An Updated Typology of Causative Constructions: Form-Function Mappings in Hupa (California Athabaskan), Chungli Ao (Tibeto-Burman) and Beyond
Taking up analytical issues raised primarily in Dixon (2000) and Dixon & Aikhenvald (2000), this dissertation combines descriptive work with a medium-sized (50-language) typological study.
Chapter 1 situates the dissertation against a concise survey of typological-functional work on causative constructions from the last few decades, and outlines the major research questions.
Chapter 2 presents a case study of causative encoding in Hupa (California Athabaskan). I describe the morphosyntax and semantics of the Hupa syntactic causative construction, and analyze its
distribution with respect to Dixon's (2000) proposals for a typology of causative constructions. I demonstrate that causee control (Dixon's parameter 3) over the caused microevent is a significant
semantic factor in licensing this construction. I show that Næss' (2007) model of transitivity as a set of semantic prototypes, as applied to causative event participants, nicely accounts for other
aspects of the distribution--particularly, why some events can be encoded in a causative construction, while other, quite similar, events must be encoded in a result clause or purpose
clause. I end with an examination of how Hupa encodes causal chains within single lexical items, bringing together a small corpus study and a case study of Hupa encoding of cutting and breaking
Chapter 3 presents a case study of causative encoding in Chungli Ao (Tibeto-Burman). I offer a description of Chungli Ao's morphological and periphrastic causative constructions, and analyze
their distribution with respect to Dixon's (2000) model, showing that directness of causer action (Dixon's parameter 6) can determine the choice between a lexical causative (where one is available), and a non-lexical causative construction. I present an in-depth case study of Chungli Ao encoding of cutting and breaking events, including a discussion of the semantics and distribution of several lexical suffixes and the elements of the causal chain they encode. Also
discussed are syntactic alternations systematically available to CUT-class verbs and BREAK-class verbs, respectively.
Chapter 4 moves away from language-specific case studies, and turns to crosslinguistic research in order to address some broader research questions about universal tendencies in the encoding of
causative relationships. I bring a new, larger body of empirical evidence to bear specifically on 2 Dixon's (2000) account of the formal and semantic factors influencing (or forcing) a speaker's
choice of one causative construction over another. Dixon's (2000) claims about prototypical patternings of compact vs. less compact constructions are not well-supported: in order for the claims to be well-supported, the values of Dixon's nine parameters would have to be correlated in individual languages at a statistically significant frequency. Parameter 5 (partially versus completely affected causee) was not found to be crucially encoded at all in a sample of 114
constructions from 50 languages. Parameters 7 (causer intentionality), 8 (naturalness of causer action) and 9 (causer accompanying or not accompanying causee) nearly always occur in isolation. The one instance of a solid correlation among parameter values (with a sample of more than 10 relevant pairing events) was that between parameters 3 (causee control) and 4 (causee
willingness). These two parameters were correlated: they were encoded together at particular values (both low or both high), or were both systematically absent, at a statistically significant
frequency. Since, in general, Dixon's proposed semantic parameters tend not to co-occur, the study shows that it is not feasible to model the distribution of causative constructions around
notions of semantic prototypicality. A much larger sample may show Dixon to be correct in his proposal regarding correlations between values of semantic parameters, but the current findings
suggest that his nine parameters are not likely to pattern together in subgroups. I conclude the chapter by presenting a series of constructions, encountered in the course of compiling the
typological sample, that obligatorily encode other semantic information along with causativity. I note potential patterns and propose questions for research to build on my typological study.
Chapter 5 summarizes the contributions of the dissertation, and recapitulates various proposals for further study.