You are on page 1of 11

56 FRANK BOERS

Kovecses, Zoltan. 1995. "The Container Metaphor of Anger in English, Chinese,


Japanese and Hungarian". From a Metaphorical Point of View: A Multi-
disciplinary Approach to the Cognitive Content of Metaphor, ed. hy
LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS
Zdravdo Radman, 117-147, Berlin and New York: Walter de Gruyter.
Lakoff, George. 1987. Women, Fire and Dangerous Things: What Categories
Reveal about the Mind. Chicago and London: University of Chicago Press;
GERARD STEEN
-------- 1990. "The Invariance Hypothesis: Is Abstract Reasoning Based on Image~
Tilburg University and Free University Amsterdam
Schemas?" Cognitive Linguistics 1. 5-38.
Miller, Arthur I. 1995. Imagery and Metaphor: The Cognitive Science Con"
nection, ed. by Zdravko Radman, 1995. 199-224.
does the cognitive linguist get from linguistic metaphor to conceptual
Rigotti, Francesca. 1995. The House as Metaphor, ed. by Zdravko Radman, 4191-
~taphor? Is there a procedure for the determination of conceptual metaphor when
446. Berlin and New York: Walter de Gruyter. ~taphoricallanguage has been encountered? These are the questions that are
Roediger, Henry L. 1980. "Memory Metaphors in Cognitive Psychology".
idressed in this chapter, which aims to build a bridge between linguistic and
Memory & Cognition 8. 231-246. onceptual metaphor by proposing a series of five analytical steps. Together they
the beginning of a procedure for conceptual metaphor identification in

The procedure is meant to constrain the relation between linguistic and


nceptual metaphor. It has sometimes remained an act of faith that particular
..taphors in language reflect particular metaphors in thought. This does not mean
there is no linguistic support for the existence of conceptual metaphors. And
there are many clear cases in which the name of a particular conceptual
is used in a linguistic expression, as can be demonstrated by a brief
at the by now classic list of references Lakoff & Johnson (1980), Johnson
Lakoff (1987), Turner (1987), Lakoff & Turner (1989), and Lakoff(1993).
these clear cases serve the purpose of demonstration; they have not been
~tematIcally and exhaustively collected. from large stretches of discourse, but
have been selected for th~ir persuasive power. Now that the theory of
l~eptual metaphor has been firmly established as one important component of
theory of metaphor, providing one of the main inspirations to cognitive
~gulSt1CS as a general approach to language, it is time to reverse the perspective.
the question arises how stretches of discourse can be said to express
conceptual metaphors as opposed to others, and this is a difficult issue.
presupposes a generally accepted procedure of deriving conceptual metaphors
lipguistic metaphors encountered in on-going discourse, and that is currently
available.
Most readers will be familiar with some of the examples of metaphorical
~rrespondences between conceptual domains such as the following:
58 GERARD STEEN FROM LINGUISTIC TO CONCEPtuAL METAPHOR IN FIVE STEPS 59

THE LOVE-AS-A-JOURNEY MAPPING particular expression is to be counted as a one-shot metaphor or as a systematic


The lovers correspond to travelers. metaphor: he or she first has to identify metaphorical expressions and determine
The love relationship corresponds to the vehicle. what the conceptual nature of the metaphorical expression in question is. Only
The lovers' common goals correspond to their common destinations one the once this has been achieved, can the metaphorical concept be examined as to its
journey. possible relation(s) with other metaphorical concepts, which then leads to a
Difficulties in the relationship correspond to impediments to travel.
(Lakoff,1993:207) decision about one-shot conceptual metaphoricity versus systematic conceptual
metaphoricity. Such a comparison across metaphors presupposes that the other
But from the present perspective, these are at best the output of the last step of the metaphorical concepts have also been collected from discourse analysis in the
envisaged procedure, and this would probably only hold in ideal cases. What I am same fashion. What I am focusing on, then, is the procedure for collecting such
interested in is to explicate the assumptions that lead linguists to arrive at such metaphorical concepts with the purpose of examining their systematic relations.
conceptual mappings in departing from metaphorical expressions in discourse. If one insists on regarding as conceptual metaphors only those metaphors which
This chapter is a logical reconstruction of these assumptions in an attempt to reach are systematic (as opposed to one-shot metaphors), which I do not, then a sixth
agreement about the steps that are inevitable when one goes from linguistic to step will have to be added to the procedure, saying that the output of the first five
conceptual metaphor identification. steps is to be compared across large numbers of metaphors in order to establish
It is noteworthy that this explication can be related to a number of more or less systematic groups of metaphorical concepts, labeling the largest
theoretical issues which were previously discussed in the seventies, before the systematic groups as conceptual metaphors.
advent of conceptual metaphor theory as we now know it (Cohen & Margalith, I have to add one more caveat from the beginning. I wish to emphasize that
1972; Van Dijk, 1975; Reinhart, 1976; Cohen 1993; Miller 1993). In retrospect, I am dealing with metaphor analysis, not metaphor understanding. Metaphor
most of these references can be seen as attempts to make the jump from linguistic analysis is a task for the linguist who wishes to describe and explain the structure
to conceptual metaphor in one way or another, but they failed to do so in an and function of language. Metaphor understanding is a cognitive process which
optimal manner because of the lack of a well-developed conceptual theory of is the object of investigation of psycho linguists and discourse psychologists who
metaphor. The time is now ripe to return to these issues in order to put conceptual are conducting behavioral research. This chapter does not deal with behavior. This
metaphor theory on a firmer linguistic footing. It is ironic that cognitive linguists does not mean that it cannot make use of theories of metaphor understanding for
are going out of their way to show that linguistic metaphor is fundamentally the identification of specific stages in the analytical procedure; on the contrary, it
conceptual, but that in doing so, they have neglected the method for showing how would be odd if there were no connection between understanding and analysis.
they get from linguistic metaphor to conceptual metaphor in the first place. However, metaphor analysis is a goal- and norm-related activity in the pursuit of
My recourse to these sources has one consequence which may be misleading data collection. It is the intentional technical identification of conceptual
and which has to be circumvented from the beginning. Some or most of the metaphors from metaphorical language in discourse. This chapter is concerned
examples discussed by theorists in the seventies were not of the conventional kind with a logical reconstruction of the discrete steps involved in that activity.
that have since become popular in the literature. In present-day terms they might The logical reconstruction may then be transformed into a procedure for
be seen as one-shot and often poetic metaphors rather than systematic conceptual practical use in linguistic research. From the perspective of the cognitive linguist,
metaphors. Moreover, I do not address the question whether my illustrations of who is interested in the analysis of discourse and the way it reflects concepts and
metaphor are actually found in other expressions of a similar kind, which is the cognition, it is essential that there is such a procedure for relating linguistic
generally accepted approach to establishing conceptual metaphors in cognitive metaphor to conceptual metaphor in a reliable fashion. Ultimately, the cognitive
linguistics. These may be surprising features of a chapter titled "From linguistic linguist has to begin with stretches of discourse and determine which linguistic
to conceptual metaphor in five steps." However, I believe that they are actually expressions are metaphorical and related to which conceptual metaphors, and this
immaterial to the purpose of this particular contribution, which is to reconstruct is no trivial matter. However, such a procedure is also important for constructing
how the linguist gets from linguistic metaphor to conceptual metaphor. For the link between cognitive linguistics and psycholinguistics, in which precise
methodologically speaking, the linguist has no a priori knowledge whether a descriptions of literal and nonliteral materials are needed for the development of
well-controlled linguistic stimulus materials. Manipulating texts and expressions
60 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 61

with a view to activating particular conceptual metaphors requires the same solid linguistic expression used nonliterally in the discourse. This means that the focus
foundation in linguistic methodology. expression activates a concept which cannot be literally applied to the referents in
the world evoked by the text. The concept RIDING-ON cannot be literally applied
1. Metaphor focus identification to the relation projected between the entities referred to by MERMAIDS and WAYES,
The first step of the procedure involves the identification of metaphorical and the concept ROYAL COURT cannot be literally applied to the entity referred to
expressions in discourse. This naturally involves the theoretically thorny issue of by LIONS. It will be noted that referents can be entities, relations, and attributes in
the definition of metaphor itself. As it is the purpose of this contribution to present some situation evoked by the discourse.
a procedure for conceptual metaphor identification, which in itself more or less The twin concept related to focus is frame, which I will define, with
presupposes that We know what a metaphor is, I will cut a long story short and Reinhart and others, as the immediate linguistic environment of the metaphor
make the following assumption. It seems best to adopt the most widely accepted focus. In (1), the frame is waves, according to Reinhart (1976:385). However,
definition of metaphor that is currently available, the Lakoffian one of metaphor Reinhart does not explicate the linguistic frame of (2). This is probably because
as a set of correspondences between two conceptual domains, with linguistic (2) is a special class of metaphor: the focus is not non-literal in relation to the rest
metaphor deriving from conceptual structures. The presence of two domains is of the linguistic expression, is going to hunt; there is no semantic tension between
intended to capture the fact that we are dealing with nonliteral similarities between focus and frame. The kind of metaphor exemplified by (2) is purposefully left
entities and relations at some level of the analysis, which rules out other types of aside in Reinhart's attempt at ordering the theoretical concepts for the analysis of
mappings like metonymies. metaphor. As we shall see below, a cognitive linguistic approach which includes
The first step consists of identifying metaphorical expressions in discourse, linguistic and conceptual as well as other discourse aspects of analysis is better
and we now need to become more precise about the nature of this operation. For equipped to handle these issues.
it is not true that identifying metaphorical expressions is tantamount to identifying There are some important issues in Reinhart's discussion of focus and
linguistic metaphors. I will now show that when expressions are identified as frame, as well as of other notions like tenor and vehicle, and some of them will
metaphorical, it is the focus of the metaphor that we are dealing with, and this is return later. But our present concern is the first step in the conceptual metaphor
only one part of the complete metaphor. In effect, it depends on a number of identification procedure. It may now be appreciated more fully that the first step
factors what one will call the complete metaphor in the first place. This observa- is largely concerned with metaphor focus identification, not linguistic metaphor
tion needs to be explained with reference to a number of familiar but sometimes identification as a whole. The reason is this. Many metaphor foci may be related
elusive concepts in metaphor theory. to a literally used concept which is explicitly expressed in the metaphor frame, as
One of the more interesting discussions in this respect is Reinhart (1976), in the case of (1); however, there is also a good number of metaphor foci located
who has aligned some of the classic theories of metaphor such as Richards (1936), in metaphor frames without a linguistic expression of the literal concept of the
Beardsley (1958), and Black (1962). The main point of interest at present is best metaphor, as in (2). When both literal and nonliteral concept are present in the
introduced by comparing the following metaphors, discussed by Reinhart (1976): frame, it is possible to identify a complete linguistic metaphor in the first step of
the analysis. However, when the literal concept is not expressed in the frame, the
(1) I have seen the mermaids riding seawards on the waves (T.S. Eliot, 'The love song linguistic metaphor cannot be identified in the first step; then it is only the focus
of J. Alfred Prufrock') that is identified in the first step. These are implicit metaphors (Steen 1999) and
they need explication through propositional analysis, which clarifies to which
(2) The royal court is going to hunt concept the nonliteral concept expressed by the focus is applied. Fdr instance, .
words like scene, shit, heat, and so on are often used metaphorically without the
As an aside, the actual line from Eliot does not contain 'the mermaids' but 'them' , literal concept to which they are applied being expressed expiicitly in the
but Reinhart has explicated the anaphoric pronoun for expository purposes. discourse. As it is the metaphor focus that can be identified throughout all classes
Reinhart explains that the focus of the metaphor in (1) is riding on. The of metaphor in the first step, this is why step 1 of the procedure is called metaphor·
focus of the metaphor in (2), which is about 'lions stalking their prey' (1976:391, focus identification.
fn. 8), is the royal court. Together, (1) and (2) illustrate that the focus is the
62 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 63

2. Metaphorical idea identification presents the structural relations between the concepts contained by the idea units.
Now that we have a metaphor focus as a result of step 1, the question arises All of this is relatively uncontroversial in discourse psychology (Britton & Black
to what other kind of element the focus applies. As a matter of principle it cannot 1985; Perfetti & Britt 1995; but cf. Garnham 1996).
be the linguistic frame, as the linguistic frame merely provides a linguistic The metaphorical idea that is present in T.S. Eliot's line is found in
background in which the focus is a nonliteral expression which mayor may not proposition P2. The nonliteral part of the metaphorical proposition is the concept
stand out as the case may be. In other words, the focus is not a focus on account RIDE-ON, which functions as a predicate, and it can be seen to apply to two other
of its relation to the linguistic frame, as we have observed, for it may even exhibit concepts, MERMAIDS and WAYES. They are conceptualizations of the literal
a literal relation to the frame, as was shown by (2). What does make a focus into referents in the projected text world about which something is said in a metaphori-
a focus is the fact that it expresses a concept which is to be related to another cal manner, namely that they are in a relation ofthe one 'riding on' the other. This
concept to which it cannot be applied in a literal fashion: 'riding on' cannot be relation could also have been conceptualized in a literal manner, for instance by
literally applied to 'mermaids' doing something to 'waves', and 'the royal court' means of FLOATING, which would have yielded a literal expression. This is also
cannot be literally applied to 'lions'. accounted for by Reinhart's definition of focus:
The other, literal, concept has been variously referred to as the tenor or the
topic or the principal subject of the metaphor, but I will call it the literal part of the Given a metaphorical expression Fj[Ej]
metaphorical idea. As not all literal parts of metaphors are explicitly expressed in Ej is the focus if it is possible to substitute Ej for Ej, so that Fj[Ej ] is a literal
discourse, as is the case for (2), they sometimes have to be inferred. This is why expression and Fj[Ej ] is similar in meaning to Fj[EJ (1976:391)
metaphor identification is a matter of concepts, propositions, and reference. As
these are general aspects of discourse analysis which are not limited to metaphor, Whereas RIDING-ON does not refer literally to the presumed relation between the
and propositional analysis was specially designed to cater for them, it is now time entities of mermaids and waves in the projected textworld, FLOATING does and
to turn to propositional analysis. Propositional analysis was developed independ- produces a similar meaning.
ently of the study of metaphor and also aims to bridge the gap between discourse The conceptual basis of this approach is clarified by considering the other
and conceptualization (see the contributions to Britton & Black, 1985; in metaphor and its propositional analysis:
particular Bovair & Kieras, 1985).
Consider the following propositional analysis of (1), a personal, notational (4) The royal court is going to hunt
variant of mine of the technique used by Bovair & Kieras (1985): PI (REF COURT LIONS)
P2 (HUNT COURT)
(3) I have seen the mermaids riding seawards on the waves P3 (MOD COURT ROYAL)
PI (SEEIp2)
P2 (RIDE-ON MERMAIDS wAYES) The crucial proposition is PI, where it is clarified that one concept in S2, COURT,
P3 (DIRECTION p2 SEAWARDS) refers to another concept, LIONS, which is available from the previous (or
~911owing) discourse, be it co-text or context. This explicit and immediate form
This analysis of (1) is maximally consistent with the one of Reinhart (1976), even Qf reference assignment between arguments is standard procedure in Bovair &
though I can see that at least one other analysis is possible, namely one in which Kieras' (1985) method of propositionalization, and is aimed at ensuring referential
P2 merely consists of (RIDE MERMAIDS) and at). additional P4 is needed to capture coherence between concepts in consecutive sentences, especially when different
that (ON p2 WAYES). However, since this difference is immaterial to my argument, expressions are used about the same entities in a projected text world. Of course
I will not unnecessarily increase the complexity of the exposition and leave this 'tthis is precisely what is needed to solve the problem of cases like (2) for the theory
point aside. t~f metaphor: in order to be able to interpret the main idea unit P2 correctly, it is
Irrelevant details aside, (3) is a linearly and hierarchically ordered list of . 'fji\st necessary to clarify potentially confusing elements such as 'court' when there
propositions, each consisting of a predicate and one or more arguments. The list ,,110 literal entity of that kind in the projected text world. This is the function of
captures the meaning of (1) as a series of minimal idea units, or propositions, and which thereby automatically identifies a complete metaphor: it relates the
!Ii
i
i

64 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 65

nonliteral concept COURT to the literal concept it is supposed to apply to, LIONS, the idea unit of the proposition rather than some rank of linguistic form, be it the
by means of the specially designed predicate 'REF' , and reveals that the metaphor word, the phrase, the clause, or the sentence. The complete metaphor is only
of (2) is Pl. All of these observations are intended to capture the propositional identified when the appropriate literal and nonliteral concepts in the proposition
structure of the language: they are not meant to be read as claims about compre- have been identitied. Metaphor identification, even linguistic metaphor identifica-
hension or other forms of processing. tion, is fundamentally a matter of conceptual analysis. It deals with the concepts
Propositional metaphor analysis can apply to all kinds of metaphors. It also referred to by the words and is at the core of a functional approach to language.
lays bare how metaphors can differ from each other with respect to important What is more, the complete metaphor is not always expressed as a complete
dimensions of conceptual structure. For instance, Miller (1993 :387) discusses the metaphor in the surface of the discourse: there are metaphors which are only
metaphor a watchdog committee. I did not find it coincidental that he presented signalled by means of their focus expression, and their literal part has to be
the metaphor in precisely this form, and ran an automatic Microconcord search of inferred by means of propositionalization (implicit metaphors). Propositional
the Times corpus of approximately 1,000,000 words coming with that program. metaphor analysis hence also throws into relief the nature as well as the limited
There were 10 occurrences of watchdog in all, and they were all metaphorical in role of the notion of frame: frames are nothing but the immediate linguistic
the above sense. What is interesting is the formal variation between the instances. environment of the focus and hence do not always help in setting off the focus as
There was one case of Miller's structure, a watchdog organisation. It exemplifies a nonliteral expression, as in (2). Frames have to receive a semantic and pragmatic
an explicit metaphor, because it contains both the literal and the nonliteral concept interpretation in terms of reference and intentions before metaphorical idea
of the underl:ying proposition. It is also a reduced metaphor, in that the linguistic identification can succeed, and this requires going beyond the surface of the
structure is not equivalent to a proposition itself, but is a nominal phrase. (A full frame: metaphorical idea identification (step 2) has to follow after metaphor focus
version of this metaphor would be something like: the committee is a watchdog, identification (step 1). If many linguists were still uncomfortable with sQch an
which is not likely to be found in genuine discourse, as is shown by the results of approach in the seventies, the advent of discourse analysis, cognitive linguistics,
this search.) The metaphor is also simple, in that there is no additional material and related functional studies of language use is now alleviating this problem.
attached to the nonliteral concept itself. This should be contrasted to the following Propositional analysis also raises an interesting question about the scope of
class: the National River Authority, the new watchdog/or the water industry, and the frame: in Reinhart's analysis of (1), the frame is riding X, which in effect is the
the Audit Commission, the local authority financial watchdog. In the latter case, next grammatical level up in comparison to the focus (the waves). However, as is
the concept WATCHDOG is modified by additional concepts, such as NEW, turning shown by the propositional analysis, other reaches of the frame may be imagined
the metaphorical focus into a complex structure needing more than one proposi- as well, ranging from the clause through the utterance to the sentence:
tion. Another class of metaphor may be illustrated by a national watchdog and an
independent watchdog: these are implicit metaphors, because the literal concept (5) a. riding on the waves
to which WATCHDOG is applied is not expressed in the text; and they are complex b. riding seawards on the waves
metaphors, because WATCHDOG does not stand by itself. Half of the ten watchdog c. the mermaids riding on the waves
metaphors were of this kind. The only example of an implicit and simple d. the mermaids riding seawards on the waves
metaphor was found in a headline: watchdog may be too fierce; contextual e. I saw the mermaids riding seawards on the waves
assumptions about the use of words in headlines apparently have relaxed the need
This is no mere terminological or technical matter, for it determines whether one
for adding information to the nonliterally used concept WATCHDOG which was
says that (1) contains the metaphor 'riding on the waves' or, for instance,
exem.plified by all of the other cases. Propositional analysis is hence a valuable
'mermaids riding on the waves'. I would prefer the metaphorical proposition as
tool in metaphor classification and raises consequential questions for interpreta-
the indicator of the complete metaphor, as it embodies the turning point between
tion and processing (Steen to appear).
language and conceptualization, as we shall see in a moment. What is important
Propositional metaphor analysis also clarifies some interesting aspects ofthe
to point out here is that such decisions have effects on the coding and counting of
previous discussion of Reinhart (1976). In accordance with Reinhart and the
metaphors of varying linguistic structures in corpus analysis, precise operational
theorists she discusses, the use of propositional analysis emphasizes that the
nature of metaphor is conceptual and that the analyst can only access it through
66 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 67

definitions being required for the reliable annotation of data. This is precisely one elements left out, and he proposes three specific rewrite rules which transform
of the issues that a procedure for metaphor identification needs to resolve. metaphorical propositions into nonliteral comparisons. That is also why he
requires propositional analysis as providing the input for his analysis of the
3. Nonliteral comparison identification underlying comparisons (1993:375-6): metaphorical idea identification (step 2)
The output of step 2 is a proposition in which we have a nonliterally used has to precede nonliteral comparison identification (step 3). Let us examine how
concept (expressed by the linguistic focus identified in step 1) that is related to one this works in practice.
or more literally used concepts identified in step 2 which evoke the relevant literal The metaphorical proposition of (1), (RIDE-ON MERMAIDS WAVES), may be
referent(s). The literally used concept(s) may be explicitly expressed in the frame rewritten as a nonliteral comparison with the following structure:
or may have been inferred by the analyst from co-text or context. The input of step
3 is hence a metaphorical idea in the form of a proposition with literal and (7) (RIDE-ON MERMAIDS WAVES) -> (3F) (3y,y ') {SIM[F(MERMAIDS, WAVES), RIDE-ON
nonliteral concepts. (y, y')]}
We have assumed that metaphors are sets of correspondences between
conceptual domains in which nonliteral similarity or comparison plays a pivotal A paraphrase of this formal notation of the conceptual structure of the implied
role. Therefore we need to work from the metaphorical proposition towards a comparison would be the following: there is an activity (or relation) F and two
conceptual representation of the mapping between the two conceptual domains entities y and y' such that there is a similarity between mermaids and waves' doing
involved. The next step in our procedure is hence to begin to set up the compara- F' on the one hand and y riding on y' on the other. The input of the rule is the
tive structure that is implicit in the nonliteral mapping between domains for every output of our step 2, while the output of the rule is the automatic product of
conceptual metaphor. There is an excellent source for this objective in Miller Miller's rewrite rules.
(1993), who presents a sophisticated view of the comparison theory and has been Miller adds, 'The first step in interpreting [such a] comparison would be to
unjustly neglected in conceptual metaphor theory. find appropriate values for the missing terms' (1993:384). This will be our step
Th~ essential point of Miller's contribution is that '[R]econstruction of the 4, nonliteral analogy identification. To jump ahead, in the present case we may fall
implied comparison is a critical step in understanding a metaphor' (1993:381). back on the analysis suggested by Reinhart (1976:388-90):
That metaphors imply comparisons is shown by the postulation of conceptual
metaphors like LOVE IS A JOURNEY, LIFE IS A GAME, and so on. Metaphors in (8) (RIDE-ON MERMAIDS WAVES)-> SIM[FLOAT(MERMAIDS,WAVES),RIDE-ON (JOCKEY,

discourse hence require 'reconstructing the conceptual basis of the comparison' HORSE)]

(1993:382) in order to be interpreted. This statement begs some fundamental


questions about on-line comprehension behavior, but it certainly holds true for the Filling in the missing terms in this manner will be step 4 of the procedure, and
off-line analysis of the relation between linguistic and conceptual metaphor, with fleshing out the resulting analogy into a full-blown nonliteral mapping, step 5. For
which we are concerned here. now, we have to Goncentrate on step 3, identifying the underlying comparison.
When an author uses a metaphor, Miller (1993:384-5) says, 'The claim is The derivation of the comparison statements from the metaphorical
[... ] that he had a general concept--resemblance, comparison, analogy--that we are propositions created by step 2 is highly mechanical. Miller introduces three
trying to appreciate. Such concepts have a structure, and S 1 makes that structure general rules which automatically create comparison statements from proposi-
explicit.' S 1 is the general form of all comparison statements, and encompasses tional input, for three classes of metaphors: nominal, verbal, and sentential.
nonliteral and metaphorical c'omparison statements: (Actually Miller uses the term 'predicative', but to the linguist this is misleading,
for the nominal part of a metaphor like 'Man is a wolf is also used predicatively.)
(6) S1. SIM[F(x), G(y)] The rules are as follows:

This is a conceptual analysis of similarity and comparison, as Miller repeatedly (9) a. M1. BE (x,y) -> (3F) (3G) {SIM [F(x), G(y)]}
emphasizes (1993:377; 381; 382; 385; 389). Miller shows that it allows for an b. M2. G(x) -> (3F) (3y) {SIM [F(x), G(y)]}
c. M3. G(y) -> (3F) (3x) {SIM [F(x), G(y)]}
analysis of metaphorical propositions as nonliteral analogies with conceptual
68 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 69

Although the structure of the comparison statements is complete, not all of the a matter of interpretation' (1993 :384). But Miller is right, and, what is more, there
concepts of the comparisons are known: the unknown ones are thematized, as it is no contradiction between these warnings and our undertaking. Even if our goal
were, by the two existential propositions preceding the comparison structures. For is one of analysis and identification, it will be good to remember that there are
instance, in the nominal metaphor 'Man is a wolf, also discussed by Miller objective limitations imposed on our pursuit and that interpretation needs to be
(1993:382), the underlying comparison suggests that there is a property of men kept on the leash.
such that it is similar to some other property of wolves. It is the task of step 4, It is fortunately possible to say a little more about the actual procedure of
nonliteral analogy identification, to attend to these unexpressed properties and fill nonliteral analogy identification. For this purpose it is useful to return to Reinhart
them in at the appropriate slots of the incomplete comparison statement generated (1976), whose main pointwas to introduce a distinction between focus and vehicle
by rule MI. However,. the derivation of the incomplete nonliteral comparison identification and interpretation as two distinct aspects of understanding metaphor.
statements themselves is much less interpretive, for it follows from the analysis These two aspects may now be put to use in the identifieation of the nonliteral
of a stretch of discourse as containing a metaphorical proposition of a particular analogy. I will begin with focus interpretation, but this does not mean that this
type, nominal, verbal, or sentential. reflects the actual order of proceeding.
Let us examine another of Reinhart's examples. She discusses another
4. Nonliteral analogy identification metaphor from T.S. Eliot's 'Love Song of J. Alfred Prufrock', reproduced under
The fourth step handles the reconstruction of the complete nonliteral (10):
comparison statement by inferring the implied concepts for the empty slots. Miller
calls this the reconstruction of the comparison, but I would like to call this (10) The yellow fog that rubs its back upon the window panes
nonliteral analogy identification. It would serve well to suggest what step 4 adds
to the previous steps of (1) metaphor focus identification, (2) metaphorical idea Reinhart offers the following analysis of the focus:
identification, and (3) nonliteral comparison identification: by filling in the empty
slots of the comparative structure produced by step 3, the incomplete nonliteral In this metaphor, the focus is rubs its back upon, since we can substitute
comparison'statement is turned into a full-blown nonliteral analogy in step 4.1£ another expression for it, e.g. touches, swirls against, or comes up against,
step 3 turns complete propositions into comparisons between two incomplete to yield a literal expression, such as The yellow fog that touches the window
propositions by means of rules Ml through M3, step 4 fills out each of these panes. (1976:391)
incomplete propositions into complete ones. Nonliteral analogy identification
(step 4) has to follow after nonliteral comparison identification (step 3). Let us examine how this analysis fares in our procedure and how it can be inserted
. The term nonliteral analogy identification also suggests that the aim of the into step 4. If we follow Reinhart in identifying the focus as she does, we
entire undertaking is still to obtain reliable analyses of metaphorical language. Our rediscover this in the following list of propositions produced by step 2:
frame of research is the kind of claim that a particular stretch of language is an
(11) The yellow fog that rubs its back upon the window panes
expression of a specific conceptual metaphor, and if such claims are to be upheld,
PI (RUB FOG BACK)
we need to be able to support them with reliable identification procedures at every
P2 (UPON pI PANES)
stage of the analysis. In other words, the step from incomplete nonliteral
P3 (MOD PANES WINDOW)
comparison statements to complete nonliteral analogies is supposed to be one of
identification, in which interpretation is to be kept under firm control: We are to
If we combine all of these propositions and take 'rub its back upon' as a complex
pi<;:k out the underlying analogy that may plausibly be conjectured to play a
verbal group used in a nonliteral fashion, we need Miller's rule M2 to create the
guiding role in interpreting the incomplete comparative structure produced in step
following incomplete comparison statement as step 3:
3. In this connection it may be disheartening to find Miller issuing the waming
that 'there can be no uniquely correct comparison statement' (1993:384).
(12) Rub back upon (fog, window panes) -> (:3F) (:3y, y') {SIM [F(fog, window panes),
Somewhat later, he writes, '[T]he search for suitable values to convert [an
rub back(y,y')]}
incomplete comparison statement to a nonliteral analogy, GS] is, strictly speaking,
70 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 71

To fill this in, as part of step 4, we require the operation of Reinhart's focus Reinhart's vehicle is hence identical to the second existential proposition of the
interpretation. It is nothing but the search for the predicate F which is needed to incomplete comparison statement quoted above, (::Jy,y'): we need to find a frame
cement the relation between the fog and the window panes in the first incomplete consisting of two literal entities which may be related to each other by means of
proposition of the comparison. The choice of 'touch' by Reinhart is one possible the function 'rub its back upon' so that they form one coherent state of affairs. The
option (although it is metaphorical, too). This would fill out the empty slot on the resulting proposition is supposed to be similar in a nonliteral fashion to the fog
literal side of the equation quoted as (12). Reinhart comments: touching the window panes. Reinhart suggests that 'cat' would be a good
candidate for the vehicle thus defined, and the poetic context of the line in
Focus-interpetation assigns a reading to the focus expression which is a question provides corroboratory evidence.
matter of selecting those properties associated with the focus expression There are again many aspects which would deserve further treatment.
which are relevant to the context. Thus among the properties of rubbing However, let us conclude this section by pointing out that there is an important
one's back, the properties of physical contact and of being in movement are
difference between the two aspects of step 4. Focus interpretation is usually more
consistent with the context of Eliot's metaphor, hence they can be selected.
This procedure provides a rough understanding of what the metaphor is richly constrained by the context of the metaphor, in that it involves the
about, or what the actual situation which is being depicted is (the fog construction of a literal proposition against the background of the topic and
swirling against the window panes), and how it ties in with the wider content of the previous discourse (see Reinhart's remarks above, and compare
context of the metaphor. (1976:391-2) Miller 1993:394). Vehicle identification does not have this rich contextual
constaint--it has to activate prototypical or default knowledge about the source
Focus interpretation produces a partial but basic understanding of the metaphor. domain. Miller writes: 'usually it is sufficient to take as y whatever the most
In Miller's terms, it deals with 'the cast of characters in the reader's concept of the generic argument G is conventionally predicated of' (1993:393). Indeed, part of
text' (1993:382). Note the comparable referential style in Reinhart's passage, the difficulty of vehicle identification is precisely that there may be more than one
when she uses 'what the metaphor is about, or what the actual situation which is source domain which can be involved in the interpretation of second completed
being depicted is.' proposition (cf. Vosniadou and Ortony 1989). This is clearly notthe casefor focus
However, the focus interpretation of back rubbing can only take place if it interpretation, in which the conceptual domain is identical to or part of the
is also at least partly interpreted as rubbing one's back (see Reinhart's selection conceptual domain of the stretch of discourse in which the metaphor is located.
of 'the properties of physical contact and of being in movement' in the above Another crucial aspect is the relation between the two sides of the analogy.
quotation). In other words, the second incomplete proposition of (12) also has to Part of the meaning of y may have to be filtered out by its lack of relevance to the
be part of the equation. This is the beginning of vehicle'identification. And to literal topic of the metaphor and the discourse. However, many possible
construct the full set of conceptual correspondences between the two conceptual assumptions about the source domain can usually be maintained with varying
domains involved in the metaphor, the second incomplete proposition of (12) degrees of strength in the context of the target domain: a range of possible values
leads us into a relatively independent consideration of the source domain. The for the empty slots is usually possible, that is, compatible with the textual world,
second incomplete proposition needs to be fleshed out itself, too, primarily by which makes it rather difficult to hit upon the best vehicle for the eventual
filling in the empty slots of the arguments relating to rubbing one's back. This is analogy. That focus interpretation and vehicle identification thus exert a mutual
not just the beginning but nothing less than Reinhart's vehicle identification. influence on each other when one attempts to align the two parts of the analogy
Reinhart's definition of the notion of vehicle squarely falls within our is self-evident and the cornerstone of the interaction theory (cf. Miller 1993:394
operation of filling in the missing terms of the nonliteral side of the analogy: and Reinhart 1976:389). This does make step 4 highly dependent on the specifics
of every particular metaphor.
The vehicle is the frame Fj [ ] in which the occurrence of Ej results in a
literal expression, Fj[EJ, where Fj[I;] is not similar in meaning to F;[EJ 5. Nonliteral mapping identification
(1976:391) The last step in the procedure is to identify the complete nonliteral mapping.
This is done by filling out the conceptual structure of the two sides of the
non literal analogy, the source and the target domain. Other concepts and the
72 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 73

relations between them are to be listed for each domain, and their interdomain not happen anywhere between steps 1 and 4. Step 4 is crucial, because that is
relations are to be projected. Some of these additions may be motivated where interpretation comes in, albeit as tightly controlled by the aims discussed
semantically whereas others are based on our know ledge of the world in particular above as possible, and step 5 acts as a verification of step 4 in spelling out its
domains (cf. Miller 1993:382 and Reinhart 1976:388). The result of this operation consequences in a more complex and explicit manner.
is a conceptual network from which the analyst may derive sets of correspon- It is clear that the last two steps of the procedure form the weakest parts of
dences such as those illustrated at the beginning of this chapter (Lakoff 1993 :207). the chain, with step 5 being the weakest of all. Future research will have to
The transition from analogies to mappings is not given much principled concentrate on strengthening these stages of the analysis. Otherwise, the gap
attention by Reinhart and Miller: they seem to relegate it to the domain of between metaphorical language and conceptual metaphor may never be bridged
interpretation about which they do not offer systematic observations. It is a theme in a reliable fashion. I will now move on to a brief illustration of the complete
which, to my knowledge, hlfS not been discussed by Lakoff either,except in a sequence of steps.
negative fashion, as in (1993:210): 'Mappings should not be thought of as
processes, or as algorithms that mechanically take source domain inputs .and 6. Conclusion
produce target domain outputs.' One of the more prolific writers in this area, I have suggested that the identification of conceptual metaphor in discourse
however, is Dedre Gentner (1982; 1983; 1989; Gentner & Jeziorski 1993), who requires five steps:
has done experimental research on behavior as well as AI modeling of knowledge
structures. She has listed six principles of analogical reasoning which may be used (1) metaphor focus identification
to constrain ·the production of mappings from analogies. However, Gentner (2) metaphorical idea identification
(personal communication) has also admitted that the technical analysis of such (3) nonliteral comparison identification
mappings is still basically an art. I cannot escape the impression that another (4) nonliteral analogy identification
(5) nonliteral mapping identification
alternative in this area, Turner and Fauconnier's (1995) conceptual blending
theory, fares no better.
These steps are all called identification, because they give an answer to the
Thus the question arises how step 5 of the procedure may be better
question 'What is ... ?' The first three questions are easiest to answer: what is the
constrained. Arid indeed, asking this question invite.s reviewing the other steps and
metaphor focus, what is the metaphorical proposition, and what is the metaphori-
inspecting the relation between step 5 and step 4 in particular. For identifying the
cal comparison? Question 4 is more difficult to handle, because it involves filling
analogy in step 4 seems to require at least a partial identification of the nonliteral
in empty slots in an analogy on the basis of focus interpretation and vehicle
mapping; which is step 5, in order for the proposed analogy to be plausible. To
identification in mutual interaction. However, context and default language use,
take this further, it might be suggested that identifying the analogy is a kind of
respectively, act as guides to provide an answer to question 4. The answer to this
summary or abstraction of the nonliteral mapping identified in step 5, and this
question has to be seen as a searchlight for constructing a nonliteral mapping,
could lead to the query why step 4 and 5 are not reversed. In other words, one
which then has to be checked against the discourse regarding its appropriateness.
might also entertain an order in which the incomplete comparison statements of
This step is the least reliable step in the procedure, while it is essential in order to
step 3 are first fleshed out into conceptual domains with conceptual correspon~
arrive at metaphors as sets of conceptual correspondences.
dences and are only afterwards condensed, as it were, into neat explicit analogies. Let us return to the watchdog metaphor discussed above. Identifying the
However, to raise the question is to answer it. For one needs a kind of
focus of the metaphor involved checking that the word watchdog was not used in
searchlight in the construction of the conceptual domains and their relations, its literal sense, referring to an animal guarding some property or people. This was
which involves a kind of propositional interpretation of the implied comparison. not difficult on account of the accompanying information which explicitly
Step 4 provides such a provisional interpretation in the form of an analogy, and signalled that the watchdog was 'The National River Authority' or 'the local
this analogy acts as a target for the·construction of the more complex mapping (or authority financial watchdog' . Such explicit lexical signals make focus identifica-
mappings, if you follow conceptual blending theory) .. There is hence a special tion an easy task, but they are often unnecessary given the overt contrast between
relation between steps 4 and 5, in which step 4 provides a tentative analogy which
the domains of the metaphor focus and the topic of the text.
is rejected or retained depending on the success of the ensuing step 5. This does
74 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 75

The identification of the metaphorical proposition is not problematic either. appear). Differences between nominal, verbal, and sentential metaphor may cause
It involves constructing the proposition (BE ORGANISATION WATCHDOG) or some different situations for step 4. And so on. These and other issues are on the agenda
variant thereof. As was shown above, there is some additional conceptual for future research. One specific item is the testing of the reliability of the various
variation between the individual cases of the metaphor. steps: specialist informants may be given tasks to carry out steps I through 5,
The identification of the nonliteral comparison is done by feeding the either separately or in a row, with diverging sets of materials in order to reveal
proposition into rule MI. This yields the paraphrase "Some property of a more specific difficulties of applying the procedure to real discourse data.
committee is like some property of a watchdog." The general conclusion, however, is that we have offered a logical
Identifying the nonliteral analogy next concerns the finding of the properties reconstruction of the analytical process, and that it holds a promise for practical
left open in the nonliteral comparison. I will begin with vehicle interpretation for use as a descriptive tool in semantics. Again, the procedure does not pretend to
the sake of showing that there is no predetermined order between the two aspects. model the comprehension process. Moreover, as was suggested above, some steps
Taking the canonical values of what watchdogs usually do, we can suggest that the in the analysis make greater interpretative jumps than others, and this is where
relevant property is 'guarding property or people' . The property of the committees cognitive linguists should be aware of alternative explanations of their data and
in question is similar, in that they have to guard the public interest in the economic their beliefs in conceptual metaphors. It is hoped that the procedure can pinpoint
domain. One case has the following additional information: 'To protect the some of the more risky moments of analysis and that it can help in recording
interests of non-Treasury use of economic data' . experiences in negotiating these moments.
Fleshing out the analogy into a complete mapping involves listing attributes
of both committees and watchdogs and attempting to match them. The following Author's note
provisional list may be entertained: I wish to thank Ray Gibbs, Rachel Giora, Lachlan Mackenzie, and Wilbert Spooren for
their acute observations and comments on an earlier version of this paper.
THE COMMITIBE-AS-WATCHDOG MAPPING
The committee corresponds to the watchdog. References
The organizational domain corresponds to the yard. Beardsley, M. 1958. Aesthetics: Problems in the Philosophy of Criticism. New
The interest or activity at risk corresponds to the property. York: Harcourt Brace Jovonovich.
Malpractice corresponds to trespassing. Black, M. 1962. Models and Metaphors. Ithaca, N.Y.: Cornell University Press.
Monitoring corresponds to watching. Bovair, S., &D.E. Kieras. 1985. "A Guide to Propositional Analysis for Research
WamiQg the public corresponds to barking. on Technical Prose". Britton & Black 1985. 315-362.
Britton, B.K., & J.B. Black, ed. 1985. Understanding Expository Text. Hillsdale,
Other aspects can be included, and the list may be adjusted according to context. N.J.: Lawrence Erlbaum Associates.
As I have acknowledged at the beginning of this paper, there may be another Cohen, L. J. 1993. "The Semantics of Metaphor" . Metaphor and Thought, second
step, which is to compare the analysis of one metaphor with those of others. This edition, ed:by A. Ortony, 58-70. Cambridge: Cambridge University Press.
would be the last step in determining whether a metaphor is part of a systemati- --------, & A. Margalith. 1972. "The Role of Inductive Reasoning in the Interpreta-
cally organized set of metaphorical concepts (conventional conceptual metaphors) tion of Metaphor". Semantics of Natural Language, ed. by D. Davidson, &
or not. In our case, the analysis would not be dramatically different for every case G. Harman, 722-744. Dordrecht: Reidel.
of the ten instances of the watchdog metaphor, and we might come to the Dijk, T.A. van. 1975. "Formal Semantics of metaphorical Discourse". Poetics 4.
conclusion that the metaphor is relatively conventional, depending on other 173-198.
frequencies. Gamham, A. 1996. "Discourse Comprehension Models". Computational
There are many other issues which have also had to be left aside. For Psycholinguistics: AI and Connectionist Models of Human Language
instance, etymological and sociolinguistically restricted metaphor may pose Processing ed. by T. Dijkstra, & K. de Smedt, 221-244. London: Taylor &
special. problems in step 1. Implicit, reduced, complex, multiple, extended, and Francis.
mixed metaphors provide special challenges to steps 2 and 3 (Steen 1999 and to
76 GERARD STEEN FROM LINGUISTIC TO CONCEPTUAL METAPHOR IN FIVE STEPS 77
Gentner, D. 1982. "Are Scientific Analogies Metaphors?" Metaphor: Problems Vosniadou, S., & A. Ortony. 1989. "Similarity and Analogical Reasoning: A
and Perspectives, ed. by D. Miall, 106-132. Brighton: Harvester. Synthesis". Similarity and Analogical Reasoning, ed. by S. Vosniadou, &
-------- 1983. "Structure-Mapping: A Theoretical Framework for Analogy". A. Ortony, 1-17. Cambridge: Cambridge University Press.
Cognitive Science 7. 155-170.
-------- 1989. "The Mechanisms of Analogical Learning". Similarity anti
Analogical Reasoning ed. by S. Vosniadou, & A. Ortony, 199-241.
Cambridge: Cambridge University Press.
--------, & M. Jeziorski. 1993. "The Shift from Metaphor to Analogy in Western
Science". Metaphor and Thought, second edition, ed. by A. Ortony, 447-
480. Cambridge: Cambridge University Press.
Johnson, M. 1987. The Body in the Mind. Chicago: University of Chicago Press.
Lakoff, G. 1987. Women, Fire, and Dangerous Things: What Categories Reveal
about The Mind. Chicago: University of Chicago Press.
-------- 1993. "The Contemporary Theory of Metaphor. Metaphor and Thought,
second edition, ed. by A. Ortony, 202-251. Cambridge: Cambridge
University Press.
--------, & M. Johnson. 1980. Metaphors We Live By. Chicago: University of
Chicago Press.
--------, & M. Turner. 1989. More Than Cool Reason: A Field Guide to Poetic
Metaphor. Chicago: University of Chicago Press.
Miller, G.A. 1993. "Images and Models, Similes and Metaphors". Metaphor and
Thought, second edition, ed. by A. Ortony, 357-400. Cambridge: Cambridge
University Press.
Perfetti, C.A., & M.A. Britt. 1995. "Where Do Propositions Come From?"
Discourse Comprehension: Essays in Honor of Walter Kintsch, ed. by C.A.
Weaver, ill, S. Mannes, & C.A. Fletcher, 11-34. Hillsdale, N.J.: Lawrence
Erlbaum Associates.
Reinhart, T. 1976. "On Understanding Poetic Metaphor". Poetics 5.383-402.
Richards, lA. 1936. The Philosophy of Rhetoric. London: Oxford University
Press.
Steen, G.J. 1999. "Metaphor and Discourse: Towards a Linguistic Checklist for
Metaphor Analysis". Researching and Applying Metaphor, ed. by L.
Cameron, & G. Low, 81-104. Cambridge: Cambridge University Press.
-------- To appear. "Analyzing Metaphor in Literature: With Examples from
William Wordsworth's 'I wandered Lonely as a Cloud"'. Poetics Today 20.
Turner, M. 1987. Death Is the Mother of Beauty: Mind, Metaphor, Criticism.
Chicago: University of Chicago Press.
--------, & G. Fauconnier. 1995. "Conceptual Integration and Formal Expression".
Metaphor and Symbolic Activity 10. 183-204.

You might also like