Does conventionality affect metaphor processing? Comparing fMRI results to computational models
Aubrie Amstutz (email@example.com)
Evi Hendrikx (firstname.lastname@example.org)
Xiaoyu Tong (email@example.com)
Feel free to contact us for more details!
University of Amsterdam Brain and Cognition Master’s Summer School
Metaphor is ubiquitous in language use, and fundamental in the human conceptual system.
A metaphor is a conceptual cross-domain mapping. The mapping eventually results in emergent meaning that does not reside in either the source or the target domain. One of the most influential theories of metaphor processing, the Career of Metaphor theory, contends that the meaning of a metaphor can emerge from two main types of processes: comparison or categorization (Bowdle and Gentner, 2005). Comparison involves establishing mappings between the target and source domains in order to generate emergent meaning. Categorization entails finding a superordinate category to which both the target and the source belong, with the source being a good representative of the category. For example, in the metaphor “My job is a jail,” one could either compare the similar properties of job and jail (e.g. unpleasantness, long hours spent in one place), or one could consider a category to which both belong (e.g. places one feels trapped). Both of these methods of processing a metaphor allow for a new perspective on the first entity (often called the “topic,” or the target domain) due to the additional information provided by the second entity (often called the “vehicle”, or the source domain).
The question that then arises is under what conditions a metaphor is processed through categorization or comparison. In the Career of Metaphor theory it is proposed that conventional metaphors are processed using the categorical method, while novel metaphors are processed using the comparison method. The conventionality explanation (and its alignment with the categorical or comparison method) has been challenged since the theory was proposed, and the debate has not been resolved. This project aims to compare the processing of conventional and novel metaphors using computational modelling and fMRI brain scans of human subjects.
Utsumi (2011) tests three hypotheses concerning different moderators of metaphor processing (conventionality, aptness, and interpretive diversity) using a computational model that represents categorization and comparison processes. The predictions of the model were compared to participant ratings of the three properties. They showed correlations between subjective measures of conventionality and the working of a computational model on these same metaphors. However, since categorization and comparison processes are not yet verified in the brain, we believe conclusions about similar processes taking place during human processing of metaphors are premature. This is what we want to improve on.
Using participant ratings of the conventionality of a metaphor is problematic as it would assume that ordinary people are able to distinguish conventional and novel metaphors, which even metaphor researchers are not able to accomplish without an agreed upon procedure. Therefore, the conventionality judgements which were used in this study are of questionable reliability. We will thus use a more objective procedure for conventionality judgement in this study.
In the current study we propose to compare brain data (fMRI scans) to computational models that employ algorithms which simulate the processing methods of categorization and comparison to answer the research question: Does the conventionality of a metaphor determine whether a metaphor is processed through categorisation or comparison?
If the Career of Metaphor theory is correct, a conventional metaphor would be processed through categorization, while a novel metaphor would be processed through comparison. Our results would then show that the model that employs a categorization algorithm has a higher correlation with the brain data that were collected during the presentation of conventional metaphors than the model that employs a comparison algorithm or a control model. On the other hand, the brain data collected during metaphors would show a higher correlation with the model that includes the comparison algorithm than with the other models.
Bowdle, B. F., & Gentner, D. (2005). The career of metaphor. Psychological Review, 112(1), 193–216.
Utsumi, A. (2011), Computational Exploration of Metaphor Comprehension Processes Using a Semantic Space Model. Cognitive Science, 35: 251-296. doi:10.1111/j.1551-6709.2010.01144.x
Steen, G. J., Dorst, A. G., Herrmann, J. B., Kaal, A. A., Krennmayr, T., & Pasma, T. (2010). A method for linguistic metaphor identification: From MIP to MIPVU. Amsterdam, the Netherlands: John Benjamins.
Chiang, J. N., Peng, Y., Lu, H., Holyoak, K. J., & Monti, M. M. (2019). Neural and computational mechanisms of analogical reasoning. bioRxiv, 596726.
Brain icon made by [Freepik] from www.flaticon.com
Ear icon mayd by [Roundicons] from www.flaticon.com