Sunday, October 31, 2004

Metaphor I: A Brief History of Metaphors in Cognitive Science

From Aristotle through speech act theories, metaphor had been viewed as a secondary type of language, built on literal speech which is, in turn, the true nature of language. However, since the 1970s, cognitive scientists have become increasingly convinced that metaphor is not only central to thought, something that Aristotle would admit, but that it is also a central aspect of language, and no less priveleged than literal language. Metaphors are processed as quickly as literal language, and as automatically. In addition, metaphors, while generally literally false, are difficult to label as literally false. This new status for metaphor has led to a great deal of attention among cognitive scientists, and a wealth of theories and models of metaphor. It would be prohibitively time-consuming (and unthinkably pointless) to detail each of these models along with the evidence for them, so instead I will focus on two models, the structure mapping model of Gentner and colleagues, and the attributive categorization model of Glucksberg and Keysar. These models are the most prominent, and the most empirically viable. Furthermore, they capture two diametrically opposed (at least ostensibly) ways of viewing metaphor. Also, I refuse to talk about Lakoff and cognitive linguistics until after the election, which means that cognitive linguistics theories of metaphor are off the table. Before I get to the current theories, though, I'll give a very brief history of cognitive views of metaphors.

Originally, almost all theories of metaphor were two-process theories, which involved metaphorical statements first being processed as literal statements, and when it was discovered that the statements did not work as literal statements, they were then, and only then, processed as metaphors. This view was already on its way out in 1982 when a classic experiment by Glucksberg et al. 1 made it completely untenable. In their experiment, participants read three types of class-inclusion statements. The statements were either literally true ("Some birds are robins"), literally false and "anomolous" (i.e., they couldn't easily be interpreted metaphorically, as in "Some birds are apples"), or metaphorical ("Some lawyers are sharks"). Participants were asked to judge whether the statements were literally true, and their reaction time in making this judgement was measured. Participants were quick to judge literally true statements as true, and literally false and anomolous statements as false, but they were significantly slower when rating metaphorical statements as literally false. This was interpreted to mean that the participants were automatically interpreting the metaphorical statements metaphorically, and a subsequent literal interpetation required more time. This rules out the view that the statements must be interpreted literally first, prior to any metaphorical interpretation. If this were in fact the case, then participants should have noticed that they were literally false in about the same time that they noticed the falsity of the anomolous statements.

Cognitive scientific views of metaphor that do not treat them as less priveleged than literal statements began to surface in the 1970s, andsince then almost all of them have arisen out of the work of Max Black. Black2 discussed several different cognitive theories of metaphor, the most prominent of which were the substitution, simile, and interactive theories. In the substitution theory of metaphor, the cognitive processing of metaphor involved substituting a property of the vehicle for the vehicle itself. So, "My surgeon is a butcher" becomes "My surgeon is sloppy," or "My lawyer is a shark" becomes "My lawyer is aggressive." The simile theory is similar, in that it views metaphors as highlighting properties of the vehicle that can be attributed to the topic. However, under this view, metaphors are essentially substituted with a corresponding simile (shades of Aristotle). "My surgeon is a butcher" becomes "my surgeon is like a butcher," and the comparison highlights the pragmatically relevant common attributes. Finally, the interactive theory of metaphor again involves a comparison between the vehicle and topic, but in this case, the properties that are highlighted by the comparison are determined by the interaction of the topic and vehicle.

In general, the interactive theory of metaphor has been the most influential of Black's various theories, and is ultimately the one that he himself accepts. However, it does suffer from a glaring problem: it's extremely abstract. I wish I could tell you that it was more detailed than my short description in the preceding paragraph, but it's really not. How do the topic and vehicle interact? How are the relevant properties selected? These questions would need to be answered before we could claim that the interactive account of metaphor is sufficient as a cognitive theory. To rectify these defects, several theories based on the interactive theory have been proposed over the last few decades. The first of these, called the salience imbalance theory3, solves the problem of feature selection by positing that metaphors involve comparisons between topics and vehicles that exhibit an imbalance in the saliency of the properties the creator of the metaphor wishes to hilghight. For instance, sharks exemplify aggression in a much more salient way than lawyers, and by making the comparison in the metaphor "My lawyer is a shark," the salience of the "aggressive" feature in sharks serves to highlight this feature in my lawyer. One of the primary motivations for this theory is the assymetry in metaphorical comparisons. "My lawyer is a shark" is a much more acceptable, and powerful statement (in most contexts) than, "The shark is a lawyer." This theory, and many subsequent theories, have been motivated by a desire to explain this inherent assymetry in all metaphors.

The salience imbalance theory, while it is an improvement on the interactive theory, still suffers from many problems. It doesn't detail how the comparison takes place, and its explanation of how properties are highlighted is still somewhat abstract. After the salience imbalance approach, theories of metaphor split into three different types, all still roughly based on Black's work, but with different solutions to the problems of Black's and Ortony's theories. The first type treats metaphor as an instance of analogy, or more accurately, as using the same processes as analogy. Under this view, called structure mapping, metaphor involves mappings between the topic and vehicle. Thus, it is also a comparison view of analogy, but one with a much more sophisticated comparison process than either the interactive or salience imbalance theories. The second type treats metaphors as categorization statements. Specifically, under the attributive categoriztion theory, metaphors are treated as class-inclusion statements, which involve placing the topic into a category defined by a feature or features exemplified by the vehicle. For example, "My lawyer is a shark" involves placing "My lawyer" in the category of "aggressive things" of which "shark" is a highly salient or typical member. The final widely used type of cognitive theory of metaphor comes out of cognitive linguistics, and is either based on Lakoff's conceptual metaphor theory, or blending. However, talking about this will have to wait until after the election, because I don't want to type type name "Lakoff" again until then.

In the next post, I'll describe the structure mapping view of metaphors in more detail. This theory is very interesting because it comes up with a way to classify different types of metaphors based on the sorts of arguments that are mapped between the topic and vehicle. After that, I'll discuss the attributive categorization model in a third post. After that, who knows what will happen?

1 Glucksberg, S ., Gildea, P., & Bookin, H. (1982) . On understanding non-literal speech : Can people ignore metaphors? Journal of Verbal Learning and Verbal Behavior, 1, 85-96.
2 Black, M. (1955). Metaphor. Proceedings of the Aristotelian Society, 55, 273-294, Black, M. (1962). Models and Metaphors. Ithaca, NY: Cornell University Press, and Black, M. (1979). More about metaphor. In A. Ortony (Ed.), Metaphor and Thought (pp. 19-43). Cambridge, England: Cambridge University Press.
3 Ortony, A. (1979). Beyond literal similarity. Psychological Review, 86, 161-180, and Ortony, A. (1993). The role of similarity in similes and metaphors. In A. Ortony (Ed.), Metaphor and Thought 2nd Ed., (pp. 342-356). Cambridge, England: Cambridge University Press.


8 comments:

Anonymous said...

Have you read any of Ricoeur's stuff on metaphors?  

Posted by Clark Goble

Anonymous said...

I have read Ricoeur's work on metaphor. As far as I can tell, it hasn't really had an influence on cognitive scientific views of metaphor. At least, I've never seen him cited, except maybe in some cognitive linguistic literature. In fact, most cognitive accounts would be inconsistent with his view of metaphor. In the cognitive linguistic view, literal speech is built on top of metaphorical language (an inversion more Derridaian than Ricoeurian), and in more mainstream cognitive science, literal and metaphorical language are equally priveleged and involve the same processes (either those involved in other types of comparisons, or those involved in other class-inclusions). Ricoeur's view is more similar to speech act theories of metaphor (even if they are meant, in part, to oppose them). In those theories of metaphor, figurative language is "parasitic" on literal language. In Ricoeur's, metaphor is built on the "ruins" of the literal, only after the literal has been destroyed. This is different from parasitism, but it still requires that metaphor be built on top of, and more importantly, after, the literal. This might be a good way to think about creative metaphors (e.g., those found in poetry or fiction), but it's pretty clear that in commonplace metaphorical speech, metaphor doesn't come after anything. We create, notice, and interpret metaphors automatically, and without the literal meaning either getting in the way or being destroyed in the process.  

Posted by Chris

Anonymous said...

Interesting. I confess that his work on metaphor is one of the main works by Ricouer I've never read up on, although he mentions it a bit in Time and Narrative. I was, coincidentally, reading Peirce on the notion last night. Of course I know Peirce is very influential among various people in cognitive science. 

Posted by Clark Goble

Anonymous said...

Pierce is influential in cognitive science, but I don't think he's had much direct influence on cognitive approaches to metaphor production and comprehension. His discussion of metaphor is relatively sparse, and difficult to penetrate. He gives very good descriptions of images and diagrams, but when he talks of metaphors, it's always fairly abstract. I, for one, am not exactly sure how to analyze instances of metaphor from a Piercean perspective. A metaphor is a diagram of a diagram, but comparisons are diagrams. Does this mean that metaphors can be considered diagrams of comparisons? If I say, "My lawyer is a shark," what is the diagram, and what is the diagram of the diagram? Is the lawyer the diagram, and the shark the diagram of the lawyer? As you can problably tell, when it comes to Pierce's discussion of metaphors, I'm lost. 

Posted by Chris

Anonymous said...

You might be interested in the following paper. Metaphor-Icon Link in Poetic Texts: A Cognitive Approach to IconicityThere's also an article (I confess I've not read) in the Transactions of the C S Peirce Society, back in 1965 on Metaphors by Peirce.  

Posted by Clark Goble

Anonymous said...

Whoops, that should have been the 1984 issue of TotPS not 1965. 

Posted by Clark Goble

breakdown said...
This comment has been removed by the author.
Jairo said...

Good job!! I like your blog!!