

Language and Literature 2002 11(1)

by Paolo Magagnin on February 8, 2007

http://lal.sagepub.com

Downloaded from

Wyszukiwarka

Podobne podstrony:
Cultural Identity and Globalization Multimodal Metaphors in a Chinese Educational Advertisement Nin
Metaphor interpretation and emergence
Identifcation and Simultaneous Determination of Twelve Active
Metaphor, Relevance and the Emergent Property Issue
Am I Who I Say I Am Social Identities and Identification
Metaphor interpretation and emergence
Identity and the Young English Language Learner (E M Day)
Cruz Jansen Ethnic Identity and Racial Formations
Identification and fault diagnosis of a simulated model of an industrial gas turbine I6C
Cruz Jansen Ethnic Identity and Racial Formations
19657249 subaltern studies collective schizophrenic identity and confronted politics a revision
Generic Identity and Intertextuality
0415165733 Routledge Personal Identity and Self Consciousness May 1998
Akerlof Kranton Identity and the economics of organizations
identifying and discussing feelings 201
Variation in NSSI Identification and Features of Latent Classes in a College Population of Emerging
Susan B A Somers Willett The Cultural Politics of Slam Poetry, Race, Identity, and the Performance
Work motivation, organisational identification, and well being in call centre work

więcej podobnych podstron