Open Access Publications from the University of California

## When phi-Agreement Targets Topics: The view from San Martín Peras Mixtec

• Author(s): Ostrove, Jason
• et al.
Abstract

This dissertation presents what is, to my knowledge, the first in-depth theoretical syntactic investigation of a Mixtec language: San Martín Peras Mixtec. The syntax of this language is explored through the lens of a process called pronoun doubling, which duplicates the phi-features of a nominal with a preverbal morpheme that is morphophonologically identical to a regular pronoun.

The narrow question of this dissertation is what the synchronic analysis of pronoun doubling is. Ultimately, I argue that, despite initial appearances, the pronoun'' in pronoun doubling constructions is not a true pronominal, but rather the morphophonological exponent of a phi-agreement process that targets topics.

But as this process replicates $\varphi$-features, I consider two possibilities. First, I investigate if pronoun doubling could be a form of clitic doubling. I argue that clitic doubling is best derived by phrasal movement of the doubled nominal, followed by reduction of the higher copy to a clitic. Therefore, in order to diagnose clitic doubling within this Move-and-Reduce'' framework, we must look for evidence of movement in pronoun doubling constructions.

I first consider A-movement. I describe a novel A-movement process which I term quantifier fronting.'' I then show that none of the A-movement properties of quantifier fronting can be found in pronoun doubling. I also show that A-movement in SMP Mixtec is typologically unusual in being much less restricted than A'-movement.

A'-movement in SMP Mixtec shows several interesting properties. First, it demonstrates an A'-extraction restriction reminiscent of syntactic ergativity: external arguments may not undergo A'-movement. I investigate this restriction through the properties of two A'-constructions. I first investigate restrictive relative clauses, which require a resumption strategy in order to avoid the ban against extraction of external arguments. I then investigate content wh-questions, which avoid this ban by employing a movement construction that demonstrably lacks the properties of both A-movement and A'-movement. Crucially, no evidence for this extraction restriction is replicated in pronoun doubling constructions.

As clitic doubling within this Move-and-Reduce framework does not seem like a viable option, I consider the possibility that pronoun doubling is a form of phi-agreement. Despite appearances, this turns out to be the best fit for several reasons. First, pronoun doubling obeys the PIC yet allows Closest Conjunct Agreement. Second, pronoun doubling yield novel variable-binding possibilities. That said, pronoun doubling is not vanilla phi-agreement for two reasons. First, both subjects and objects are equally accessible, depending on which is interpreted as a topic.

To account for this system, I develop a theory in which phi-agreement can target topics. This functions similarly to a traditional minimalist Case theory, in which nouns must stand in an Agree relation with some functional head in order to receive Case. I adapt this general approach to topic agreement by proposing that nominals which are to be interpreted as topics must stand in an Agree relation with a head Top, and Top also acts as a phi-probe.

The dissertation is organized as follows. Chapter 1 provides the necessary background on this language, including a brief syntactic and phonological overview. Chapter 1 also provides a detailed discussion of the prosodic conditions which restrict the appearance of topic agreement. Chapter 2 provides an extensive literature review of both phi-agreement and clitic doubling, including an exhaustive evaluation of the diagnostics which have been put forward to distinguish the two. In this chapter I also argue for why clitic doubling is best understood within a Move-and-Reduce framework. This approach to clitic doubling allows for intuitive argumentation for either clitic doubling or phi-agreement: clitic doubling constructions show the characteristics of movement chains, while phi-agreement does not.

With this background, Chapter 3 begins the investigation of movement constructions in San Martín Peras Mixtec by investigating A-movement. It begins with a detailed discussion of the syntax of embedded clauses. This is because failure to extract from finite embedded clauses is a common feature of A-movement cross-linguistically. In this language, though, it is difficult to tell which embedded clauses are finite because there is no non-finite verbal morphology and all embedded clauses have an overt subject. I argue that, despite appearances, this language distinguishes between Obligatory Control and simple pronominal coreference, despite this language's equivalent of PRO always being overt. This relation is mediated through a CP layer, which faithfully tracks when Obligatory Control occurs.

Once these aspects of the clause structure have been established, I demonstrate the A-movement properties of quantifier fronting, such as failure to reconstruct for scope or binding, and the ability to expand the variable and anaphora binding possibilities within a clause. I also show that A-movement in this language is much less restricted than A'-movement, despite being clause bounded. While A'-movement forbids extraction of possessors, objects of prepositions, and, as shown in Chapter 4, external arguments, A-movement shows no such restriction. Additionally, A'-movement, particular content wh-questions, require pied-piping with inversion, while A-movement is incompatible with it. Importantly for our purposes, pronoun doubling fails to show any of these A-chain properties: it fails to affect variable or anaphora binding, and it may not target objects of prepositions or possessors.

As pronoun doubling fails to show the properties of an A-chain, Chapter 4 considers the possibility that the movement chain involves is an A'-chain. This requires an extensive investigation of A'-movement in this language. I demonstrate that San Martín Peras Mixtec shows a strict A'-extraction restriction, in addition to the restictions shown in Chapter 3: external arguments may not undergo A'-extraction. This restriction, while similar in many ways, is fundamentally different from syntactic ergativity for two reasons. First, San Martín Peras Mixtec is not morphologically ergative. Second, intransitive predicates show varying behavior, with unergative subjects patterns with transitive subjects while unaccusative subjects pattern with transitive objects.

After this extraction restiction is established, I demonstrate its properties in more detail. First I investigate restrictive relative clauses, which require resumption of external arguments. I show that this resumption strategy fails to show any properties of movement. This allows the language to straightforwardly avoid the A'-extraction restriction. Second, I investigate the syntax of content wh-questions. These also show the effects of the A'-extraction restriction, albeit the repair is different. Content wh-questions of external arguments do involve movement, but I demonstrate that this movement process fails to show the hallmark characteristics of either A-movement or A'-movement. This also allows the language to satisfy the restriction against A'-extraction of external arguments. I conclude by demonstrating that this extraction restriction that is so crucial to understanding A'-constructions in this language is not replicable in pronoun doubling constructions because external arguments may be freely targeted.

As pronoun doubling fails to show the diagnostic movement chain properties of clitic doubling, Chapter 5 considers the possibility that this process is a phi-agreement. After cataloging the phi-agreement properties of pronoun doubling, such as observance of the PIC, Closest Conjunct Agreement, and the failure to affect variable and anaphora binding, I develop a theory of topic agreement. First, I demonstrate that topic agreement is not an especially rare phenomenon cross-linguistically, having been reported in a range of languages from Northern Khanty (Uralic, Russia), Oromo (Cushitic, Ethiopia), and many others. While the semantic investigation is cursory, I demonstrate that topichood is the best notion to determine which nominals will be targeted in pronoun doubling constructions. Finally, I present the formal analysis of topic agreement, which I described above.