Text

theses

For over 50 years, linguists have raised questions about the nature of switch-reference (SR): is it a syntactic or a semantic phenomenon? One of my main goals in this dissertation is to argue that SR is one of grammar’s multiple ways to express anaphora, and that as such, it cannot be characterized as simply syntactic or semantic: it involves coordinated work from all modules of grammar. My main source of primary data is the Yawanawa language, which along with its Panoan relatives, has an especially rich paradigm of SR: not only does it have the same-subject (SS) and different-subject (DS) markers that are found in a number of SR languages around the world, but also it includes the rare object=subject (OS) marker, which has not been documented outside the Panoan family (Valenzuela 2003; Fleck 2003; Zariquiey-Biondi 2011; Baker and Camargo Souza 2019, 2020; Clem 2019; Neely 2019). I argue that the existence of the OS morpheme supports the view that SR has a strong syntactic component (Finer 1984 and much subsequent work): it needs to make reference to the grammatical functions of subject and object. However, syntax is not all there is to it. SR is not construction- specific and as such, it gives rise to different types of coconstrual. The nature of the coconstrual is up to semantics to define, according to the syntactic structure that is
shipped to interpretation, and the types of nominal expressions therein. The possible coconstruals we find are not SR-exclusive: they are the ones that are made available by Universal Grammar and therefore familiar to us.

I argue that the syntactic component of SR is Agree-based (Chomsky 2000, 2001), which at first may seem somewhat paradoxical because SR morphemes do not expone the phi-features of DPs it coconstrues. I solve this apparent paradox by proposing that when Agree-link obtains in a certain configuration but Agree-copy does not, this leads the semantic module to interpret the links as a coconstrual relation. As such, we get Agree without agreement. More specifically, the configuration is one in which a single syntactic terminal is linked to two DPs. When it comes to interpretation, I argue that whenever c-command obtains between the DPs being coconstrued, we get bound variable anaphora; when it does not, we get dynamic binding or coreference, depending on the types of nominal expressions involved. I find further support for my argument that SR expresses familiar types of referential dependencies by exploring patterns of anaphora to sets with quantified and plural referents (c.f. Thomas 2019). I show that what is special about SR languages is that they ‘choose’ to morpho-syntactically express certain types of cross-clausal anaphora, not that they express patterns of anaphora that are unique or unattested crosslinguistically.

ETD_11269

Ph.D.

Includes bibliographical references

External ETD doctoral

doi:10.7282/t3-k2nr-6m84

The author owns the copyright to this work.