The Weekly Newsletter of MIT Linguistics

Issue of Monday, October 31st, 2016

Syntax Square 10/31 - Chris O’Brien

Speaker: Chris O’Brien (MIT)
Title: Linearization and complete dominance: Deriving the right-edge restriction on RNR
Date/Time: Monday, Oct. 31, 1-2pm
Location: 32-D461

This talk concerns two puzzles involving the multidominant analysis of right-node raising structures (McCawley 1982, Wilder 1999, Bachrach & Katzir 2015, a.o.). The first is how such structures can be assigned a well-formed linear ordering at PF. Following Wilder (1999), I argue that this is because linearization of multi-dominant structures is sensitive to complete dominance. That is, when one phrase A is ordered before a phrase B, everything completely dominated by A must be ordered before everything completely dominated by B.

The second puzzle concerns the “right-edge restriction” on RNR structures (Wilder 1999, Bachrach & Katzir 2015). It turns out that, while the pivot of an RNR structure may appear in a non-rightmost position within the final conjunct, it must be merged in the rightmost position of of each non-final conjunct. My proposal depends on one crucial property of complete dominance: The notion that x completely dominates y must be defined with respect to some larger structure (or set of structures). I argue that linearization is computed compositionally at each step in the derivation. For any phrase A = Merge(B, C) , where some linear precedence rule says that B < C, then all terminal nodes which are completely dominated by B within A will be required to precede all nodes completely dominated by C within A. This turns out to derive the right-edge restriction. I end with discussion of some recalcitrant problems involving internal merge structures.

Phonology Circle 10/31 - Ting Huang

Speaker: Ting Huang (MIT)
Title: Contrast and context-dependent merger: Evidence from Malaysian Mandarin sibilants
Date/Time: Monday, October 31, 5:00–6:30pm
Location: 32-D831


This study reports an ongoing merger of Mandarin sibilants spoken in Malaysia. The contrast of dental/alveolar vs. palatal sibilants in Malaysian Mandarin (MM) is neutralized in the context of high-front vowel. Specifically, while the contrasts between [ɕa] vs. [sa] and [ɕu] vs. [su] exist, [si] is the only surface form of the coronal sibilant followed by a high-front vowel /i/ ([*ɕi] is not allowed) in MM (we ignored the retroflex sibilants here, which is irrelevant to this study). We provide evidence from palatography and linguography to show a fine-grained difference among these sibilant variants in place of articulation. The results of spectral moments analysis (Forrest et al. 1988; Jongman 2000; Lee 2014) and F2 onset values (Li 2008; Wilde 1993) also support the argument that the MM sibilants are incompletely neutralized, especially for speakers of younger generation. The phenomenon in question may be attributable to language contact-induced sound change. This also casts doubt to the feature-based account (Clements 1991; Hume 1992) in explaining why [-anterior] of [ɕ] can be retained when followed by a following vowel that is specified with [dorsal] (e.g. [u], [a]), but not by those with [coronal] (e.g. [i]). We extend the line in Flemming’s (2003) that tongue-body position should be specified under [coronal], and argue that distinctiveness of sibilant contrasts may rely as well on tongue-body position of vowels.

LFRG 11/2 - Peter Alrenga

Speaker: Peter Alrenga (Boston University)
Time: Wednesday, November 2nd, 1-2pm
Place: 32-D831
Title: At least and at most: Ignorance and variation in focus.

A hallmark feature of the scalar operators “at least” and “at most” is their capacity to convey speaker uncertainty: from an utterance of (1), a listener would typically infer that the speaker does not know the exact number of points that LeBron scored.

(1) LeBron scored at least / at most 20 points in last night’s game.

These uncertainty implications tend to disappear in the presence of modals: under their most salient interpretations, neither (2a) nor (2b) need convey any uncertainty regarding what is necessary or required:

(2) a. (In order to win the scoring title), LeBron needs to score at least 45 points in tonight’s game.
b. One person can submit at most one abstract as sole author and one abstract as co-author (or two co-authored abstracts).

Rather, the most salient interpretations for these sentences convey variation in what the speaker deems to be sufficient or permissible. Similar variation implications can also be observed in combination with nominal quantifiers:

(3) a. Every player scored at least 10 points in last night’s game.
b. Individuals can give to as many federal candidates as they want, so long as they give at most $2600 to any single candidate in an election cycle.

The question of exactly how “at least” and “at most” manage to convey uncertainty and variation in (1)-(3) has attracted considerable scrutiny. Recent work has converged on the view that these implications are implicatures arising from the interaction of the basic semantic properties of at “least / at most” with general pragmatic mechanisms. A near-universal impulse of these pragmatic approaches is to draw an analogy to disjunction, which gives rise to a similar pattern of uncertainty and variation implications. But capitalizing on this analogy has proven surprisingly difficult. In its most direct form, it amounts to the view that “at least” and “at most” form n-ary disjunctions over their associated scalar terms and all higher / lower ones. While such a view correctly characterizes the truth-conditional contribution of “at least”, it appears to to mischaracterize its pragmatic behavior. And without further amendment, it fails to even adequately capture the truth-conditional contributions of “at most”.

In the first part of this talk, I argue that a version of the simple view can indeed be maintained for “at least”, once it is recognized that (i) the scales that “at least” and “at most” operate over are fundamentally pragmatic/contextual in nature, and (ii) these scales are never ordered by entailment. While the simple n-ary disjunction view cannot be maintained for “at most”, I show how its essential insights into “at most“‘s pragmatic behavior nevertheless can be. In the second part of the talk, I apply the resulting analysis to certain unresolved problems concerning the interactions of these scalar operators with modals and other quantifiers.

ESSL/LacqLab Meeting 11/01 — Jie Ren

Speaker: Jie Ren (Brown University)
Title: Underspecification in Toddlers’ and Adults’ Lexical Representations
Date: Tuesday, November 1st
Time: 1—2pm
Location: 32-D461

Theories of underspecification claim that certain unmarked features are empty/underspecified in lexical representation. This hypothesis predicts asymmetrical judgments in lexical processing. In particular, noncoronal tokens such as paan can putatively activate a coronal entry taan, but not vice versa. Studies with both younger infants and adults had found that participants are more sensitive to noncoronal-to-coronal than to coronal-to-noncoronal changes. In this talk, I will report a series of studies that examined toddlers’ and adults’ sensitivities to these two types of changes in mispronunciations of familiar words using the visual world paradigm. Unlike the prediction of underspecification, 19-month-olds and adults showed significant effects in both directions of mispronunciations, and no asymmetries were attributable to underspecification of coronal sounds. Toddlers’ lexical representations appear to be as detailed as those of adults, and there is a striking developmental continuity between early and mature lexical representations. Finally, I will report a computational model which suggests that discrepancies between the current findings and those of previous studies appear to be due to methodological differences that cast doubt on the validity of claims of psycholinguistic support for underspecification.

Ling-Lunch 11/03 — Jon Rawski

Speaker: Jon Rawski (Stony Brook)
Tittle: Homeostatic Reinforcement Learning for Harmonic Grammars
Date/Time: Thursday, November 3/12:30pm-1:50pm
Location: 32-D461

The main idea of this talk is to bridge a particularly thorny divide between linguistics and neuroscience. Reinforcement Learning (RL), despite being one of the most widely used and neurologically robust learning algorithms, has an uneasy history with generative grammar. Specifically, the requirement of an internal, restricted hypothesis space and other learnability restraints is inadequately satisfied by externally defined “naive” reward (Chomsky 1959).

Reparation of RL and linguistics is made urgent by the discovery that: 1) phonology is at most a regular language (Kaplan & Kay 1994, Heinz 2011), meaning it is restricted to finite-state automata, and 2) RL is perfectly computed by cortical neurons (Schultz et al 1997). One recent attempt is Charles Yang’s (2002) “Naïve Parameter Learner”, which uses RL to successfully model acquisition of overt [WH-movement] and [V2] parameters, yet fails to provide more than an ad-hoc definition for “reward”.

In this talk I show that recent insights from computational neuroscience offer a possible strategy. A recent framework called Homeostatic Reinforcement Learning (HRL) (Keramati and Gutkin 2014) treats “reward” as an internal satisfaction of multiple, parallel constraints in a homeostatic space. This immediately suggests Harmonic Grammar. I posit that the weighted constraints in Harmonic Grammar constitute a homeostatic space, and the Harmony function is a necessary and sufficient condition for RL in constraint-based grammars. I then show that this model successfully learns final obstruent devoicing in Russian, among others. I conclude with some tentative hypotheses for homeostasis in bilinguals and in late-L2 learners. Apart from interesting models and simulations, this approach offers prospects for uniting ideas from neural and linguistic theory in order to provide a more coherent explanatory neurolinguistics.

Colloquium 11/4 - Peter Svenonius

Speaker: Peter Svenonius (University of Tromsø)
Title: Emergent Extended Projections
Time: Friday, November 4th, 2016, 3:30-5:00pm
Venue: 32-155

The theory of extended projections (Grimshaw 2005) is built on a strongly universalist/innatist premise, especially in its cartographic implementation (Cinque 1999 inter alios). On that view, the LAD (language acquisition device) matches instantiated categories in the input to a prespecified sequence of hierarchically arranged categories in UG. In this talk, I explore the implications of a sparer UG. I suggest how extended projections might emerge from the primary data, given certain assumptions about the LAD. I suggest that these assumptions give a more satisfying understanding of mixed projections (Abney 1987) and some other phenomena than do the standard assumptions about extended projections.

Pumpkin carving!

On Wednesday October 26th, MIT Linguistics celebrated the Halloween season with an annual pumpkin carving party. Some of the results:

(photo credit: Snejana Iovtcheva)