Skip to main content
eScholarship
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Constructional and Conceptual Composition

Abstract

Goldberg's (1995) recognition that, in addition to various word-level constructions, sentences also instantiate meaningful argument structure constructions enables a non-polysemy-based analysis of various verb `alternations' (Levin 1993). In such an analysis, meaning variations associated with the use of the same verb in different argument realization patterns are analyzed as resulting from composition of the same verb meaning with different meaningful argument structure constructions. This compositional analysis raises some important semantic questions: Which specific elements of verb and argument structure construction meanings motivate their composition? And what is the precise nature of the meaning that results from this composition? I argue that to answer these questions, it is essential that we look more closely at the underlying conceptual systems utilized by these constructions.

In this dissertation, I examine a set of simple single clause sentence examples which describe a variety of basic experiences involving motion, change of location, action, force, causation, and affectedness. One key assumption that the Neural Theory of Language (NTL) makes about language understanding concerns simulation semantics, the idea that the neural circuitry we use to understand descriptions of experiences is closely similar to that which is activated by the experiences themselves (Feldman 2006). Accordingly, I use linguistic, neuroscientific, and other forms of evidence to define a lattice of interconnected schemas that represent schematic structures and interrelations associated with the types of experiences described in these examples, and use these schemas in the meaning representations of the verb and argument structure constructions instantiated in these examples. In addition, the grammar described in this dissertation includes various word-level, phrasal and clause-level constructions, along with schemas to represent their meanings, all of which are represented using the Embodied Construction Grammar (ECG) formalism (Bergen and Chang 2005, Feldman et al. 2010). Significantly, use of this formalism has made it possible to utilize many of the schemas and constructions in this grammar in a computational implementation of a compositional constructional analysis of sentence meaning. This implementation, called the Constructional Analyzer (Bryant 2008) determines the best-fitting interpretation of an utterance in context based on a consideration of syntactic, semantic, and contextually-specified constraints. The composed meanings of the constructions instantiated in this best fit interpretation serve as a semantic specification for the simulation of the event described by the utterance (simulation is not carried out as part of this implementation). Through simulation, the relatively schematic meanings specified by the constructions instantiated in a given utterance give rise to a much richer and fuller understanding of that utterance, via activation of additional conceptual structure related to an understander's experiences, beliefs, etc.

The schemas and constructions that comprise the grammar discussed in this dissertation have many complex interrelations. One important benefit of using the ECG formalism is that computational implementations facilitate the development of an internally consistent, wide-coverage, complex grammar. In this respect, it is similar to other unification grammars, such as HPSG (Pollard and Sag 1994) and LFG (Dalrymple 2001). However, unlike these other grammars, ECG has a deep commitment to embodied semantics, and is consistent with the NTL theoretical framework. Moreover, as the schemas and constructions presented in this dissertation illustrate, this formalism makes possible the integrated expression of several important cognitive linguistic insights, including those related to recurrent schematic conceptual structure such as: (1) image schemas (Johnson 1987, Lakoff 1987) and frames (Fillmore 1982); (2) basic patterns of cognitive organization such as prototypes (Rosch 1975, 1978) and radial categories of A-S constructions (Lakoff 1987, Goldberg 1995), and; (3) different attention-related phenomena such as perspective and profiling (Talmy 1996; Langacker 1987, 1991).

Together, the different elements described above enable the development of a grammar that supports compositional constructional analyses of a range of different sentence examples that instantiate various combinations of different verb and argument structure constructions. Significantly, these analyses capture the sometimes subtle similarities and differences in the event conceptualizations described by these examples. Moreover, this compositional account of sentence meaning also provides a compositional account of semantic roles that enable us to recognize a variety of cross-cutting generalizations that can be made about these roles.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View