|
|
Questions and Summary
by Randall Rose and Koichi Miyakoshi
Summary of Ray Gibbs's
"How Language Reflects the Embodied Nature of Creative Cognition",
and other experimental readings
by Randall Rose
This reading considers the psychological evidence for metaphor which Gibbs's
experiments have found. (The summary below also discusses the published
experiments and gives some commentary.)
Gibbs argues that nonbodily concepts often arise from bodily experience; to
describe the link between the concept and the bodily experience, he uses the
term "metaphor". He acknowledges that there can be metaphors which are not
based on bodily experience, but they aren't what he's concerned with.
Problem #1: What's the difference between "Embodied Metaphor"
and "Abstraction from Experience"?
Gibbs's "provocative" claim, that abstract concepts arise by metaphor from
embodied reality, is sometimes hard to distinguish from the traditional
empiricist view, that these concepts come from abstraction from experience.
Generally the same evidence can be used to support either view. Gibbs
acknowledges that the "metaphor" which forms an abstract concept does not
preserve all aspects of the original bodily experience; so "abstraction from
experience" might be a better term.
The real originality of the Embodied Metaphor view is the claim that something
of the "embodied" nature of the original experience still operates whenever an
abstract concept is used. (This claim is often hedged, but it's clearly
implied in several places.) Gibbs states that the original experience persists
in a schematic form called an "image schema". He defines image schemas as
"analog representations of spatial relations and movements in space". These
image schemas are "constitutive of our understanding of intentional action" and
of other abstract concepts.
For example, Gibbs lists a number of bodily experiences of the form "Some part
of the body is too hot/cold/etc.":
My head is too hot
My hand is too cold
My mouth feels too dry
My sinuses are swollen
My posture's too tilted
He states that these are all instances of a single image schema, "the balance
schema" (although only the last example is spatial). This image schema is
claimed to be used in many kinds of abstract thought. For instance, in
Hamlet's speech "To be or not to be", Gibbs says that "one quickly recognizes"
an allusion to the balance schema because it involves making a difficult
decision, even though Shakespeare's text never mentions any bodily experience
like balance. It's not clear how Gibbs knows that all these things are
understood through the same image schema.
Problem #2: How do we know that metaphors and image schemas
are used in understanding?
To provide evidence that embodied metaphor is actually used on-line in
thought, Gibbs performs several kinds of experiments.
Experiment Series 1: Literary Metaphor
Gibbs asked subjects to write about love and to analyze love poems. The
subjects' writing showed a limited number of metaphors for love, which shows
that some metaphors have become standard in our culture. Subjects also showed
that they had a good conscious understanding of these metaphors. The subjects'
commentaries on love poems tended to use the same metaphors as the poems
themselves did.
These experiments are the only ones in which Gibbs touches on "creative
cognition". It's clear that people are able to consciously understand
literature through conventional metaphor. But it still seems unclear whether
this is really enough to show that metaphor is essential to creative thought.
Experiment Series 2: Idioms (in the "Idiom Comprehension" paper)
Gibbs et al. (1997) examine idioms like "blow one's stack" which have three
kinds of meanings: an idiomatic meaning, a literal meaning which is the
historical source of the idiomatic one, and a rival literal meaning which is
unrelated.
Meanings of "to blow one's stack"
Idiomatic meaning Source literal meaning Rival literal meaning
(historic source of idiom)
to get angry to explode by boiling to thrust air
(key word: "heat") through a column
These experiments used a word (e.g. "heat") which was directly relevant only
to the source literal meaning, not to the other two meanings. The experiments
showed that this word was easier to recognize after the idiom was used in its
idiomatic meaning ("to get angry"), but not after the idiom was used in its
rival literal meaning ("to thrust air through a column").
What this seems to show is that one literal meaning of the phrase is able to
inhibit another literal meaning; but an idiomatic meaning doesn't inhibit the
literal meaning which was the idiom's original source. Gibbs concludes that a
metaphor, the relation between the idiomatic meaning and its literal source, is
accessed when the idiom is understood. But this conclusion doesn't seem
necessary; perhaps the literal meaning is simply not a competitor for the
idiomatic meaning, and is not inhibited for that reason.
Experiment Series 3: Polysemy (in the "Taking a Stand" paper)
This series of experiments (Gibbs et al., 1995) considered 32 different uses
of "stand" ("to stand against great odds", "it stands to reason", "the parts
stand for the whole", etc.). The experiments attempted to classify these
senses into significant groups and to show that these groups are motivated by
image schemas.
Gibbs et al. asked the subjects to sort these 32 uses of "stand" into groups
of their own choosing. To get a consistent classification, the experiments
counted the number of times each each pair of senses were classed together by
subjects and performed a cluster analysis, which ended up classifying the 32
uses into 3 groups. The authors admit that none of the three resulting groups
seem to have a coherent meaning; e.g. "to stand the test of time", "to stand
firm", and "to stand against great odds" were placed in three different groups.
There's probably some degree of meaningfulness that's reflected in this
classification, but my impression was that the cluster analysis failed to pick
out the most meaningful divisions.
Gibbs et al. then asked other subjects (who had been asked to pay attention to
their experience of standing and moving) to rate each of the 32 uses of "stand"
according to how well they fit various image schemas; the 5 image schemas used
were the ones that previous subjects had rated as most closely related to their
experience of standing. The assumption was that uses of "stand" which the
original cluster analysis had classified together would also have similar
image-schema ratings. A discriminant analysis produced a significant
discriminant function which, when given the image-schema ratings for a given
use of "stand", could predict with 80% accuracy which of the 3 cluster-analysis
groups it should be assigned to.
Interestingly, when Gibbs et al. replaced the 32 phrases containing "stand"
with near-synonyms that lack "stand", the new phrases turned out to have
slightly different image-schema ratings, from which it was impossible to
significantly predict the original classification. The data turn out to
undermine Gibbs's point, however: the near-synonyms without "stand" ended up
having slightly higher image-schema ratings than the originals with "stand",
even though these image schemas were chosen for their close relation to the
experience of standing. If image schemas were as closely related to meaning as
Gibbs suggests, near-synonyms which lack the word "stand" should score lower on
"stand"'s favorite image-schemas.
These experiments do make it clear that image-schema ratings reflect at least
a part of the meaning of phrases, and that they're sensitive enough to capture
differences among near-synonyms. But any semantically-related measure should
be able to do this; to show that image schemas play a fundamental role in
meaning, one would expect to see much more reliable predictions from the
image-schema ratings. We still haven't seen sufficient evidence for any strong
claim.
--Randall Rose (rrose@ling.ucsd.edu)
Summary of Gibbs (1996): "Why many concepts are metaphorical", Cognition 61:
309-319.
(summarized by Koichi Miyakoshi)
There is currently a debate over metaphor between Raymond Gibbs and Gregory
Murphy in the journal Cognition. Gibbs argues for it, and Murphy against it.
The following is a summary of Gibbs's (1996) reply to Murphy's (1996) arguments
against metaphor (Murphy, G. 1996. "On metaphorical representation", Cognition
60: 173-204). After the summary, I raise some questions, touching upon
Murphy's (1997) response (Murphy, G. 1997. "Reasons to doubt the present
evidence for metaphoric representation", Cognition 62, 99-108).
Gibbs (1996) addresses the following four issues questioned by Murphy (1996):
[1] the necessity of metaphor, [2] the motivation for metaphor, [3] the
validity of linguistic arguments for metaphor, and [4] the consistency of
multiple metaphors in the mental representation.
[1] The necessity of metaphor
Murphy argues that the notion of structural similarity is sufficient to
account for linguistic expressions (both conventional and novel), and
questions the existence and necessity of metaphor. For example, he claims that
polysemy can be explained by the abstract similarity between different senses
of a word without postulating the existence of metaphor. Gibbs disagrees with
this view, making the following two arguments: First, he argues along with
Brugman and Lakoff (1988) and many others that it is almost impossible to
specify the abstract features or relations of similarity underlying all the
different senses of a polysemous word. Second, he argues, following Sweetser
(1990), that the similarity view fails to capture the asymmetrical relationship
between the physical and nonphysical meanings of a polysemous word in the
course of historical development. Consider the English verb "see" for example.
It can be used in physical (e.g., I can see a bird) and nonphysical (e.g. I can
see the point) senses. Even if this polysemous word can be described in terms
of the abstract similarity between the two senses, this view cannot explain the
observation that the nonphysical sense is acquired AFTER the physical sense in
English and many other Indo-European languages, and that the same kinds of
meaning changes have occurred over and over again in the history of those
languages. The metaphor view, on the other hand, can explain why this is the
case: Given the conceptual metaphor UNDERSTANDING IS SEEING (the metaphorical
mapping of our knowledge about vision onto the domain of understanding), then
it is likely that words meaning SEE extend to their meanings to UNDERSTAND
whereas the mapping in the opposite direction is unlikely.
[2] The motivation for metaphor
Murphy questions the motivation for metaphorical representations, and argues
that metaphor theories fail to explain why certain conceptual metaphors, but
not others, are used by people in speaking about abstract concepts. In
response to this question, Gibbs points out that Murphy does not pay enough
attention to the important role of human embodiment in theories of metaphor,
suggesting that it leads him to underestimate the significance of cognitive
linguistic and psycholinguistic work on metaphor. Gibbs argues along with
Johnson (1987), Lakoff (1987), and many others that metaphorical understanding
of abstract concepts is grounded in nonmetaphorical structures that emerge from
our everyday bodily experiences; more precisely, people's recurring bodily
experiences give rise to the development of various image schemas (experiential
gestalts), which are in turn metaphorically elaborated to help them understand
more abstract domains of experiences. For example, consider the metaphor ANGER
IS HEATED FLUID IN A CONTAINER. Our recurring, kinesthetic experiences of
bodily containment give rise to the development of the image schema for
CONTAINMENT, and it provides the (partial) basis on which we can understand and
structure abstract concepts about emotions like anger. That is, the conceptual
metaphor takes the image schema for CONTAINMENT as part of its source domain
and maps this image-schematic structure onto anger, thereby allowing us to
understand and structure anger experiences in terms of containment. This
metaphor helps us make sense of figurative expressions like the following:
(1) His pent-up anger welled up inside of him. (We know that when the
intensity of anger increases, the fluid in the container rises.)
(2)a. Bill is getting hot under the collar. b. Jim's just blowing off
steam. c. He was bursting with anger. (We know that intense heat produces
steam and creates pressure on the container.)
(3) She blew up at me. (We know that when the pressure of the container
becomes too high, the container explodes.)
Gibbs argues that figurative expressions like these are overwhelmingly
ubiquitous, and it illustrates the embodied motivation for metaphor and reveals
the metaphorical nature of human cognition.
[3] The validity of linguistic arguments for metaphor
Murphy repeatedly claims that arguments for metaphorical representations on
the basis of linguistic evidence (like the one reviewed just above) are dubious
since they presuppose that a certain pattern in speech directly reflects
underlying conceptual structure and thought. He thus argues that nonlinguistic
evidence is needed to buttress the idea that metaphor is a fundamental mental
capacity by which we conceptualize many experiences. In reply to this claim,
Gibbs points out that there are psychological studies showing the influence of
metaphor in gesture, categorization, decision-making, learning, and memory (cf.
Gibbs 1994). He also points out that there are psycholinguistic studies
suggesting important connections between people's use and understanding of
figurative, linguistic expressions and their underlying, nonlinguistic
understanding of concepts. For instance, Gibbs (1992) experientially examined
what inferences people draw when they understand idioms motivated by the
metaphor ANGER IS HEATED FLUID IN A CONTAINER (e.g., blow your stack, flip your
lid, hit the ceiling, etc.). The result was that participants' inferences were
remarkably consistent. For example, they inferred that the cause of anger is
internal pressure, and that the expression of anger is unintentional, and is
done in an abrupt, violent manner. These inferences are all predictable from
our nonlinguistic knowledge about heated fluid in a container. Thus this
result suggests that an important connection exists between people's
nonlinguistic understanding of the source domain concepts and their
understanding of the metaphorically motivated linguistic expressions.
[4] The inconsistency of multiple metaphors in the mental representation
Cognitive linguists have argued that one concept (e.g. love) can be understood
through several different metaphors (e.g., LOVE IS A JOURNEY, LOVE IS INSANITY,
LOVE IS AN OPPONENT, LOVE IS A VALUABLE COMMODITY). Murphy warns that such
multiple metaphors would result in inconsistency in the mental representation
since some of them are inconsistent with one another. Gibbs replies that
concepts are not fixed, static structures but rather temporary, dynamic
representations that are created on the spot in working memory on the basis of
generic and episodic information in long-term memory (which he calls
"knowledge" as opposed to "concept"). He argues that this dynamic view of
concepts allows us to conceptualize one experience in different ways at
different times and to access different aspects of one piece of knowledge. For
example, The LOVE IS A JOURNEY metaphor might be appropriate to create a
particular conceptualization of love in certain situations, whereas LOVE IS AN
OPPONENT might be suited for creating another concept of love in other
situations. Gibbs argues that concepts are independent of each other as a
temporary representation in working memory apart from source domain information
in long-term memory, thereby accommodating multiple metaphors while avoiding
the problem raised by Murphy.
Questions:
I basically agree with Gibbs's arguments and find his (and many others') story
about the embodied motivation for metaphor particularly interesting. However,
there are a number of fundamental questions yet to be answered (or to be asked
in the first place). Here I would raise three types of questions: questions
about (1) the mental status of the crucial notions, (2) the leaning mechanism,
and (3) directionality of metaphorical mapping. First of all, as Murphy (1997)
points out, many crucial notions have not been explicated yet. Even the most
important notions "metaphor", "metaphorical concepts", and "image schemas" have
not been made very explicit yet: Exactly what are these concepts? How are
they represented in the mind/brain? How are they accessed and manipulated?
To take the metaphor UNDERSTANDING IS SEEING as an example again, is this
entrenched in the mind/brain with some mental status? If so, exactly what is
the mental status of this metaphor? As we saw above, Gibbs assumes that
metaphorical concepts are created in working memory, but does it follow from
this assumption that the lexical entry of the verb "see" does not even
explicitly list the sense of "see" which metaphorically means "understanding"
(i.e., the sense of "understanding" used in sentences like "I see the point")?
If it turns out not to be listed, then he might not have to revise his analysis
(although the story seems highly counter-intuitive). But if it turns out to be
listed (which seems likely), then his claim about temporary concepts would need
to be qualified. In any case, is there any evidence for it? What about image
schemas? (He suggests a few possibilities in Gibbs (1997) and Gibbs and
Colston (1995): (i) they are "dynamic analog representations of spatial
relations and movements in space", (ii) they "never get encoded as explicit
mental representations", and (iii) they are "characterized as emergent
properties of our ordinary conceptual systems".) In any case, it should be
addressed whether they are represented in the mind/brain, if they are, exactly
how. If not, how are they made use of in our metaphorical conceptualization?
Secondly, the learning mechanism of metaphor is also unclear, although the
overall picture (image schema formations based on bodily experiences followed
by metaphorical extensions) is very clear and seems to be on the right track.
How are metaphors like ANGER IS HEATED FLUID IN A CONTAINER learned, if at all?
Is there any general learning principle? What about image schemas and
metaphorical concepts? If they are really emergent properties and only have
temporary representations in working memory, then it follows that they are not
learned in the first place, but is there any evidence for that?
Thirdly, as Murphy (1997) also points out, it is true that metaphorical
mappings are usually unidirectional from a concrete, physical domain to a more
abstract, nonphysical domain (recall LOVE IS A JOURNEY), but not always. For
example, consider the metaphor LOVE IS A FINANCIAL TRANSACTION. In this
metaphor, the source domain concept (a financial transaction) is not bodily and
even seems less bodily than the target domain concept (love). Thus the
question is how to reconcile metaphors of this kind with the basic idea about
the embodied motivation for metaphor.
|
