The Weekly Newsletter of MIT Linguistics

Issue of Monday, November 12th, 2018

Syntax Square 11/13 - Danfeng Wu (MIT)

Speaker: Danfeng Wu (MIT)
Date and Time: Tuesday, November 13, 1-2pm
Location: 32-D461

There has been a debate about whether non-clausal elements can be coordinated, i.e. whether coordination of elements smaller than TP and VP is possible, such as DP coordination and PP coordination (e.g. Gleitman 1965, Ross 1967, McCawley 1968, Hankamer 1979, Schein 2017, Hirsch 2017). In this talk, I will provide arguments supporting coordination of non-clausal elements based on the position of ‘either’ and the scope of disjunction. 

Once we allow non-clausal elements to be coordinated, we can explain a puzzle about adjunct fronting in ‘whether‘-clauses and ‘if‘-clauses. My analysis of this puzzle involves the following components: (a) non-clausal elements can be coordinated; (b) ‘whether’ can pied-pipe adjacent materials in moving to Spec, CP, whereas phonologically null elements cannot pied-pipe; (c) it is not possible to sub-extract out of an A’-position (e.g. a topicalized phrase does not allow sub-extraction).

I will briefly describe the puzzle below. While both ‘whether‘-clauses and ‘if‘-clauses allow the Alt(ernative) reading in (1), when the PP disjunction is apparently fronted to immediately follow ‘whether’/’if’, only ‘whether‘-clauses obtain the Alt reading (2), and the Alt reading is lost in ‘if‘-clauses (3).

(1) I don’t know whether/if John will arrive on Saturday or on Sunday.
✓Alt Reading: I don’t know whether John will arrive on Saturday or he will arrive on Sunday.
✓Yes/No (Y/N) Reading: I don’t know whether John will arrive on a weekend day or he won’t arrive on a weekend day at all.

(2) I don’t know whether on Saturday or on Sunday John will arrive. (✓Alt; ✓Y/N)
(3) I don’t know if on Saturday or on Sunday John will arrive. (*Alt; ✓Y/N)

This puzzle parallels another puzzle previously observed in the literature. When ‘or not’ is at the end of the sentence, we can get Y/N reading for both ‘whether‘-clauses and ‘if‘-clauses (4). But when ‘or not’ is apparently fronted, only ‘whether‘-clauses obtain the Y/N reading (5), and ‘if‘-clauses become ungrammatical (6).

(4) I don’t know whether/if John will arrive this weekend or not. (✓Y/N)
(5) I don’t know whether or not John will arrive this weekend. (✓Y/N)
(6) *I don’t know if or not John will arrive this weekend. (*Y/N)

LF Reading Group 11/14 - Frank Staniszewski (MIT)

Speaker: Frank Staniszewski (MIT)
Title: Deriving the properties of until phrases
Date and time: Wednesday, November, 13, 1-2 pm
Location: 32-D461


     The well-known paradigm in (1-2) presents a challenge for a semantic analysis of until phrases.

(1)      a. He was angry until the end of the conference
           b. He wasn’t available until the end of the conference.
(2)      a. *The bomb exploded until yesterday.
           b. The bomb didn’t explode until yesterday.
          Condoravdi (2008)

Stative/progressive predicates can appear with until phrases with or without negation, as in (1a-b). Perfective predicates, however, can appear with until phrases only under negation, as shown in (2a-b).
Most analyses of this paradigm fall into two basic camps. One proposes that there is only one until, which selects for “durative” predicates. This means that without negation, it can combine only with stative/progressive predicates, thus explaining the badness of (2a) (Klima (1964), Mittwoch (1977), a.o.). In this system, the acceptability of (2b) is a consequence of the assumption that negation is a predicate modifier that can create a durative predicate.
      The other camp proposes that until is lexically ambiguous between two untils. One is a “durative” until that combines only with statives/progressives, and another is a “punctual” until that combines with perfectives, and is additionally assumed to be an NPI (Horn (1970, 1972, Karttunen (1974), Condoravdi (2008), a.o.). Here, the NPI property of “punctual” until derives the ungrammaticality of (2a).
      In this presentation of work in progress, I explore another possible approach. I propose that there is only one until, with a basic meaning similar to before, that is compatible with both stative/progressive as well as perfective predicates. I then hypothesize that this basic meaning of until is obligatorily strengthened in positive environments to its “durative” interpretation by exhaustification, adopting a mechanism similar to ones that have been proposed for obligatory strengthening of other quantificational elements (Bar-Lev & Margulis (2014), Bowler (2014) a.o.). Given this strengthening, until phrases will be incompatible with perfective predicates in upward-entailing environments, thus deriving the badness of (2a).
      I will discuss some arguments for the various approaches, and then develop an analysis that builds on a presuppositional entry for until (Karttunen (1974), Condoravdi (2008)), which in turn interacts with an exhaust operator as proposed in Bar-Lev & Fox (2017) in an attempt to derive the attested empirical patterns.

Phonology Circle 11/14 - Iris Berent (Northeastern University)

Presenter: Iris Berent (Northeastern University) 
Title: Amodal Phonology
Date/Time: Wednesday, November 14, 5:00-6:30pm
Location: 32-D831

This talk evaluates the hypothesis that the phonological grammar is algebraic and amodal. The first part of the talk examines the computational properties of the phonological grammar. Using the restrictions on doubling as a case study, I show that doubling restrictions generalize across the board. Such generalizations are evident in both spoken and signed language phonology, and they require algebraic rules.

In the second part of the talk, I turn to show that the restrictions on doubling are demonstrably dissociable from the phonetic system. I first show that a single linguistic stimulus (e.g. panana) elicits conflicting responses (preference vs. aversion), depending on the level of analysis (phonology vs. morphology). I next show that speakers with no command of a sign language spontaneously project these two parses to novel ABB signs in American Sign language. Moreover, the chosen parse (for signs) is constrained by the morphology of their spoken language. Together, these findings demonstrate that (i) a single, invariant phonetic form can elicit conflicting linguistic parses; whereas (ii) a linguistic parse can remain invariant when the phonetic substance is radically altered – from speech to signs.

These results suggest that doubling in speech and signs is constrained by a single set of linguistic principles that are amodal, algebraic and abstract.


Ling-Lunch 11/15 - Dominique Sportiche (UCLA)

Speaker: Dominique Sportiche (UCLA)
Title: PRO’s Rigid Dependence and some possible consequences.
Date and time: Thursday, 11/15, 12:30-1:50pm
Location: 32-D461

An examination of the study of the behavior of obligatorily or non obligatorily controlled PRO  yields the conclusion that PRO and its antecedent must rigidly covary, a requirement  I call the  `de re ipsa’ requirement. Why this holds suggests ways to derive why `de se’ or `de te’ readings mandatorily arise in obligatory control constructions under attitude control predicates, and in fact elsewhere too. In turn, this behavior of PRO suggests that binding theory could be about the sharing (or non sharing) of recipes for reference which would provide a different way of looking at some binding puzzles discussed in Heim (1994), and Sharvit (2011).

CompLang 11/15 - Hao Tang (MIT CSAIL)

Speaker: Hao Tang (MIT CSAIL)

Title: Automatic speech recognition without linguistic knowledge?

Date and time: Thursday, 11/15, 5-6pm

Location: 46-5165


Building state-of-the-art speech recognizers, besides a large corpus of transcribed speech, require several additional ingredients, such as a phoneme inventory, a lexicon, and a language model.  These ingredients carry linguistic constraints to make training more feasible and more sample efficient. Recently, there has been a push towards building a speech recognizer end to end, i.e., using few or even none of the aforementioned ingredients.  This raises a fundamental question: is it possible to train a speech recognizer without any linguistic constraint?

  How much data do we need to make it possible?  What linguistic constraints are necessary for building a speech recognizer?  In this talk, I will review the inner workings of conventional and end-to-end speech recognizers, and to help answer some of those questions, I will present empirical results in training end-to-end speech recognizers without any linguistic constraints.

Lab Meeting 11/16 - Sherry Yong Chen (MIT)

Speaker: Sherry Yong  Chen (MIT)

Title: Resolving scope ambiguity under (negative) degree questions

Date and time: Friday, 11/16, 2-3pm

Location: 32-D831


This talk presents work-in-progress that provides evidence for facilitation effects of certain types of Question under Discussion (QUD) on scope ambiguity resolution. Specifically, when an ambiguous sentence involves a negation and a comparative quantifier (CQP), a negative QUD facilitates access to the inverse scope reading, but only when the non-negated question nucleus is expected to hold in a given context. We discuss our results with references to the ignorance inference of comparative numerals, maximal informativeness, question answer congruence.