Abstract Barnden

Metonymy Considerations Arising from a Metaphor Project

John A. Barnden

School of Computer Science
University of Birmingham, England

Extended Abstract

I have implemented a system called ATT-Meta for handling some of the reasoning needed in the understanding of metaphorical utterances (see Barnden and Lee references below). The focus has been on metaphors for mental states and processes, but the implemented approach is not restricted to these. As part of the work I have conducted some metaphor data collection from real discourse (click here for some of the results), and the system has been applied to simplified versions of some real examples. Despite the simplifications the required reasoning is still complex and subtle.

Although the ATT-Meta system does not handle metonymy, I am interested in making it do so eventually. Some interesting issues about metonymy have come to the fore in the work on the project. These issues are described below. Some are related to metonymy used in language about pictures, models, etc. My talk will largely be a general discussion of some of the issues (it is unlikely I will be able to address all of them). I also include below a metaphor issue that does not immediately bring in metonymy but that may be of interest to the workshop.

ATT-Meta is purely a reasoning system, and does not currently take natural language input. Rather, a user specifies logical forms encoding the information in a possible discourse chunk, rules encapsulating domain knowledge, rules encapsulating known metaphorical mappings, and a query. In trying to answer the query by means of the provided information and general reasoning processes, the system is intended to create useful connotations of the discourse chunk.

ATT-Meta does not discover new metaphorical mappings, but seeks to exploit existing mappings to the full. Indeed, one fundamental principle in the research is the claim that much or most metaphorical understanding should seek as far as possible to avoid creation of new mappings or the extension of old ones. There is an unwarranted bias in the metaphor literature towards mapping-creation. (Mapping-creation/extension is, nevertheless, sometimes needed.)

ATT-Meta incorporates a powerful general-purpose engine for qualitatively uncertain reasoning. This is because metaphorical understanding strongly involves uncertainty of various different types.

A Common Mental Metonymy

It appears that a metonymic schema I call THING FOR IDEA OF IT is often used in metaphorical utterances about mental states, as in the sentence


         China was right at the surface of John's mind.

The metaphorical element here is the use of the physical-location word ``surface.'' I therefore analyze the sentence as using the very common metaphorical view of MIND AS PHYSICAL SPACE.

Now, presumably it is some idea or ideas associated with China, not China itself, that was (metaphorically speaking) physically inside John's mind. Thus, the word ``China'' is being used metonymically to refer to an idea or collection of ideas connected to China. (Note, in passing, that it is therefore likely that a further metaphorical view of IDEAS AS PHYSICAL OBJECTS is being appealed to by the sentence.)

As for what idea or ideas are actually being metonymically referred to, one possibility is that it is John's idea of China as a whole that is being referred to. In this case I say that the metonymic schema of THING FOR IDEA OF IT is involved. Another possibility is that the reference is to certain issues that have arisen in the discourse context and that concern China. The metonymic schema in this case is THING FOR ISSUES ABOUT IT. I count issues as a type of idea.

There is a competing, entirely metaphorical analysis, viz. that the sentence uses a MIND AS PHYSICAL SPACE view that allows external objects themselves, rather than ideas of them, to be within the mind-space. Although this account is an interesting possibility, there is no independent reason to adopt it, and several reasons not to.

For one thing, there is independent motivation for proposing that sentences such as the above involve the metonymic schema of THING FOR IDEA OF IT. The schema is a special case of THING FOR REPRESENTATION OF IT, which is often used in describing pictures, models, etc. Reference to an object is often used metonymically to refer to a part of a picture, etc., that represents that object. For instance, in talking about a picture depicting various people, we might say ``John is right at the edge of the picture.''

Mixing of Metonymy and Metaphor

In the previous section we saw an example of the mixing of metonymy and metaphor. Such potential mixing makes metaphorical understanding yet more a matter of uncertain reasoning than it would otherwise be, because

it can be uncertain as to whether there is in fact a mix of metonymy and metaphor at all (recalling that in the previous section there was a competing non-metonymic analysis),
it can be uncertain what the precise mix is, and
in metonymy generally, there can be uncertainty as to what it is that is being indirectly referred to (in the China example, is it the idea of China as a country, or is it issues connected to China?).

There have been few attempts to devise computational systems that can handle metaphor and metonymy in an integrated way. Fass (1997) and Hobbs (1990) provide major approaches to this, but much further work remains to be done, especially with regard to handling the uncertainties.

Mental States Ascribed to Non-Persons

Mental states are often to non-persons such as institutions, nations, +committees, geographical areas, vehicles, buildings and so on, as in the sentence


        The committee thinks that the procedure should be scrapped.
        
        That car wants to overtake me.

Such ascriptions are often analyzed metonymically in the literature. According to this view, the phrase ``The committee'' refers metonymically to some majority of the membership, or possibly to an imaginary typical member; and the phrase ``That car'' refers metonymically to the driver.

However, I suspect that such utterances may often be better handled as cases of personification metaphor. Although metaphor tends to be viewed as a more complex phenomenon than metonymy, in my own approach the representations that would be used in a metaphorical approach to the examples would be simpler than the representations an understander one would need to use if he/she/it cashed out the metonymic references into representations of the real referents. For instance, in the committee example one might need some way of referring to those members who form the majority, without having any idea of which subset constitutes the majority (or even what counts as a sufficient majority in the workings of the particular committee). Also, as is often pointed out, if part of the virtue of metonymy is sometimes to avoid the issue of exactly what things are being referred to, then the understander might need to list alternative real referents, or to have very vague yet technically complex representations like ``some people relevantly associated with the committee.'' By contrast, in my metaphorical approach, the internal representation would just cast the committee as an entity that can have mental states --- i.e., in effect the metaphor is preserved in the internal representations themselves.

Such gains in representational economy have consequent gains in economy of reasoning. Indeed, there may be qualitative gains as well: the metaphorical approach makes possible certain commonplace uncertain inferences that would be considerably harder to justify on the metonymic approach.

Distinguishing Between Metonymy and Metaphor

In seeking to construct a definition of metaphor that relies on a firm notion neither of what a ``domain'' is nor of what ``literal'' language about a domain is, I have encountered considerable difficulty in distinguishing adequately between metonymy and metaphor. In particular, without a secure notion of domain it becomes more difficult to say what mappings should count as metaphorical, and in particular which should count as metaphorical as opposed to metonymic. An appeal to the notion of similarity is tricky because in some cases it may be difficult to justify saying that a given mapping is one of similarity without first having judged it to be metaphorical; and metaphors can create similarity rather than simply exploit already known similarity.

In addition, even if one does divide knowledge up into domains, the two (or more) domains involved in a metaphorical utterance can overlap in important ways (cf. Kittay 1989), whereas the literature often gives the impression that the domains do not overlap. This further complicates the notion of a metaphorical mapping.

Minds as World Definers

An additional issue for discussion --- largely unrelated to metonymy but strongly related to considerations about diagrams, models and so on --- is a commonly-used metaphorical view that I call MIND AS WORLD-DEFINER. In this view a person's mind is viewed as being or containing a definition of a world. This is illustrated in sentences like


    In Stephen's mind, the man had knocked the woman over deliberately.

This, I claim, is much like saying ``In the story, the man had knocked the woman over deliberately'' or ``In the picture, the man was knocking the woman over deliberately.'' If we take stories and pictures to define (mini-)worlds, then is reasonable to suggest that phrases like ``in Stephen's mind'' (often) also cast the mind as a world-definer.

If this stance is correct, then considerations concerning how maps, diagrams, pictures, and models are used and talked about can illuminate the question of how to deal with an important type of mental metaphor, and vice versa.

It would be interesting to investigate to what extent the types of metonymy that can be used in talk about models, etc., are related to any that can be used in conjunction with MIND AS WORLD-DEFINER.

Acknowledgements

The work is currently being supported by a grant from the Engineering and Physical Sciences Research council of the U.K. and has previously been supported by grants from the National Science Foundation of the U.S.A.

I am indebted to Mark Lee, Marina Barnden, David Brooks, Fiona Ferguson, and Miriam Howe for valuable help in the research.

References

Barnden, J.A. (1998). An AI system for metaphorical reasoning about mental states in discourse. In Koenig, J-P. (Ed.), Discourse and Cognition: Bridging the Gap. Stanford, CA: CSLI.
Barnden, J.A. (1998). Combining uncertain belief reasoning and uncertain metaphor-based reasoning. In Procs. Twentieth Annual Meeting of the Cognitive Science Society, pp.114--119. Mahwah, N.J.: Lawrence Erlbaum Associates.
Barnden, J.A. (1999). An implemented system for metaphor-based reasoning, with special application to reasoning about agents. In C.L. Nehaniv (Ed.), Computation for Metaphors, Analogy, and Agents, Lecture Notes in Artificial Intelligence, Vol. 1562, pp.143--153. Springer.
Barnden, J.A. & Lee, M.G. (1999). An implemented context system that combines belief reasoning, metaphor-based reasoning and uncertainty handling. In P. Bouquet, P. Brezillon & L. Serafini (Eds), Second International and Interdisciplinary Conference on Modeling and Using Context (CONTEXT'99, Trento, Sep 9--11, 1999), Lecture Notes in Artificial Intelligence, 1688, pp.28--41, Springer.
Fass, D. (1997). Processing metaphor and metonymy. Greenwich, Connecticut: Ablex.
Hobbs, J.R. (1990). Literature and cognition. CSLI Lecture Notes, No. 21, Center for the Study of Language and Information, Stanford University.
Kittay, E.F. (1989). Metaphor: its cognitive force and linguistic structure. (Paperback ed.) Oxford, U.K.: Clarendon Press.
Lee, M.G. & Barnden, J.A. (2000). Reasoning about mixed metaphors with an implemented AI system. Technical Report CSRP-00-11, School of Computer Science, The University of Birmingham, U.K. To appear in 2001 in the journal Metaphor and Symbol, 16 (1 & 2).

to SFB-Homepage

created by: Anke Weinberger (2000-08-30).
maintained: Anke Weinberger (2000-10-04).