Extemporaneous Blending: Conceptual Integration in Humorous Discourse from Talk Radio

Abstract

An Off the Leash cartoon by W. B. Park depicts a dozen or so pigs feeding at a trough. One pig, however, has his head raised, as if addressing the approaching farmer. The pig's words are apparently expressed in the cartoon's caption, Garcon! The cartoon thus compares the farmer in the cartoon to a waiter in a French restaurant, and the viewer is left to speculate about the nature of the correspondence between expensive French food and the contents of the feeding trough. Douglas Hofstadter and Liane Gabora, pointing to the analogical nature of this joke, pose the term frame blend for a frame whose elements and relations are constructed from a combination of two frames that share some abstract structure. Gilles Fauconnier and Mark Turner have shown how frame blends occur in a wide variety of cognitive phenomena, and they have developed an elaborate theory of conceptual integration, or blending, to explain the representation of composite descriptions (Conceptual Integration, Projection, Way). Previous work in this area suggests that conceptual blending plays an important role in cases of verbal humor. But whereas this earlier work has addressed conceptual integration needed to comprehend carefully crafted humorous narratives, the present study addresses blends that people use in the slightly less scripted world of talk radio. Below I provide a brief introduction to conceptual integration theory, describe its application to humorous interaction between two hosts on a radio call-in show, and consider how people exploit the creative process of meaning construction in conversational interactions. 1. Conceptual Integration Theory Among the basic concepts in conceptual integration theory are mental spaces, frames, or cultural models, and mappings. Mental spaces can be thought of as buffers in working memory that represent relevant information about a particular domain (Fauconnier, Mental). A mental space contains a partial representation of the entities and relations of a particular scenario as construed by a speaker. Spaces are structured by elements that represent each of the discourse entities and simple frames to represent the relationships that exist between them. Frames are hierarchically structured attribute/value pairs that can either be integrated with perceptual information or be used to activate generic knowledge about people and objects assumed by default. Socially shared frames are called cultural models. Finally, mappings are abstract correspondences between elements and relations in different spaces. When speakers produce language, listeners use that linguistic input along with background and contextual knowledge to set up simple cognitive models in mental spaces (Coulson, Semantic Leaps). Similarly, when people look at cartoons, or, indeed, the events of the world, they partition the input into different mental spaces, each structured by cognitive models from a relevant domain. For example, in the barnyard cartoon described above, the artist is evoking an analogy between aspects of the domain of human restaurants and the domain of barnyards. Since a mental space is used to represent certain aspects of conceptual structure from a particular domain that is relevant to the ongoing discourse context, to understand the barnyard cartoon, we set up one mental space to represent relevant aspects of the barnyard domain and another to represent those of the restaurant domain. Although our knowledge of restaurants and barnyards is fairly extensive (though, admittedly, most urban dwellers know much more about the former), the conceptual structure activated in a mental space is but a small subset of the totality of our knowledge of these domains. In the barnyard cartoon, for example, the barnyard space is structured by a few contextually relevant elements, including the pig, farmer, food, and trough that are depicted in the cartoon, and by a simple frame that represents the relationship between them. …

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 88 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.