Whamit!

The Weekly Newsletter of MIT Linguistics

Issue of Monday, May 13th, 2019

GLOW42 @ University of Oslo

The 42nd GLOW conference (GLOW = “Generative Linguistics in the Old World”) took place at the University of Oslo from May 7th to May 9th. Rafael Abramovitz (4th year) presented at the workshop on “Anaphora at the syntax-semantics-pragmatics interface in endangered and understudied languages”. The title of his talk was What can agreeing anaphors tell us about the Anaphor Agreement Effect?.

Gregor Williamson (University College London), a current visitor at MIT, also prented at the main session (Adverbial Adjunct Clauses and their LFs). Recent alumni Benjamin Storme and Coppe van Urk also presented.

LF Reading Group 5/15 - Gregor Williamson (UCL/MIT) & Jacopo Romoli (Ulster)

Two presenters at LFRG this week: Gregor Williamson (UCL/MIT) and Jacopo Romoli (Ulster University, joint work with Lyn Tieu and Cory Bill). The details of their presentations are below.

Speaker: Gregor Williamson (UCL/MIT)
Title: Conditional Antecedents as Polar Free Relatives
Time: Wednesday, May 15th, 1-2PM
Location: 32-D461

Abstract:
Kratzer (1986, 2012), building on Lewis (1975), develops one of the most successful accounts of conditionals to date: the if-clause-as-restrictor account. It maintains that if contributes no meaning to a conditional construction. Rather, the antecedent of a conditional simply denotes a proposition, which may act as a restrictor for a (covert) modal operator in the consequent. A popular alternative to this account, the if-clause-as-a-definite-description-of-worlds account (Schlenker, 2001), has been argued by Bhatt & Pancheva (2006) to be supported by syntactic facts which suggest that if-clauses are free relatives which denote a definite description of worlds. We propose a syntax-semantics for if-clauses which treats them as free relatives formed via a polar question operator (see also Arsenijević, 2009). We show that such an account provides a more transparent syntax-semantics mapping than that of Bhatt & Pancheva. The proposed account maintains a Kratzerian semantics, while nonetheless capturing the well-established syntactic behavior of if-clauses.

…and:

Speaker: Jacopo Romoli (Ulster, joint work with Lyn Tieu and Cory Bill)
Title: Homogeneity or implicature: An experimental study of free choice

Abstract:
A sentence containing disjunction in the scope of a possibility modal, such as (1a), gives rise to the free choice inference in (1b). This inference presents a well-known puzzle in light of standard treatments of modals and disjunction (Kamp 1974 and much subsequent work). To complicate things further, free choice tends to disappear under negation: (2a) doesn’t merely convey the negation of (1a), but rather the stronger dual prohibition reading in (2b). There are two main approaches to the free choice-dual prohibition pattern in the literature, based on implicature and homogeneity. We present experimental findings that favour the homogeneity approach, and further discuss how the implicature approach could be developed to account for the results.

(1) a. Sue is allowed to buy the boat or the car. b.Sue is allowed to buy the boat and is allowed to buy the car

(2) a. Sue is not allowed to buy the boat or the car. b. Sue is not allowed to buy the boat and is not allowed to buy the car

Phonology Circle 5/15 - Boer Fu (MIT)

Speaker: Boer Fu (MIT)
Title: Contrast Preservation in Mandarin R-suffixation
Time: Wednesday (5/15), 5:00pm-6:30pm
Location: 32-D831

Abstract: The low vowel /a/ has three surface forms in Mandarin, a front [a] when followed by an alveolar nasal, a back [ɑ] before a velar nasal, and a central [ɑ̟] in an open syllable. This variation is called the rhyme harmony. My project looks into how these three stem forms change, respectively, in r-suffixation, where a diminutive suffix /-r/ is attached to the stem. In the two vowel-nasal forms, the nasal stop is dropped before the /-r/ coda. However, the back [ɑ] retains its nasalization, while the front [a] is loses its nasalization. Zhang (2000) suggests that this is to maximize the contrast in degree of nasalization that is already in place in the stem forms. He also argues against a contrast preservation account because the front [a] and central [ɑ̟] are neutralized, now that the nasalization on [a] has disappeared. This is indeed true in the Beijing dialect, which the standard Mandarin is closest to. However, I have found that in the Liaoning dialect, such a neutralization does not happen. And [a] and [ɑ̟] are still distinguished from each other by the difference in the quality of the /-r/ coda. I argue that this is contrast preservation, and can be analyzed using MinDist constraints.

CompLang 5/13 - Ethan Wilcox (Harvard)

Speaker: Ethan Wilcox (Harvard University)
Title: Neural Network Language Models as Psycholinguistic Subjects: The Case of Filler—Gap Dependencies
Time: Thursday, 5/13, 5-6pm
Location: 46-5165

Abstract:
Recurrent Neural Networks (RNNs) are one type of neural model that has been able to achieve state-of-the-art scores on a variety of natural language tasks, including translation and language modeling (which is used in, for example, text prediction). However, the nature of the representations that these ‘black boxes’ learn is poorly understood, raising issues of accountability and controllability of the NLP system. In this talk, I will argue that one way to assess what these networks are learning is to treat like subjects in a psycholinguistic experiment. By feeding them hand-crafted sentences that belie the model’s underlying knowledge of language I will demonstrate that they are able to learn the filler—gap dependency, and are even sensitive to the hierarchical constraints implicated in the dependency. Next, I turn to “island effects”, or structural configurations that block the filler—-gap dependency, which have been theorized to be unlearnable. I demonstrate that RNNs are able to learn some of the “island” constraints and even recover some of their pre-island gap expectation. These experiments demonstrate that linear statistical models are able to learn some fine-grained syntactic rules, however their behavior remains un-humanlike in many cases.

6th TripleA workshop @ MIT

The TripleA workshop will take place at MIT from May 31st to June 2nd. Invited speakers are Diti Bhadra (Harvard), Sandra Chung (UC Santa Cruz), Virginia Dawson (UC Berkeley) and Ken Safir (Rutgers). Registration is open until May 18 and the program can be found here: http://triplea6.mit.edu/program. Frank Stanizweski (3rd year) will be presenting his work on Wolof, with title Wolof (non) polarity sensitive item dara.

MIT@FASL28

The 28th Workshop on Formal Approaches to Slavic Languages took place at Stony Brook University from May 3rd to May 5th. Several of our students presented:

The Fourth American International Morphology Meeting (AIMM4) was also held at Stony Brook University on the same dates. Colin Davis gave a talk titled Plural suppletion in Barguzin Buryat: Case containment versus *ABA.