The Oxford Handbook of Cognitive Linguistics Part 23 ppsx

2. Basics of Conceptual
Metaphor Theory

The most fundamental notion of conceptual metaphor theory (CMT) is the map-
ping.
2
This term borrowed from mathematics refers to systematic metaphorical
correspondences between closely related ideas. For example, the common concep-
tualization of a nation (or other political body) as a ship includes correspondences
between the ship and the state conceived as wholes, but also between the course of
the ship and the historical progression of the state; the seas traversed by the ship
and the political and other circumstances with which the state is faced; and so
forth. Rather than existing as isolated specimens, the metaphorical usages within a
passage like the following depend on this conventional pattern of conceptual
associations.
The blueprints drafted last week will ensure that the ship of the Commonwealth
truly remains one for the ages The House Budget will allow the state to
withstand even the stormiest weather Continued commitment to our most
needy and the Commonwealth’s most essential obligations is critical ballast for
every successful ship of state. (from Massachusetts State Representative Paul
Casey’s Web page: empha-
sis mine)
In the CMT system, the course of the ship is said to ‘‘map’’ or ‘‘be mapped’’ (or
‘‘projected’’) onto the historical progression of the state, and other elements of the
conceptual domain of ships and navigation (the ‘source’ domain) are likewise
‘‘mapped’’ onto elements of the conceptual domain of nations and politics. The
source domain of a metaphor (here, ships and navigation) supplies the language
and imagery which are used to refer to the domain which is actually at issue in the
discourse (the ‘‘target’’ domain—in this case, politics and states). As it is used
in popular discourse, the metaphor includes at least the following conventional
cross-domain correspondences (see Grady, Oakley, and Coulson 1999):

state ship
State’s policies/action Ship’s course
Determining policies/actions Steering the ship
Success/improvement of the state Forward motion of the ship
Failures/problems Sailing mishaps (e.g., foundering)
Circumstances affecting the state
(e.g., on the political or economic levels)
Sea conditions
Other conventional metaphorical patterns involving multiple correspondences
between source and target domains include death is a reaper (where people are
plants to be harvested, etc.) and life is a journey (where difficulties are obstacles,
objectives are destinations, etc.; see Lakoff and Turner 1989).
3
Note that some
patterns are quite a bit more specific than others; for example, life is a journey is
190 joseph e. grady
a very broad pattern, of which love is a journey (We have reached a crossroads in
our relationship) might be considered a special case. That is, metaphors exist in
hierarchies of specificity, in which a more specific pattern can be said to inherit a
more general one. In each metaphorical pattern, the mapping between source and
target is constrained by what cognitive linguists have called the invariance principle
(Lakoff and Turner 1989; Brugman 1990; Lakoff 1990; Turner 1991; etc.): the require-
ment that the mapping not violate the basic topological structure of the target do-
main (see Oakley’s discussion of ‘‘Image Schemas’’; this volume, chapter 9).
The systematic projection of elements from one conceptual domain onto el-
ements of another involves not merely the objects and properties characteristic of
the domain (e.g., buildings, sturdiness vs. flimsiness, etc.) but also the relations,
events, and scenarios that characterize the domain. In short, CMT is concerned with
the mapping of inferences from source to target. If a person ‘‘blows off steam,’’
then the person should feel a reduced intensity of anger afterwards. If one ‘‘gets

past obstacles’’ in one’s work, then he or she should achieve greater success in his
or her objectives. On one level, inference mapping is another illustration of the
richness of the conceptual structures upon which metaphorical usages are based.
On another, it is a strong demonstration that metaphor is more than an innovative
use of language or of the figurative application of a single term to a new referent.
Besides systematicity, the asymmetrical directionality of conceptual metaphors
is one of the features most strongly emphasized by Lakoff and Johnson and cog-
nitive linguists since. While a term like weather may be used metaphorically to refer
to a set of economic and political circumstances, the reverse metaphor is not pos-
sible, linguistically or conceptually (e.g., the nonsensical idea of referring to an ac-
tual storm as a recession). Likewise, it is meaningful to refer to a person as warm
but meaningless to refer to a cup of tea as affectionate, meaningful to refer to the
foundations of a theory but meaningless to refer to the postulates of a building. Note
that this usage is not only unconventional but uninterpretable. We can guess what
white-hot anger might be like even if we have never heard the phrase, thanks to the
underlying conceptual pattern that projects heat onto intense emotions; there is no
corresponding pattern which allows us to understand parts of a building in terms
of postulates or other elements of logic.
This directionality is more than an interesting and characteristic feature of
metaphorical conceptualizations; it is evidence against a traditional and still com-
mon view of metaphor, in which a metaphorical usage is most fundamentally a
reflection of ‘‘similarity’’ between the source and target ideas. If the ultimate basis
for the Theories-as-Buildings pattern (see discussions in Lakoff and Johnson 1980;
Clausner and Croft 1997; and Grady 1997b) is an underlying similarity (called a
ground, in some philosophical studies), metaphorical substitutions might be just as
valid in either direction; clearly they are not. More importantly, there are cases
where it is hard to identify what the similarity might be between the source and
target concepts in a metaphor, even in a simple and familiar pattern such as the
understanding of happiness as brightness. We refer to a sunny disposition and a
bright future, but what could the similarity be between a mood and a degree of

metaphor 191
luminance? Or between physical coldness and lack of emotion? Philosopher John
Searle acknowledged this puzzle when he wrote:
I think the only answer to the question, ‘‘what is the relation between cold
things and unemotional people that would justify the use of ‘cold’ as a meta-
phor for lack of emotion?’’ is simply that as a matter of perceptions, sensibili-
ties, and linguistic practices, people find the notion of coldness associated in their
minds with lack of emotion. The notion of being cold just is associated with being
unemotional. (1981: 267)
Coldness and lack of emotion are not ‘‘similar’’ in any straightforward way, yet cog-
nitive linguists have been able to point to a more particular reason than Searle
recognizes in this passage: the conceptual domains of temperature and emotion are
associated in our experience, for instance, because intimate interactions can entail
physical proximity which leads to body heat being shared.
The emphasis on ‘‘experiential motivation’’ is another of the central principles
of CMT, and one which most sharply distinguishes the approach from alternative
theories. Lakoff and Johnson (1980), for example, discuss motivations for the met-
aphorical pattern they call more is up. In their account, elevation and quantity are
conceptual domains closely related in experience, since whenever we see a pile of
objects or liquid in a contained space, we are aware of the connection between the
height which the pile (or whatever) reaches and the number of objects or amount of
the liquid. In this way, the mapping between quantity and height is well motivated,
rather than arbitrary, but does not depend on similarity per se. This experiential
analysis is typical of Cognitive Linguistics’ concern with metaphors not only as in-
teresting linguistic phenomena requiring explanation, but as important elements of
conceptual structure and reflections of ways in which humans experience the world.
3. Primary metaphors
and ‘‘Neural CMT’’

There is a set of pervasive conceptual metaphors which seem to reveal with special

directness the deep relationships between word usage, conceptual structure, and
the way we experience the world. Primary metaphors (Grady, Taub, and Morgan
1996; Grady 1997a; Lakoff and Johnson 1999; etc.) are simple patterns, like Lakoff
and Johnson’s more is up, which map fundamental perceptual concepts onto
equally fundamental but not directly perceptual ones. Source concepts for primary
metaphors include up, down, heavy, bright, forward, backward, sweet,
varioussimple‘‘force-dynamic’’concepts(inthesenseofTalmy1988),andsoon.Cor-
responding target concepts are such basic building blocks of mental experience
192 joseph e. grady
as dominant, sad, difficult, happy, success, the past, appealing, and
compulsion. (Many of the metaphors for emotion discussed by Ko
¨
vecses 1990 are
primary metaphors.) These metaphors appear to arise directly from experience in
ways that more complex metaphors such as theories are buildings do not.
There are experiential correlations between quantity and height, as we have seen,
and likewise between other concepts paired in primary metaphors: happiness and
brightness (we feel safer and more content in sunshine than in the dark); difficulty
and heaviness (we experience strain when we try to support or manipulate heavy
objects); and so forth. There are no such experiential correlations between theo-
ries and buildings, ships and nations, or between various other complex con-
cepts which are linked in conventional metaphorical pairings.
The unidirectionality of primary metaphorical patterns is consistent and ab-
solute. In each case, the perceptual concept is the source and is mapped onto the
nonperceptual target concept. Success is easily spoken of and conceived as motion
forward (e.g., We’ve made great strides forward this year), but simple forward
motion is not thought of as success (consider a car rolling slowly downhill because
its brake has not been set). An important matter is heavy or weighty, but we cannot
communicate that one laptop computer is heavier than another by saying it is more
important.

4
This strong unidirectionality is especially significant given that there are
many metaphors which are not asymmetrical in exactly the ways emphasized within
most CMT accounts. For instance, Lakoff and Turner (1989: 89–96) discuss a cat-
egory of ‘‘image metaphors,’’ which depict one thing in terms of another based on
shared perceptual features (visual or otherwise): ‘‘My wife whose waist is an
hourglass,’’ ‘‘My horse with a mane made of short rainbows,’’ and so on. In each of
these cases, the target-source relationship can be reversed and still yield a perfectly
understandable metaphor, if not as poetic an image: the ‘‘waist’’ of an hourglass, a
rainbow as a horse’s mane, and so on.
5
There are also metaphors which appear to be
based on shared qualities which are not perceptual: when we refer to a person as
some type of animal based on a personality trait, for example (e.g., as a ‘‘pig,’’
‘‘snake,’’ or the more classical, not to mention complimentary, ‘‘lion’’), we are ap-
parently invoking a commonality which we believe unites the person and the
animal (or some stereotype of the animal).
6
Like image metaphors, these are based
on conceptual relationships which can be reversed and still be meaningful: we refer
to lions as the ‘‘king of beasts’’ and might even equate a particular lion with a par-
ticular human exemplar of stout-heartedness.
Strict unidirectionality, then, appears to be a special feature of correlation-
based primary metaphors. Complex metaphors which are also thoroughly asym-
metrical, like theories are buildings, appear to be analyzable as elaborations of
conceptualizations which are, at bottom, primary metaphors—for example, one in
which logical organization is understood as physical part-whole structure and
another in which continued functionality or existence is understood as persisting in
a standing position (Grady 1997b). Each of these metaphorical patterns is entirely
unidirectional and can plausibly be accounted for in terms of recurring correlations

metaphor 193
in experience. (An account on these lines also has the important advantage of
explaining why many of the most salient aspects of buildings from the point of view
of the human interacting with them—floors, walls, occupants, rent, and so on—are
not parts of the conventional understanding of theories; the metaphors are about
understanding complex abstract entities as erect physical structures, rather than as
buildings per se.)
Given that humans everywhere share the basic patterns of perception and
experience that are reflected in primary metaphors, these patterns ought to show up
in languages around the world.
7
In fact, it does appear that primary metaphors are
widespread across languages that are not related genetically, areally, or culturally.
An example of a pattern with broad (if not universal) cross-linguistic distribution is
the semantic extension from ‘large’ to ‘important’, observed in senses of Hawaiian
nui , Malay besar, Russian krupnij, Turkish bu
¨
yu
¨
k, and Zulu -khulu, for example
(cf. uses of English big, such as Today is a big day for the company). In each case, a
basic word referring to size is conventionally used to refer to importance in utterly
nonphysical situations (presumably based on the frequent correlation in our ex-
perience between the size of an object and its salience or importance to us). Addi-
tional patterns found broadly across languages include ‘(spatially) close’ for ‘inti-
mate’, ‘(spatially) close’ for ‘similar’, and ‘warm/hot’ for ‘agitated’ (Grady 1999b).
Primary metaphors, then, are natural or even inevitable consequences of recurring
associations in daily life.
Analysis in terms of primary metaphors was one of several concurrent devel-
opments that led to the ‘‘Neural Theory of Language,’’ and more specifically, a

‘‘neural’’ version of CMT (Lakoff and Johnson 1999). Within this framework (which
represents cognitive structures in computational ‘‘neural nets’’; see the discussion
of computational models of metaphor below), the mappings that constitute pri-
mary metaphors are treated as neural circuits linking representations of source and
target concepts—circuits which are automatically established when a perceptual
and a nonperceptual concept are repeatedly co-activated. This automaticity effect
within Neural CMT is an implementation of hypotheses about the origins of met-
aphor in children’s thought and language. The concepts linked in primary meta-
phors are so closely associated with each other in experience that conflated repre-
sentations may arise as integrated wholes in conceptual structure even before they
are understood as associations between distinct ideas. C. Johnson (1999) has shown
that children may have no basis for distinguishing between literal and metaphorical
senses of a term like see based on the ways in which they hear the term used. If
parents and others regularly use see in contexts where it can mean either ‘perceive
visually’ or ‘learn; find out’—as in Let’s see what’s in the box versus Let’s see what this
bell sounds like—children may hypothesize a sense of the term which conflates
literal and metaphorical meanings from the adult point of view, and later need to
perform a process of deconflation before they understand that there are two distinct
senses of the word, linked by a conventional pattern of metaphor. Such a devel-
opmental trajectory may be natural for many or most terms with conventional
senses that are licensed by primary metaphors.
194 joseph e. grady
4. The Cognitive Reality
of Conceptual Metaphors

Are the mappings identified within CMT cognitively ‘‘real,’’ rather than mere clever
inventions of the analyst trying to account for data (like competing phonological
analyses which share nothing except that they can generate the same set of forms)?
This is a more crucial question within Cognitive Linguistics than it is within other
areas of linguistics, since practitioners of Cognitive Linguistics take seriously the

idea that they are describing relationships between language and other cognitive
functions. There are several types of evidence that conceptual metaphor mappings
are psychologically real. First, there is the fact of systematicity itself; if sets of terms
and ideas from particular conceptual domains are systematically and predictably
associated with such sets from other domains, it is easier to conclude that the do-
mains are connected on some level of understanding than that mere accidents of
usage have led to the data patterns or that the patterns have arisen through analogy,
one lexical item at a time. The fact that we can often immediately grasp novel
metaphorical usages like white-hot anger also suggests that the underlying concep-
tual patterns are real.
Another compelling confirmation of the reality of metaphors on a conceptual
(rather than merely a lexical) level is the way in which gestures often appear to be
motivated by metaphorical understandings for which we have evidence in spoken
language. McNeill (1992) has used the term ‘‘metaphorics’’ to refer to gestures which
are metaphorically motivated. His recorded examples include a gesture used by a
mathematician during a conversation with a colleague about the technical concept
of limits: while committing a speech error by mentioning ‘‘inverse limits’’ rather
than the direct limits he has in mind, the speaker nonetheless makes the hand
gesture associated with direct limits (an abrupt motion and stopping of the hand,
at the ‘‘end point’’), showing that his gesture is in fact motivated by his concep-
tualization of the topic—that is, an understanding in which a quantitative limit is
treated as a physical obstacle or stopping point—and specifically not by the word
he is uttering at the time. It would be difficult to argue that examples like this are
motivated by mere analogy with language; instead, they appear to reflect the same
underlying patterns of conceptualization which motivate metaphorical uses of
words.
Evidence from experimental psychology also helps confirm the cognitive reality
of conceptual metaphors. Gentner (2001) has reported on a series of experiments
designed to test whether people actually invoke metaphorical conceptualizations
when they think and speak about time. Her results indicate that when people switch

from one metaphorical system for understanding time to another, there is a cost
in reaction time. Subjects were first asked a question framed in terms of either the
so-called ‘‘ego-moving’’ model of time or the ‘‘time-moving’’ model. In the ego-
moving model, time is conceptualized as the path or landscape through which we
metaphor 195
move, with the future lying in front of us (e.g., Is Boston ahead or behind us time-
wise?). The time-moving model frames time independently of the observer (e.g.,
Is it later or earlier in Boston than it is here?), as a series of objects following each
other through space; in this model, a later time follows and is therefore ‘‘behind’’ an
earlier time. When the first and second questions were framed in terms of con-
flicting models, response times were slower than when they were framed in terms of
the same model. Furthermore, in the ‘‘conflicting’’ condition, subjects often refor-
mulated the second question for themselves, apparently in order to make it easier
to understand and reply to; this reformulation did have the effect of significantly
speeding up response times. Gentner draws the conclusion that the metaphorical
systems are playing a real role in subjects’ conceptualization of time and their un-
derstanding of temporal language. In a subsequent series of ingenious experi-
ments involving rigorous testing of alternative hypotheses, Boroditsky (2000) dem-
onstrates even more clearly that spatial schemas play a role in structuring temporal
thought.
Another kind of evidence comes from Gibbs (1994: 163–64), who has reported
that there are consistencies in the mental images described by experimental sub-
jects interpreting metaphorical idioms which go beyond the information supplied
in the idioms themselves. Conversely, metaphorical idioms suggest the same set of
inferences to people, even when those inferences go far beyond what is expressed in
the words themselves; subjects agree that when you blow your stack, ‘‘the expression
of anger is unintentional and is done in an abrupt, violent manner.’’
Despite the range of evidence in favor of the view of metaphors as entrenched
conceptual patterns, there are challenges to this position from a number of di-
rections. One common view which runs contrary to CMT perspectives is that met-

aphorical thought and language are essentially unconstrained. Philosopher Donald
Davidson (1981), for instance, has suggested that any two things can be understood,
when juxtaposed, as bearing a metaphorical connection.
8
If one person states that
‘‘Life is a kiwi fruit,’’ another will be clever enough to point out the shared features
which make these two entities comparable and which provide the ground for the
metaphorical mapping of one onto the other. In some sense, the range of possible
metaphorical correspondences would appear to be limitless, or to be limited only
by our imagination and our ability to interpret expressions based on pragmatic
context. In fact, Davidson goes so far as to claim that metaphors have no meaning
and that they merely invite us to infer whatever appropriate message we can. Ob-
viously, such a view has no room for conventional metaphors—lasting structures
which may narrow in advance the possible interpretations of a given expression
and which in practice also limit the metaphors we produce. Other philosophers,
such as Black (1955), offer a variety of accounts of metaphor which often have lit-
tle in common except for the assumption that metaphors are essentially uncon-
strained. Psychologists, too, have typically assumed that there are no metaphorical
relationships with a special status, instead looking for parameters which make in-
dividual metaphorical sentences more ‘‘apt’’ or more comprehensible, for instance
(see Katz 1989 and MacCormac 1985, among many others).
196 joseph e. grady
These philosophical and psychological accounts generally neglect the fact that
certain conceptual pairings tend to recur and to motivate a great percentage of the
actual metaphorical language we encounter. While there may be no such thing as
an ‘‘impossible metaphor,’’ metaphor scholars have had no difficulty in identifying
sets of particularly common patterns or in offering compelling accounts of the mo-
tivations for these frequently encountered mappings. For example, understanding
is regularly associated with vision and with grasping in English and other languages
of the world, but there are no such widespread mappings between understanding

and fighting, for instance. If metaphor is not constrained in an absolute sense by
underlying cognitive mechanisms, then at least there are patterns to be explained in
the metaphors which tend to arise as opposed to those which do not. Proponents
of CMT would argue that these more-likely metaphors are, crucially, better mo-
tivated. In short, traditions which focus on metaphor as an expression of similarity
have downplayed the conventionalized, structured aspects of metaphor, while CMT
scholars have focused particular attention on patterns which become entrenched in
language and conceptualization, often as a result of recurring associations in ex-
perience (see Grady 1999a).
Another type of experimental finding which is sometimes interpreted as cast-
ing doubt on the role of stored conceptual metaphors relates to processing time of
metaphorical versus literal language. One intuitively appealing view of how people
might process metaphorical language involves several stages: when we encounter
a metaphorical statement, we first try to interpret the statement as a literal one,
then seek alternate interpretations once we realize that the statement either does
not make sense or clearly does not relate to the current topic of discourse (see
Searle’s 1981 discussion of this ‘‘pragmatic’’ model of processing). To arrive at these
alternate interpretations, we might use stored knowledge of metaphorical patterns
of usage, among other tools. Some reaction-time data, however, shows that met-
aphorical utterances can be understood as quickly as literal ones, or even quicker
(e.g., Pollio et al. 1984). Such results have led researchers to question whether stored
metaphors are accessed during speech processing. What such data suggest to other
researchers, however, is that under the right contextual conditions we are ‘‘primed’’
to identify a metaphorical meaning for a sentence. We may even be slower to
recognize literal meanings in such cases.
9
Another apparent challenge to the CMT view of metaphor is what Lakoff and
Johnson (1980: 106–10) characterize as the ‘‘abstraction’’ position, namely, that
usages which might be identified as metaphorical are actually literal; an abstrac-
tionist view of a word’s sense holds that it is much more general and inclusive than

metaphor theorists allow. Jackendoff and Aaron (1991) provide a good illustration
of such a position in their review of Lakoff and Turner’s (1989) More Than Cool
Reason. Jackendoff and Aaron propose a test for metaphors based on the premise
that they necessarily involve semantic incongruity on their literal readings. State-
ments which fit naturally into the following formula involve incongruity and may
therefore be considered true cases of metaphor: ‘‘Of course A isn’t B, but if it
were, you might say that _____’’ (Jackendoff and Aaron 1991: 326). The following
metaphor 197
sentence, for example, shows that the idea of a romantic relationship which has
reached a dead end is truly metaphorical: ‘‘Of course, relationships are not jour-
neys, but if they were, you might say ours is at a dead end.’’ On the other hand,
Jackendoff and Aaron find the following sentence odd: ‘‘Of course, states aren’t
locations, but if they were, you might say I’ve gotten through my depression.’’ The
article argues that this sentence is problematic because there is no need for a hedge
in a statement like ‘‘I’ve gotten through my depression.’’ The statement is literally
true, since gotten through may literally refer to events other than spatial motion.
For Jackendoff and Aaron, this and many other expressions that CMT scholars
would identify as metaphors are instead reflections of ‘‘thematic relations’’—
abstract categories whose language refers primarily to concrete and spatial expe-
rience but whose content is much more general and abstract. ‘‘Being Circum-
stantially in a state is the thematic parallel of being spatially in a location’’ (328).
Jackendoff and Aaron argue that such usages should properly be seen as ordinary
and literal. Langacker (1987) raises similar questions about whether particular us-
ages of go are really metaphorical (e.g., This milk is about to go sour). He proposes
a generalized notion of extensionality that is not specifically tied to our concep-
tion of physical space. It is a property of many domains, both basic and abstract,
though the spatial domain stands out among them for its prototypicality and
cognitive salience. By making this distinction between extensionality in general
and physical space in particular, we can characterize ‘‘motion’’ in abstract terms
applicable to any extensional domain, without prejudging the extent to which

spatial metaphor is constitutive of these domains. (169–70)
Neither of these compelling proposals about the relationship between spatial lo-
cation and other conceptual domains, however, refutes the idea that there is a con-
ventional and metaphorical association between the domains of space and states
(for instance). Both discussions refer to usages of semantically weak terms like go
and get; a sentence like I’ve managed to crawl up out of my depression passes
Jackendoff and Aaron’s test and clearly evokes a vivid metaphorical image.
5. Conceptual Integration/
Blending

In 1994, Fauconnier and Turner introduced a new analytic framework which treats
metaphors as products of a more general process of human cognition. This op-
eration, which Fauconnier and Turner call ‘‘conceptual integration’’ (or ‘‘blend-
ing’’) involves the combination, often but not always figurative, of selected con-
ceptual material from two or more distinct sources. Like metaphor in CMT terms,
blending is understood as a pervasive phenomenon in human thought, one which
198 joseph e. grady
shows its effects regularly in everyday language. A fuller treatment of the blending
framework is provided by Turner (this volume, chapter 15), but it is useful here to
briefly consider how the blending account of metaphor adds to, and is different
from, the CMT approach. In order to do so, we need to have in mind a few basics
of the theory.
Fauconnier and Turner’s model involves four ‘‘spaces,’’ rather than the two
conceptual domains (source and target) which participate in a CMT mapping. Each
of these spaces is understood as a mental space in the sense of Fauconnier (1985,
1997; this volume, chapter 14); that is, a coherent bundle of information activated
in the mind at a particular time, representing an understanding of a scenario, real
or imagined. For example, when an artist creates a variant of Leonardo da Vinci’s
Mona Lisa (or la Gioconda) in which the original subject’s face has been replaced by
Monica Lewinsky’s, he is cleverly blending elements from our knowledge of the

famous painting with ones from our knowledge about Monica Lewinsky.
10
As we
take in this picture, our mental representations of each of these phenomena con-
stitutes an ‘‘input space’’—one real and one based on an artistic image—projecting
material into a third space, the ‘‘blend space’’ represented in the magazine cover.
The fourth space in Fauconnier and Turner’s scheme is the ‘‘generic space,’’
containing material shared by the two inputs; in this case, for example, the image
of a dark-haired young woman wearing a subtle and knowing smile. (Like CMT,
blending theory is concerned with nonlinguistic as well as linguistic examples.)
Monica Lewinsky and the Mona Lisa are counterparts in the respective input
spaces; without such counterparts to establish relations between the inputs, a blend
cannot ‘‘get off the ground.’’
Like the Leonardo da Vinci/Lewinsky cover, many of the examples discussed
by blending scholars are not exactly metaphorical, though they may be figurative in
some sense. Some blends simply involve a juxtaposition of elements that do not co-
occur in reality; Fauconnier and Turner (1994) have discussed a blend, for example,
in which a modern philosopher is engaged in a figurative ‘‘debate’’ with Immanuel
Kant. The philosophers are real and literal, as are the philosophical issues and ar-
guments. The only figurative aspect of the imagined scene is that the discourse
takes the form of an in-person debate rather than two sets of writings from dif-
ferent places and historical periods.
In this framework, metaphors are treated as a subset of conceptual blends,
characterized by particular kinds of relations holding among the various spaces.
Fauconnier and Turner (1998) present a typology of blends in which metaphors are
defined by an asymmetry in the degree to which two inputs provide the conceptual
frames that structure the blend. In Ship-of-State blends, the input space of ships
would be said to provide a structuring frame (including elements such as weather,
heading, and so forth) within which such topics as elections and policy are de-
picted. Grady, Oakley, and Coulson (1999) characterize metaphorical blends as ones

involving ‘‘fusion’’ of corresponding elements from the two inputs, where a target
concept is excluded from explicit representation in the blend in favor of its
counterpart from the source.
11
In a Ship-of-State blend, for example, the nation
metaphor 199