Position Paper on Metaphor 
George Lakoff 
University of California, Berkeley 
My research has focussed mainly on conventional metaphor built into 
the conceptual system on which our everyday language is based. What I 
and others have found is that metaphor plays a major role in semantics. It 
is the means by which domains of experience that are not highly structured 
on their own terms get structured on the basis of other, highly-structured 
domains. Domains need structure so that one can reason about them. The 
major function of metaphor is thus to supply structure in terms of which 
reasoning can be done. What we have-found is that an enormous amount 
of our reasoning is metaphorical (see Lakoff, 1986b; Lakoff and Johnson, 
1987; Johnson, 1087; Turner, 1987; Holland and Quinn, 1987; and espe- 
cially Quinn, in preparation). 
Metaphors are systematic structure mappings from one domain of 
experience to another. Metaphors occur at the level of concepts, not 
linguistic expressions, and in most cases they are fairly general, occurring 
at a superordinate level rather than the basic level. A single such general 
metaphor will give rise to many individual metaphorical expressions. I 
have encountered four general types of metaphor: 
(1) Complex schema mappings: These map complex schemas in one domain 
(e.g. WAR) into corresponding schemas in another domain (e.g., ARGU- 
MENT). Each such mapping applies both to entities (the source ontology is 
mapped onto the target ontology) and relations holding among the entities 
(knowledge about the source is mapped onto knowledge about the target). 
(2) Image-schema mappings: Image-schemas are general topological and 
orientational structures that are kinaesthetic in nature. They have an ana- 
log rather than digital character. And they have sufficient internal struc- 
ture to permit inferences. Examples include: containers, paths, linear 
scales, center-periphery, force, links, balance, contact/noncontact, cycles, 
front/back, etc. A great many conventional metaphors are based on such 
schemas. For example, purposes are understood metaphorically as destina- 
tions, and achieving a purpose is understood as traveling along a path to 
that destination. 
(3) One-shot rich-image mappings: Consider the word dunk as applied to (i) 
to cookies and milk, donuts and coffee, etc. and (ii) to basketball. There is 
a conventional rich image for cases like (i); it involves a hand putting a 
piece of food through the rim of a cup or glass into liquid. In (ii~, a hand is 
putting a basketball through the rim of the basket. There is a partial 
194 
mapping from the image in (i) to the image in (ii). The extension of the 
word dunk from food to basketball is a metaphorical extension based on 
this mapping from one conventional image to another. This is a 'one-shot' 
mapping. That is, there is no system of concepts being mapped. The 
mapping sanctions the lexical extension of only one word. 
(4) Aristotle's metaphor: This is a single, very general metaphor of the fol- 
lowing form: SOMETHING IS WHAT IT HAS SALIENT PROPERTIES OF. It 
relates entities in one domain to enties in another domain, on the basis of 
common properties. Unlike other general metaphorical mappings, which 
have fixed domains, this general metaphor seems to have variable domains. 
It is this metaphor that gives rise to relatively boring cases like Man is a 
wolf, Harry is a pig, etc., which are unfortunately the cases most cited in 
the classical metaphor literature. Oddly enough, little of a systematic 
nature is known about this metaphor (e.g., whether there are restrictions 
on its domains). 
Actually, these four types of metaphor are not mutually exclusive and 
mixed cases are common. For example, (3) and (1) are sometimes combined 
in cases where complex knowledge about the source image is carried over 
into the target domain. Moreover, since complex schemas tend to have 
image-schmatic subparts, it is common for cases of (1) to also be cases of 
(2). 
On the whole, I have found (1) and (2) the most interesting types for 
a number of reasons: 
(A) They are not based on similarity; though they may create structural 
similarities. 
(B) They tend to create structure and even entitites in the target domain. 
(C) They are used in reasoning; indeed, that is their main function. 
The image-based metaphorical mappings -- (2) and (3) -- are interesting for 
another reason: they cannot be adequately characterized in finitary terms. 
Any digital rendering will miss generalizations. 
I have found (4) to be the most boring, and also the rarest of the 
types (though none of them is particularly rare). Cases of Aristotle's meta- 
phor become interesting mainly when they are combined with metaphors of 
other types, as when the similarities they are based on are created by other 
metaphors. 
195 
Literary metaphor 
A great many literary metaphors are extensions of ordinary conven- 
tional metaphors of type (1) and (2). Allegory appears to be a case where 
an existing conventional metaphor of type (1) is sustained over an entire 
literary work (see Turner, 1987). It is useful to distinguish such stable 
metaphors in a literary work from fleeting metaphors (those that occur 
once and disappear). Fleeting literary metaphors tend to be of types (3) 
and (4), which incidentally are based on similarity and are not used all 
that much in reasoning. 
Scientific analogies 
Some scientific analogies are extensions of conventional metaphors of 
type 1. For example, the mind-as-computer metaphor is an extension of the 
old mind-as-machine metaphor. In general, useful scientific analogies seem 
to have the properties of metaphors of type 1; that is, they are systematic, 
well-defined, and map high-level abstract relations such as causation. They 
thus give rise to a great many metaphoric entailments. In this way, useful 
scientific analogies are structurally just like most ordinary conventional 
metaphors and like stable metaphors (including allegories) in literary 
works. Correspondingly, scientific analogies are unlike most fleeting 
literary metaphors (those of types 3 and 4). 
Constraints on the Mappings 
Perhaps the central theoretical question in metaphor research today 
is: What determines the details of the mappings. What determines the 
choice of source and target domain? What determines which entities get 
mapped onto which other entities? A partial answer comes from the 
theory of experiential bases: regular recurrent correlations in experience 
between source domain and entities and target domain entities determines 
the ontological mappings. This works for a certain range of image-schema 
mappings (those of type 2). In cases of complex schema mappings, explana- 
tions are sometimes attainable (see Lakoff, 1986b) in terms of inferences 
based on other metaphors with experiential bases and their interaction 
with folk knowledge. In general, our knowledge at present is sketchy, to 
say the least. 

Bibliography
Holland, Dorothy and Naomi Quinn. 1987. Cultural Models in Language 
and Thought. Cambridge University Press. 
Johnson, Mark. 1987. The Body in the Mind. University of Chicago Press. 
Kuhn, Thomas. Unpublished ms. Scientific Development and Lexical 
Change. 
Lakoff, George. 1986a. Women, Fire, and Dangerous Things. University of 
Chicago Press. 
Lakoff, George. 1986b. A Figure of Thought. In Metaphor and Symbolic 
Activity, 1:3. 
Lakoff, George. 1986c. The Meaning of Literal In Metaphor and Symbolic 
Activity, 1:4. 
Lakoff, George and Mark Johnson. 1987. The Metaphorical Logic of Rape. 
In Metaphor and Symbolic Activity, 2:1: 
Lakoff, George and Zolt£n Kovecses. 1983. The Concept of Anger Inherent 
in American English. In Holland and Quinn. 
Lindner, Susan. 1982. What goes up doesn't necessarily come down: The 
ins and outs of opposites. In Vol. 18, Chicago Linguistic Society. 
Lindner, Susan. 1981. A Lexico-Semantic Analysis of English Verb Particle 
Constructions with OUT and UP. UCSD Ph.D. Thesis. pp. 75 - 230. 
Quinn, Naomi. In Preparation. American Marriage: A Cultural Analysis. 
Rhodes, Richard and John Lawler. 1981. Athematic Metaphor. In 
Proceedings of the Chicago Linguistics Society. 
Roediger, Henry L. 1980. Memory Metaphors in Cognitive Psychology. In 
Memory and Cognition, Vol. 8 (3), 231-24{}. 
Sweetser, Eve. 1984. Semantic Structure And Semantic Change. UCBerke- 
ley PhD Dissertation. In press, Cambridge University Press. 
Turner, Mark. 1987. Death is the Mother of Beauty. University of Chicago 
Press. 
