Syntax & Semantics Circle

University of California, Santa Cruz

Spring 2018

May 31

Jed Pizarro-Guevara (UC Santa Cruz): “A tale of two grammars in Tagalog: Variation in how multiple specifiers are treated”

S-Circle and MRG joint meeting

Much ink has been spilled on the extraction restriction in Philippine-type languages. In Tagalog, it has been claimed that only the argument that is cross-referenced by voice morphology is eligible to undergo further A’-movement (Aldridge, 2012; Rackowski & Richards, 2005; Sabbagh, 2005; Schachter & Otanes, 1983). In other words, when the verb exhibits AGENT VOICE (AV), only agent-extraction is licit; when the verb exhibits PATIENT VOICE (PV), only patient-extraction is licit.

However, Ceña and Nolasco (2011), Tanaka (2016), and Hsieh (2018) have observed that agent-extraction is licit under PV—at least for some people, for some A’-dependencies. In the present study, I first provide naturally occurring data and experimental evidence that corroborate these observations. I also show that the permissibility of agent-extraction under PV is a general property of A’-dependencies but its magnitude varies by the type of A’-dependency. Taking inspiration from Han, Lidz, and Musolino (2007), I propose that these “conflicting” generalizations can be framed in terms of grammar competition: different grammars view configurations involving multiple specifiers differently. Finally, taking inspiration from Dillon, Staub, Levy, and Clifton Jr (2017), I provide computational evidence consistent with a two-grammar hypothesis.

May 25

Vera Gribanova (Stanford University): “On the interaction between syntax and postsyntax in Uzbek (non-)verbal predicate formation”

S-Circle and MRG joint meeting

In Distributed Morphology (DM), morphophonological unification of independent syntactic units is in- tegrated into the Minimalist Y-model via postsyntactic operations, which manipulate the output of syntactic structure-building to amalgamate morphosyntactic feature bundles, associate them with phonological expo- nents, and linearize them. A large body of work aims to justify these operations as independently necessary, especially since many of them perform some type of merger (see Harley 2013). But while we often find evidence for one of these operations in a given language or case study, studies of the interaction of these operations within a single system are far more rare.

This talk draws on fieldwork-based evidence involving Uzbek verbal and non-verbal predicates. I argue that despite some surface similarities, the two predication types employ radically distinct word formation strategies, including various combinations of head movement, Lowering (Embick and Noyer 2001), merger under adjacency (Bobaljik 1994), and a phonological support mechanism akin to English do-support. I pro- pose that the observed differences between verbal and non-verbal predication can be derived from a general organizing principle of Uzbek, namely the availability (indeed, necessity) of syntactic head movement as a word formation strategy for verbal, but not non-verbal, predicates. Evidence in favor of this claim is drawn from novel paradigms involving verb-stranding ellipsis in Uzbek.

The presentation is based on work in progress, and forms the foundation for a planned micro-comparative study of the syntax and postsyntax of predicate formation in the Central Asian Turkic languages.

May 3

Chris Hammerly (UMass Amherst): “A verb-raising analysis of the Ojibwe VOS/VSO alternation”

S-Circle and WLMA joint meeting

A wide variety of typologically diverse languages show an alternation between VOS and VSO word orders. In this talk, I provide an analysis of a previously undescribed alternation from Southwestern Ojibwe (Algonquian). I show that in so-called direct agreement environments, where the object is obviative and the subject proximate, the word order can alternate between VOS and VSO, while in inverse agreement environments, where the subject is obviative and the object proximate, only VOS is allowed. Using facts from the possible word orders of ditransitive verbs in these environments, and scope between negation and indefinite objects, I argue that verb initiality must be accounted for by verb raising, and that the alternation between VOS and VSO in direct environments is derived by optional scrambling of obviative arguments to the middle field. The constellation of facts presented rules out an analysis of the VOS word order based on predicate fronting or base-generation with rightward specifiers, which have previously been proposed to account for surface-similar alternations in Austronesian languages. The analysis therefore supports a view where VOS/VSO alternations across languages of the world might arise from one of a variety of distinct sources.

April 26

Kenny Baclawski (UC Berkeley): “The discourse subordination effect and the syntax-discourse interface”

Certain anaphoric phenomena are known to be licensed only when discourse subordination, a class of rhetorical relation, relates the current sentence to the sentence containing the antecedent. This discourse subordination effect has been observed for clitic right-dislocation in Catalan (López 2009), topicalization, and wh-fronting in Eastern Cham (Baclawski Jr. 2015).

In this talk, experimental evidence demonstrates the discourse subordination effect for English D-linked wh-phrases (e.g. which book). Participants in a study on Amazon Mechanical Turk rated questions with D-linked wh-phrases significantly worse in the absence of discourse subordination, though the effect disappears with non-D-linked wh-phrases (e.g. what book). This pattern is not predicted by other factors previously associated with D-linking like salience or previous mention (e.g. Pesetsky 1987; Comorovski 1996; Grohmann 1998).

To account for these findings, a new representation of discourse is proposed, "Minimal SDRT", deriving from Segmented Discourse Representation Theory, which adds hierarchical rhetorical relations to DRT (Asher 1993; Asher & Lascarides 2003). In Minimal SDRT, prior discourse is modeled by a universe of discourse referents and a discourse tree, a hierarchy of rhetorical relations. Through Minimal SDRT, once a sentence is attached to the discourse tree, the universe is updated to indicate which antecedents are accessible via discourse subordination. It is argued that this stage of discourse update occurs prior to the numeration. The discourse subordination effect, then, is analyzed with syntactic features that reference information about discourse subordination in the updated universe.

April 20

CLS Practice Talks

Tom Roberts (UC Santa Cruz): “Pragmatic Licensing of Negative Polar Questions in Estonian”

Since Ladd (1981), a significant body of work has aimed to characterize the difference between positive polar questions (PPQs) and corresponding negative ones (NPQs), as in (1) below:

(1) a. Is Mabel home?

b. Is Mabel not home?

Both (1a) and (1b) request information to settle the issue of whether or not Mabel is home; indeed, in a Hamblin/Karttunen-esque semantics for questions they are denotationally equivalent. However they differ in their use conditions. (1a) cannot felicitously be used when the speaker expects Mabel not to be home, whereas (1b) can (Büring & Gunlogson 2000). This poses a puzzle: if (1a) and (1b) are semantically equivalent, why are they felicitous in different contexts? And to what extent is this difference derived from the presence or absence of negation?

This talk makes empirical and theoretical contributions towards answering these questions with novel experimental evidence from Estonian. In Estonian, NPQs which use different question particles but are otherwise identical convey different biases: kas-NPQs being roughly associated with speaker bias for a proposition p and ega-NPQs with speaker bias for ¬p (Metslang 1981, Erelt et al. 1995). The results suggest that it is contextual evidence for ¬p is essential to license kas-NPQs, which I argue can be derived from a pragmatic constraint on congruence between the polarity of a question and the polarity of the expected answer to that question (Trinh 2014, Roelofsen & Farkas 2015). However, contextual evidence for ¬p is insufficient on its own to license ega-NPQs, suggesting that negation is not solely responsible for perceived 'bias' in NPQs, but speaker goals must also be considered.

Anissa Zaitsu (UC Santa Cruz): “The Syntax of (Non-)reduced Why-Interrogatives”

The aim of this talk is to propose a syntactic account for the reduced why-interrogative that I call Why-VP (e.g., Why take Structure of Japanese?). In this kind of question, the subject and tense are missing and the availability of the bare VP is restricted to the presence of why. Despite there being evidence for a (silent) structural subject, I argue against one of the only existing accounts that analyzes these structures in terms of ellipsis (Yoshida et al. 2015). Instead, I propose that the silent subject is PRO in the specifier of a covert infinitival T and that why is a head that selects T. In support of this analysis, I present parallels between Why-VP and (embedded) wh-infinitivals, which are both interpreted as obligatorily modal (Bhatt 1998/1999), necessarily non-past (future or present-oriented), and as having an arbitrary or specific referent for PRO. Assuming this analysis is on the right track, the patterns that emerge come to bear on the availability of PRO in matrix contexts, the typological landscape of modality and its interaction with aspect, and the special status of why in general.