NYU Syntax Brown Bag Talk Series

April 21    Kunio Kinjo (Rutgers) 10:15am
Agree by minimal search: A probe-free model of agreement

In this talk, I present a new model of Agree, elaborating the idea speculated in Chomsky (2013, 2015) that labeling and agreement employ the same search procedure (minimal search, MS). Specifically, I propose that MS, whether it is employed for labeling or agreement, applies at the phase level and cannot see lower copies. When it is employed for Agree, MS scans a phase from its root node and establishes Agree relations between matching features successively in a top-down fashion. With this model, I examine various verbal agreement patterns, mainly focusing on Bantu languages, where more than two φ-bearing elements appear within a CP phase: verbal agreement in the subject inversion constructions (Kinyalolo 1991; Baker 2008 a.o.), hyperagreement in the complex tense construction (Henderson 2006; Carstens 2011 a.o.), and alternative-agreement (a.k.a antiagreement) in subject extraction contexts (Ouhalla 1993; Schneider-Zioga 2007; Diercks 2010; Henderson 2013 a.o.).

April 14    Abdul-Razak Sulemana (MIT) 10:15am
Q-particles and the nature of Covert movement: evidence from Bùlì

It is a well known fact that wh-questions in many languages may contain an in-situ wh-phrase. The nature of this wh-phrase, however, has been a contentious issue in the literature. While some have argued that the in-situ wh-phrase undergoes covert movement at LF (Aoun, Hornstein, and Sportiche, 1981; Huang, 1982; Nishigauchi, 1990; Pesetsky, 2000, Richards, 1997; 2000; Nissenbaum, 2000; Cable, 2007; 2010; Kotek, 2014; 2016), others have argued against this view (Watanabe 1992; Chomsky 1995; Reinhart 1998). A well-known puzzle for proponents of covert movement are the apparent differences in island-sensitivity between overt and covert movement — leading Huang (1982), for example, to propose that island-sensitivity is a property of S-structure or PF but not LF. The goal of this paper is to show that wh-questions in Bùlì provide strong arguments for covert movement of wh-in situ that eliminate the need to posit any overt/covert differences in island-sensitivity cross-linguistically. The key to this demonstration is the distribution of an overt Q-marker in Bùlì, and Bùlì's status as an in-situ language.

March 10    Dunja Veselinovic (NYU) 12:30pm
Parameters of syntax of must and can (and how they must and can affect language acquisition)

This talk examines the syntax of modal verbs, and more narrowly the distinctions and similarities between epistemic and root uses of modals such as can and must. Starting from a Kratzerian semantic framework (Kratzer 1977, 1981 i.a.), I present evidence from Bosnian/Croatian/Serbian (BCS) showing that previous syntactic analyses (Cinque 1999, Hacquard 2006, 2010, i.a.) cannot fully capture the distinctions between the epistemic and root uses of verbs morati ‘must’ and moći ‘can’. Instead, I argue that epistemic modal constructions in BCS are biclausal, in that the modal complement contains a full CP (1), whereas root modal constructions are monoclausal, in that the modal complement is a vP (2). For English epistemic modals, I adopt a modified raising analysis proposed at least as early as Ross (1969) (3).

(1) [TP2 … [ModP epist. [vP … [CP [TP1]]]]]                (epistemic, BCS)

(2) [TP1 … [ModP root [vP …]]]]                                   (root)

(3) [TP2 … [ModP epist. [TP1]]]                                    (epistemic, English)

I tie this analysis to corpus and experimental work on language acquisition, which has shown that children between the age of 2 and 3 experience what is called the Epistemic Gap (EG). During EG, children use modal auxiliaries only in root contexts, with epistemic uses appearing only once children show evidence of ability to embed TPs (Cournane 2015). My corpus work based on Anđelković, Ševa and Moskovljević (2001) CHILDES corpus shows that BCS children’s Epistemic Gap (EG) is protracted, lasting at least until children are 4 years old. I argue that this is due to the parametric differences in the syntax, as children have cross-linguistically been shown to not use CP-complements until after 4 years old (Diessel 2004; De Villiers & Roeper 2016).

December 9    Ian Roberts (Cambridge/UConn)
The Null Subject Parameter in the 21st Century

I begin by giving a very brief historical sketch of work on  null subjects, and summarising the typology of null subjects, distinguishing consistent, partial and radical null-subject/null-argument systems, following Roberts & Holmberg (2010). I then present the recent important proposals in Barbosa (2014) concerning the relation between partial and radical null subjects. Following and developing a suggestion of hers (Barbosa 2014:49), this is extended to consistent null-subject languages and, further, to non-null-subject languages. This leads to a typology of null arguments which links their properties directly to the D-system in general, suggesting a cross-linguistic link between the nature of the null-subject system and the general nature of the “article system” in a given language. Next, I look briefly at the proposals for the semantics of null pronouns and pronouns in general, summarising and building on ideas of Barbosa (2014) and Elbourne (2005). This leads to a consideration of the role of the Person feature in licensing null arguments, following proposals in Longobardi (2008) and Richards (2014). Finally, a general account of “licensing pro” is put forward, which relies on the twin ideas that pro contains a variable and that all variables must be bound at the C-I interface (cf. the Bijection Principle of Koopman & Sportiche 1982/1990). 

December 9    Matt Barros (Yale) and Robert Frank (Yale)
                 Discourse Domains and Syntactic Phases: A Constraint on Multiple Sluicing

Merchant (2001) notes that the clause-mate requirement on multiple sluicing (sluicing with more than one wh-phrase survivor—Takahashi 1994) is suspended when the embedded subject is a pronoun bound by the matrix subject. 

(1) Some student-i claimed {he­-i/*Sally} met with some professor, but I don’t know which student with which professor.

Analyses of this pattern (e.g., Grano and Lasnik 2015) have focused on the role of the bound pronoun in extending the domain over which multiple sluicing can obtain. In this talk, we present new syntactic contexts where clause boundedness is relaxed in the absence of bound pronouns, and others where bound pronouns do not extend the domain, suggesting the bound pronouns are neither necessary nor sufficient. We argue instead that the clause-mate requirement in multiple sluicing is active only when the embedded subject is a DP that lacks a higher co-referential element in the matrix clause.  We label such subjects “shifty," in that they shift the focus of attention in the embedded clause from discourse referents mentioned in the matrix. In the absence of a shifty subject, the embedded and matrix clauses constitute a single domain of syntactic computation, leading to the possibility of multiple sluicing across the clause boundary. We show how this idea can be implemented in a phase-based analysis of clause-boundedness effects and show how it extends to phenomena such as scope and gapping.

December 2    Hadas Kotek (Yale)
                 Unifying definite and indefinite free relatives: Evidence from Mayan (joint work with Michael Yoshitaka Erlewine)

Free relatives (FRs) such as I liked [FR what I read] are described cross-linguistically as a type of relative clause of DP size and definite semantics (Jacobson, 1995, a.o.). Some languages also have what seem to be indefinite FRs, but Izvorski (1998), Grosu (2004), and Šimík (2011) argue that these are structurally distinct and better described as Modal Existential wh-Constructions (MECs). In this talk we present a case where definite and indefinite FRs are more similar than previously thought. We present data from several understudied Mayan languages, focusing primarily on Chuj (Guatemala), where indefinite FRs show a subset of the properties claimed to hold universally of indefinite FRs as MECs. We propose a uniform internal syntax and semantics for FRs and argue that definite and indefinite FRs differ only in their external environments. We analyze Mayan indefinite FRs as property complements of existential verbs (see e.g. Milsark, 1974; McNally, 1998) lacking a DP layer. A DP layer can then be added to form definite FRs, available in any argument position. The study contributes to the typology of FRs cross-linguistically, challenging the claim in Šimík (2010) a.o. that all apparent indefinite FRs must be MECs with modal interpretation.

November 18    Laura Kalin (Princeton)
A new model of nominal licensing

Beginning with Vergnaud 1977 and Chomsky 1980, a rich tradition in generative syntax holds that, in addition to bearing a semantic (theta) role, nominals need to be formally “licensed” by receiving abstract Case. In this talk, I examine two types of nominal restrictions that are typically taken to fall outside the purview of this system, the Person Case Constraint (Bonet 1991) and Differential Object Marking (Comrie 1979, Bossong 1991, i.a.). Based on novel observations about parallels between the PCC and DOM, I turn the predominant theoretical model of abstract nominal licensing (Chomsky 2000, 2001) on its head: I propose that licensing is not a general property of nominals, but rather, is driven by the needs of (certain) “interpretable” nominal features (e.g., [participant], [definite]), rather than by abstract Case. In other words, whether a nominal needs licensing---and in what configurations a nominal may be abstractly licensed---depends on the valued features the nominal bears. The proposed system is also driven by a distinction between obligatory and secondary licensing loci (Levin and Massam 1985, Bobaljik 1993, Rezac 2011, i.a.), with secondary licensers activated only when a feature that requires licensing would otherwise go unlicensed.

November 4    Ken Hiraiwa (Meiji Gakuin/MIT)
Decomposing Indeterminates and Composing Indefinite Pronouns

Japanese builds up various indefinite pronouns by combining “indeterminates” and particles (Kuroda 1965, 2013, Nishigauchi 1990, Shimoyama 2001, among others). While this indeterminate system is thought to be quite systematic, a close look reveals that it does have an gap that has been ignored and unexplained: the non-human universal quantifier “nani-mo” (everything) is missing.

In this talk, I will argue that this gap is by no means accidental. Probing into the internal syntax of indeterminates, I will propose that indeterminates are further decomposed into demonstrative roots and noun class elements and that the gap is precisely predicted by the decompositional approach and the labeling algorithm. I will also explore some consequences of the proposed analysis.

If this line of argument is right, what looks like an accidental and trivial gap provides a crucial window into the nature of the structure-building algorithm in narrow syntax.

October 7    Shih-Yueh Jeff Lin (NYU)
Ulivelivek Wh1 as the Pseudocleft

In this talk, I will show that in Ulivelivek, a dialect of Puyuma spoken in the south-east part of Taiwan, wh-ex-situ questions involve the pseudocleft construction (Aldridge, 2004), arguing that the word maw in the sentence-initial position is a copular verb that expresses the identity between individuals as proposed by Montague (1974). Specifically, the evidence from reduplication, affixation, cliticization, negation, and the response to polar questions shows that maw is verbal in nature. The selectional restriction that maw only co-occurs with DPs supports Montague’s semantic translation for copular verbs, as well as the pseudocleft analysis. I will also propose that the copula-initial word order is derived by VP remnant movement. The conclusion is that [i] Ulivelivek as a V1 language does not have true Wh1 word order, contra Hawkins’ (1983) Generalization that if a language has dominant V1 word order, it tends to have Wh1 word order in wh-questions, and [ii] at least some Austronesian languages do have a copula as well as category distinctions, which supports Richards’ (2009) claim. 

                 The morphological expression of nominal number in Kalenjin

The number system of Kalenjin (a Nilotic language spoken in Kenya) is representative of the systems of Nilo-Saharan languages, which are well-known for their complicated number morphology in the nominal domain (Corbett 2000). The majority of these languages have a tripartite system of number marking: some nouns are interpreted as singular in their morphologically unmarked form and form their plural by the addition of a plural affix, some nouns are interpreted as plural in their unmarked form, and form their singular by the addition of a singulative affix, while others never appear in their unmarked form: they have a singulative affix in the singular, and a plural affix in the plural. The puzzle in such a system lies in the lack of a one-to-one relationship between semantic and morphological markedness irrespective of what number value is chosen as the unmarked one. Even though these systems have gained some attention in the typological literature (Dimmendaal 2000), there has been no theoretical work on their implications for the syntax of number. The purpose of this paper is to fill this gap by describing the number morphology of nouns in Kalenjin and by showing that the data support a split theory of number (Kramer 2009), in which number features are found both on the nominalizing head (little n), and on the head of the functional projection NumP. It is argued that nouns in Kalenjin belong to different classes depending on the kind of number features present on n, in the same way that nouns in Indo-European belong to different classes depending on biological sex gender features on n (Kramer 2015). Noun classification in Kalenjin works in the same way as number-based classification in Kiowa (Harbour 2007)

September 30    Anna Maria Di Sciullo (UQAM/NYU)
On the pronunciation/silence of certain prepositions in Italian

Microparametric variation is tied to specific features and lexical items (Kayne 2005, 2006, 2012; Collins 2007; Liao 2013, a.o.).  We focus on here and there in Italian qui and li, and in Fallese aecche and aloche, a dialect spoken in Abruzzi, where the preposition a (AT/TO) can be pronounced in some cases. We argue that the variation in the pronunciation/silence of the preposition follows from a minimal derivational difference, given independent properties of the computational system, where feature valuing is done via Merge. Furthermore, assuming derivation by phases (Chomsky 2008 and seq.), and that either the Specifier or the Head must be pronounced, and that when the Spec has phonetic feature, it must be pronounced (Collins 2007), it follows that AT/TO is silent in Italian, and that it must be pronounced in Fallese when the Spec of AT/TO is not filled. We argue further that the apparent optionality in the pronunciation of the preposition in Fallese, (a)ecche, (a)loche, and its silence in Italian follows from a difference in the feature specification of the preposition in the languages under consideration. We discuss the predictions of our analysis as well as some consequences for the theory. We predict that other prepositional heads may optionally be pronounced in the historical development of prepositional structures. This is the case in the development of comitative P-Pronoun structures, from Latin cum me/co me, me cum/me co “with me, me with”, to Old Italian con me co “with me with”, to modern Italian con me, “with me” (Di Sciullo et al. 2016). Our analysis supports Minimalist syntax as the variation in the pronunciation/silence of certain prepositions, including their apparent optionality in some languages, follows from the theory.

September 23    Chris Collins (NYU), Room 104 of 10WP
The Khoisan Languages of Botswana

I give a brief overview of the Khoisan languages of Botswana. Then, I discuss the linguistic fieldwork that I did during 2015-2016, focussing on the creation of a Sasi dictionary. The talk is meant for a general audience, including non-linguists.

September 16    Kensuke Takita (Meikai/UConn)
Labeling for Linearization

The general framework of Chomsky (2013, 2015) introduces a distinction among syntactic objects, namely, syntactic objects that can be labeled and ones that cannot be, which is quite novel because they have been taken either always labeled (as in the X-bar theory and Chomsky 1994’s version of the bare phrase structure theory) or totally label-free (see, for instance, Collins 2002, Seely 2006, Narita 2014). In this talk, I examine what consequences the distinction may have for linearization. Specifically, I first hypothesize that an unlabeled syntactic object of the form {XP, YP} is invisible to linearization so that the relative order between XP and YP cannot be determined, leading to an unpronounceable result. Then, I show that this hypothesis sheds new light on several issues concerning the invisibility of “traces” to the labeling algorithm, EPP-(un)satisfaction by null elements, a potential variability with respect to the size of Spell-out domains.

May 16    Ivona Kučerová (McMaster)
                 The rise and fall of nominatives

A Syntax 101 theory of case assignment posits that nominative is an abstract case assigned by finite T. Yet, nominatives appear in environments where they cannot be licensed by finite T (as in Icelandic infinitival complements), or there can be more than one nominative noun phrase per finite T (as in NP-NP copular clauses). However, if a nominative phrase is in nominative, it triggers agreement, interacts with other DPs in PCC constructions, and shows restrictions on person in general. Morphological case theories have an easier time with the distribution of nominative forms in the structure but equally do not provide much insight as to why nominatives exhibit these consistent syntactic properties. In other words, there seems to be a notion of nominative-ness independent of case licensing.
I provide novel empirical evidence that this notion of nominative-ness corresponds to a nominal structure labeled by a D head (Chomsky 2013) where that the core labeling feature formally corresponds to CI licensing of [+PERSON] feature (see also Sudo 2012, Longobardi 2008, Landau 2010). This work thus provides new empirical evidence for a formal connection between Case and PERSON (Schutze 1997, Martin 1999, Chomsky 2000, Bejar and Rezac 2003, Richards 2008, a.o.). The core empirical evidence for the proposal comes from a micro-variation in Slavic numeral constructions. As we will see, if the internal structure of the numeral blocks PERSON from labelling the structure, the resulting structure may appear in argument positions typically occupied by nominatives, but they fail to agree, license secondary predicates, interact with person features of other DPs and create boolean conjunction.

Apr 8      Yohei Oseki (NYU)
                Primitive Functional Elements of Argument Structure

In this talk, we explore primitive functional elements in the domain of argument structure. Specifically, under the strong working hypothesis that functional elements with the same sound must have the same meaning (“No Homophony” Hypothesis of functional elements; Richard Kayne, class lecture at NYU in Spring 2016), Japanese verbs paradigmatically identified in the previous literature (Miyagawa, 1984; Jacobsen, 1992; Volpe, 2005) are syntagmatically decomposed into primitive morphemes. Following the generative constructionist approach to argument structure (Scafer, 2008; Pylkkanen, 2008; Wood, 2012; Marantz, 2013), we then propose the system in which these primitive morphemes are identified as functional heads like little v, Appl, or Voice and combined to build the argument structure skeleton. Theoretical implications are two-fold: (i) the proposal that Voice and little v are two distinct functional heads is corroborated (Harley, 2013; Legate, 2014) and (ii) the syntactic approach to morphology should be correct, where there is only one computational engine to build complex "words" and phrases (Koopman & Szabolcsi, 2000; Koopman, 2005).

Apr 1        Yuta Sakamoto (UConn)
Clausal Complement Anaphora in Japanese: Deep or Surface?

In this talk, I investigate the possibility of extraction out of both overt and covert anaphora sites in Japanese, i.e. extraction out of clausal complements that are "replaced" by soo 'so' and clausal complements that are phonologically missing. Specifically, I show that both of them allow certain types of extraction out of them, unlike clausal complement anaphora in English, where extraction is uniformly banned out of its domain. Based on the extraction possibility, I then argue that the Japanese cases in question are instances of ellipsis, not pro-forms. Furthermore, I argue that both deletion and LF-copying are available strategies for implementing ellipsis. In particular, I argue that "replaced" clausal complements are best analyzed in terms of deletion and silent clausal complement anaphora in terms of LF-copying.

Mar 25    Steven Foley (UCSC)
                Morphological conspiracies in Georgian agreement 

A conspiracy arises when more than one (e.g. phonological) process serves to enforce a single constraint on surface forms. A major theoretical advantage of an Optimality Theoretic grammar is the ability to capture conspiracies: instead of relying on otherwise unconnected rules that just happen to prevent some marked structure, OT allows us to refer directly to it by ranking a markedness constraint above relevant faithfulness constraints.

In this talk I identify a morphological conspiracy in Georgian, and use it to argue that morphology is governed by an OT grammar. Again and again, the language’s agreement system goes out of its way to avoid Multiple Exponence — the presence of more than one morpheme in a word exponing a single feature. Abstractly, when probes X and Y both Agree with a single argument for feature [F], and morphemes α and β can spell out [F] on X and Y respectively, multiple exponence of [F] is avoided by blocking the insertion of either α or β. Within Distributed Morphology (DM; Halle & Marantz 1993), such blocking relationships can be derived through some suite of postsyntactic operations, like impoverishment. However, these operations do not refer to multiple exponence directly, and thus fail to capture the conspiracy.

Instead, building on previous work in OT morphology (Kiparsky 2000, Trommer 2001, Wolf 2008, Caballero & Inkelas 2013), I propose that Vocabulary Insertion — the operation that chooses which morphemes expone which syntactic terminals — is governed by ranked, violable constraints. DM's Subset Principle is decomposed into morphosyntactic faithfulness constraints; highly ranked morphosyntactic markedness constraints (like *MultipleExponence) replace DM's postsyntactic operations. I show that Georgian's conspiracy against multiple exponence, along with other peculiarities of its agreement system, follow from standard constraint interactions.

Mar 11    Tricia Irwin (UPenn)
Verb vs. structure: Who's the boss?

In this talk I'll discuss the tension between verb meaning and syntactic structure with respect to intransitive sentences like A boy walked in. I'll show that in sentences like these, the syntactic structure that 'walk' occurs in matters more for VP meaning than the contribution of 'walk' on its own. This analysis helps explain some observations on PP extraposition that go back to Guéron (1980), and it also leads to the possibility that some sentences that we might not have thought were structurally ambiguous actually are.

Jan 29    Chris Tancredi (Keio University) 2pm, Rm 103 of 10WP
Contrastive Topic, Focus and Givenness

In this talk I will argue that contrastive topic, focus and givenness are three distinct, independent phenomena.  In particular, I will argue that Givenness cannot be analyzed as a side effect of focus interpretation and the reverse is impossible as well; that focus and contrastive topic, while behaving in similar ways, need to be formally distinguished locally; and that Given contrastive topics need to be distinguished from non-Given contrastive topics.  While this may look like the worst of all possible worlds, I will also show how Constant’s 2014 insights into how to reduce contrastive topic interpretation to focus interpretation can be preserved by making small changes to the syntactic assumptions underlying his analysis.