The Language of Stories: A Conceptual Integration Approach

This paper outlines a cognitive linguistics framework for an analysis of narratives. It uses the theory of conceptual integration to propose a way of accounting for the emergence of narrative meaning and the correlations between form and interpretation. Processing the language of a narrative text, be it a novel, a film, or a play, is a crucial component of narrative comprehension. The research reported here shows how processes driven by general linguistic and conceptual patterns of meaning construction prompt the reader's or viewer's response to the narrative artifact. I rely on the basic claims of the theories of Mental Spaces and Conceptual Integration (fully described recently in Fauconnier and Turner 2002). The theory argues that linguistic and visual forms serve as prompts for evocation and setting up of semantic frames and conceptual 'packets' called mental spaces. These conceptual structures are then subjected to processes in which new meanings emerge as contextually relevant and discourse-based interpretations. The processes involve projections from one activated conceptual 'packet' to another, selection of relevant typology, and blending of mental space and frame topologies in novel ways. The approach is particularly useful in the study of narratives. Much of the research on stories focuses on lower level linguistic choices, such as narration, speech and thought representation, or focalization, or, alternately, on general cognitive processes (see e.g. Herman 2003). I propose a conceptual integration mechanism that explains the participation of the lowest level linguistic forms in the overall construction of meaning. These processes result in the emergence of a coherent, sequential story. I also define standard narratological concepts such as 'narrator', 'story' or 'focalization' in terms of mental spaces and blending (see Dancygier 2011). The basic fact that requires an explanation is how the specific formal choices make it possible for the reader (or Copyright © 2010, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. viewer) to come up with an overall understanding of the narrative and also respond emotionally. To address this, I propose the concept of a 'narrative space' – a story-based mental space structured with local topology (time, place, participants). These lowest-level narrative spaces are blended to yield higher-level narrative spaces, such as subplots or temporally marked parts of the narrative (e.g. extended flashbacks). These subsequent levels of blending of narrative spaces eventually yield the emergent space, traditionally described as 'the story'. The final product of narrative comprehension is thus a mental construct, a mega-blend, which emerges through multiple levels of selective projection and construction of coherence. All the spaces are constructed from the viewpoint of what I refer to as the 'story-viewpoint' (SV) space, which determines the temporal and spatial frame of the narrative (such that the story is told from the present or past perspective, from within the story-space or outside of it). The viewpoint structure of the SV-space also provides a set-up wherein the narrative is conceptualized against a communicative deictic ground. The set-up creates the illusion of a human 'teller', or 'narrator', but this surface effect is in each case the result of a specific SV-space viewpoint configuration. In other words, the text constructs a narrative space in which there is a voice mimicking a human communicator. This set-up has broad interpretive consequences for the processing of the story as a whole. I argue that the primary blending mechanism responsible for subsequent integration of narrative spaces into the emergent story is viewpoint compression (Dancygier 2005). Compression is the central mechanism of narrative blending, which makes it possible for elements of various spaces that are conceptually distant to become less distant or fused in the blend. For example, characters may appear under different identities in different narrative spaces (as in the case of Dr. Jekyll and Mr. Hyde), but need to be blended in the overall interpretation. Such a re-alignment of identity requires the use of textual clues in the storyconstruction processes the reader goes through, but consists in compressions of identity across narrative spaces. Blending theory talks about compressions along dimensions such as identity, change, or causation, but narratives also involve a specific kind of compression 14 Computational Models of Narrative: Papers from the AAAI Fall Symposium (FS-10-04)