Semantic role alignment in metaphor : a frame semantic approach to metaphoric meaning

Metaphor occurs when a word or phrase is used in a way that conflicts with its usual (literal) meaning, so that part of its meaning is applied to a different semantic domain. For example, time is construed as money in “This gadget will save you hours” (Lakoff and Johnson 1980). There is a link between the concepts time and money that underlies many expressions in English; this is therefore considered a conceptual metaphor. Conceptual Metaphor Theory (CMT) has dominated metaphor research since Lakoff and Johnson (1980), but researchers (e.g. Croft 2009, Sullivan 2013) are turning to other cognitive linguistic theories such as Frame Semantics (Fillmore 1982) and Cognitive Grammar (Langacker 1987) to rectify the problems inherent in that approach. CMT does not provide tools for systematically defining metaphoric concepts and their components, which prevents the analysis of metaphor’s internal meaning. It views metaphor as a superimposition of meaning from one domain (e.g. money above) onto another (e.g. time). Corpus data has improved metaphor research methods, but sounder methodology is needed to choose which metaphors to study. This dissertation takes a novel approach to metaphor in that the data are taken from a semantically annotated corpus where their semantic domains are already assigned. The main dataset is comprised of a naturally occurring group of related metaphors that construe awareness as perception. Using the notion of frame from Frame Semantics (Fillmore 1982) as implemented in the lexicographic database FrameNet (Atkins et al. 2003) to define semantic domains and their internal components (frame elements; similar to semantic roles), this dissertation analyzes metaphors from the frame-semantically annotated database of German, SALSA (Burchardt et al. 2006, 2009), to investigate how meaning elements (semantic roles) from the metaphor’s two semantic domains align. I show that semantic roles align consistently, although not every semantic role has a counterpart in the other domain. I argue that the use of semantic and syntactic information that is associated with one domain but not the other allows emergent meaning to be created in metaphor. The analysis supports the view of metaphor as a blended space, independent of either semantic domain, as described by Fauconnier and Turner (2002).