Whamit!

Speaker: Abdul-Razak Sulemana (MIT)
Title: GETCASE is Violable: Evidence for Wholesale Late Merger
Date/Time: Monday, March 6, 5:00–6:30pm
Location: 32-D461

In this talk, I examine reconstruction effects in a class of A-bar constructions in Bùlì, building on recent proposals about the mechanisms that yield reconstructions asymmetries in A and A-bar movement Takahashi and Hulsey (2009) and the asymmetries within English preposition stranding (P-stranding) Stanton (2016).

A well known asymmetry between A and A-bar movement is that: while A-movement bleeds binding Condition C, A-bar movement doesn’t. This led to the conclusion that: while A-movement optionally leaves a trace, A-bar movement obligatorily leaves a copy (Sauerland 1998, Fox 1999). This conclusion, however, posses a serious challenge to the copy theory of movement. To resolve this, Takahashi and Hulsey (2009), extending the idea of late marge (Lebeaux 1988, Chomsky 1995) and adopting insights from (Fox 2002), argue that late merger is allowed whenever an output representation can be interpreted in the semantic component (wholesale Later merger(WLM)). By this operation, they maintain that there is no distinction between A and A-bar movement with respect to the copy theory, independent properties of grammar like Case, account for the reconstruction properties of A and A-bar movement: while WLM can apply to A movement because it involves movement from a non-Case position to a Case position, WLM cannot apply to A-bar movement because A-bar movement involves movement from a Case position to a non-Case position. The goal of this talk is to show that wh-questions in Bùlì, a Gur language spoken in Ghana, provides new evidence for WLM. In particular, I argue that the outcome of overt movement in the language is as a result of ranking the constraint LATEMERGE, which requires constituents to merge as late as possible, above GETCASE, which penalizes a Caseless NP and *TOOLATE, which assigns a violation to late merge if the relationship it establishes is not the structurally highest of its type (Stanton 2016). I argue that the interactions of these constraints are responsible for the cross linguistic variations we observe between A-bar extractions and reconstruction effects in Bùlì as well as other well studied languages, including English.

Data and Analysis: Bùlì permits wh-phrases to appear in the left periphery of the clause (1a-b). The sensitive of these phrases to islands (1c) is taken as evidence to show that they undergo movement.

(1)  a. (ká) b^wa ātì bí:ká dìgì:
Q what C child.DEF cook
‘What is that the child cooked?

b. ká lām būnā ātì bí:ká
dìgì:Q meat which C child.DEF cook
‘Which meat did the child cook?’

c. *ká b^wa ātì bí:ká dà gbáŋ ālī:
Q what C child.DEF buy book CONJ

In analyzing this data, I assume that the QP moves overtly to the Spec, of ātì. However, unlike movement of the whole QP-NP-DP complex from the base position (2), I propose that it involves movement of QP-DP followed by Late merging the NP lām ‘meat’ to the structure at the final landing site (Takahashi and Hulsey 2009, Stanton 2016). This derivation, I argue, is responsible for the lack of reconstruction effects in the language (2b). Since Ajohn foto ‘picture of John’ (2b) is merged after moving the QP and DP, the co-referential pronoun, wà_i ‘3SG’ doesn’t c-command a copy of John in the base position, hence its ability to bleed principle C.

(2) a. [ká [lām] būnā ] ātì bí:ká dìgì ká būnā]

b. ká Ajohn_i foto kūnā ātì wà_i à-yā:lī:
Q John picture which C 3SG IMPF-like
‘*Which picture of John does he like’

Speaker: Snejana Iovtcheva (MIT)
Title: An Applicative Account of Bulgarian Double Object Constructions
Date and time: Tuesday March 7, 1-2pm
Location: 32-D461
Abstract:

I will present and discuss data on Bulgarian ditransitives, in which clitic doubled Goal arguments differ systematically from their non-double counterparts. More concretely, I will demonstrate that clitic doubled ditransitive constructions, behave in par with English Double Object Constructions (DOC) of the type [I gave John the book], while non-doubled ditransitive constructions behave like Prepositional Ditransitive Constructions (PDC) of the type [I gave the book to John].

The DOC/PDC distinction is not obvious right away since the language has free word order and Goal arguments are always marked with the same preposition na. The correlation between clitic doubling in DOC and the absence of clitic doubling in PDC has already been established for Spanish (Cuervo 2003) and Romanian (Rivero & Diaconescu 2006, Diaconescu 2007). In my analysis on Bulgarian DOCs, I follow Marantz (1993), Pylkkänen (2003), Cuervo (2003) and Slavkov (2008) and I propose that na-marked clitic doubled Goals in Bulgarian are introduced by functional ‘Low’ Applicative heads. The clitic itself is treated as a spell out of the Appl⁰.

In addition to contrasting clitic doubled na-Goals to non-doubled na-Goals, I will discuss also na-marked arguments of transitive and unaccusative verbs and I propose that the language has also ‘high’ Applicative heads, which are introduced above the VP domain.

Speaker: Matthew Mandelkern
Title: Bounded Modality
Date and time: Wednesday, March 8, 1-2pm
Location: 32-D461
Abstract:

To what degree does the meaning of an epistemic modal claim like ‘It might be raining’ resemble the meaning of an avowal of ignorance like ‘For all I know, it’s raining’? Progress on this question has been made by exploring differences in how constructions along these lines embed—-in particular by exploring their behavior as part of larger constructions like Wittgenstein (1953)’s ‘It might be raining and it’s not’ and Moore (1942)’s ‘It’s raining and I don’t know it’, respectively. A variety of approaches have been developed to account for those differences. All approaches, however, agree that the infelicity of unembedded Moore sentences and unembedded Wittgenstein sentences is to be explained in roughly the same way: such sentences are classically consistent, but commitment to both conjuncts is incoherent.

In this paper I argue against this consensus. If this consensus were right, then disjoined Moore sentences, and disjoined Wittgenstein sentences, would be felicitous. This prediction is borne out for disjoined Moore sentences, but not for Wittgenstein sentence. This creates a puzzle, since there is decisive reason to think that ‘Might p’ is consistent with ‘Not p’. I propose a new theory of epistemic modals and their interaction with embedding operators which predicts that, while ‘Might p’ is indeed consistent with ‘Not p’, when evaluating their conjunction, ‘Might p and not p’, we are forced to do so relative to an accessibility relation which makes the conjunction false. I show that this theory accounts not only for Wittgenstein sentences and their disjunctions, but also for the subtle behavior of embedded modals across the board. The upshot is that there is much in common between ‘For all we know, p’ and the meaning of ‘Might p’—-and thus much that is correct in the standard semantics for the latter—-but also a crucial difference: interpretation of the latter, but not the former, depends in a striking way on the intersentential dynamics of information.

Speaker: Roni Katzir (MIT and Tel Aviv University)
Title: Choosing between theories of UG using compression-based learning
Time: Thursday March 9, 12:30-2:30pm
Room: 32-D461
Abstract:

I will discuss an approach to learning — compression-based learning — and show how it can help us choose between competing grammatical architectures in some cases where adult judgments alone are insufficiently informative.

Compression (or the principle of Minimum Description Length; also very closely related to Bayesian approaches) considers both the size of the grammar and that of the description of the data given the grammar and attempts to minimize their sum. By doing so, compression guides the learner to hypotheses that balance between generality and the need to fit the data. Compression appears to match subjects’ generalization patterns in a variety of tasks, and it has yielded working learners for realistic linguistic theories in different domains.

I will review these properties of compression-based learning and show how we can use it to compare between competing architectures with two case studies, one in phonology and one in semantics. The phonological case study concerns constraints on underlying representations (also known as morpheme-structure constraints), which were central to early generative phonology but rejected in Optimality Theory. Evidence bearing directly on the question of whether the grammar uses constraints on URs has been scarce. I will show, however, that if the child is a compression-based learner, then they will succeed in learning patterns such as English aspiration if they can use constraints on URs but run into difficulties otherwise. In semantics, I will discuss two architectures for the representation of quantificational determiners: building blocks and semantic automata. While both choices support the representation and learning of quantificational determiners, I will show a specific domain where they predict different learning paths.

Shigeru Miyagawa’s most recent book, Agreement Beyond Phi has just been published by the MIT Press as an LI Monograph. Building on his previous monograph Why Agree? Why Move?, this book investigates agreement in so-called agreementless languages in arguing for a unified view of grammatical features that includes both phi-features and discourse configurational features.

“Miyagawa opens up formal syntax to include discourse-related phenomena and thus contributes to the building of a new research agenda.”—Liliane Haegeman

Congratulations Shigeru!