The Weekly Newsletter of MIT Linguistics

Issue of Monday, February 21st, 2022

LFRG 2/23 - Elizabeth Coppock (BU)

Speaker: Elizabeth Coppock (BU)
Date and time: Wednesday February 23, 1-2pm
Location: 32-D461

Title: Triangle Equivalences

Abstract:  Denominator phrases like “per ton” can combine either with expressions denoting particular degrees, as in “$100 per ton”, or with expressions denoting measure functions, as in “cost per ton”. As we see regularly when we look at translations of “per” into other languages, there is a systematic pattern of equivalences involving these two options:

(1) a. The cost of wheat is $100 per ton. / b. The cost of wheat per ton is $100.
(2) a. the EUR 100 billion per annum shortfall / b. the per annum shortfall of EUR 100 billion
(based on English-Bulgarian aligned sentences in Europarl)
In the (a) examples, an expression denoting a specific quantity combines with a denominator to yield a fractional degree-denoting term (“$100 per ton”, “EUR 100 billion per annum”) that is somehow equated with a measure function (“cost”, “shortfall”), or its value when applied to some implicit object. In the (b) examples, an expression denoting a measure function combines with a denominator (“cost … per ton”, “per annum shortfall”) to yield an expression that is somehow equated with a specific quantity. The goal of the talk is to explain these equivalences in a way that is mathematically coherent, respecting the principle that quantities cannot be equated unless they are of the same (possibly complex) dimension (building on foundations for degree multiplication drawn from the field of quantity calculus), and in a way that captures my own intuition that in the (b) variants, division is somehow taking place at the level of the measure function. To do so, I propose that the (a) sentences are derived via a “quotient operator” analysis of “per” (one I will advocate for in my upcoming SALT contribution), and the (b) sentences are derived using two tricks: (i) a binary-Geached version of a “quotient term” analysis of “per” (one I advocated for in my contribution to last year’s SALT); and (ii) a type-shifting operation that converts units like “ton” (type d) to unit functions (type <e,d>) like “lambda x . the weight of x in tons”.

MorPhun 2/23 - Peter Grishin (MIT)

Speaker: Peter Grishin (MIT)
Title: Omnivorous agreement for third person in Algonquian
Time: Wednesday, February 23rd, 5pm - 6:30pm

Abstract: A common view is that third person is underspecified (Harley and Ritter 2002, a.m.o.): first and second person have features that third person doesn’t (e.g. [PART]), but third person doesn’t have any features that first and second lack. This kind of featural representation beautifully captures the default nature of third person: for instance, things like expletives and default agreement are invariably third person in language after language after language.

I want to show that, unfortunately, this view is untenable (see also Nevins 2007). Third person must have a feature that first and second person lack, because there is omnivorous agreement for third person. In Menominee, Innu-aimûn, and Plains Cree, the peripheral agreement suffix (standardly analyzed as a probe in C; Halle and Marantz 1993, Branigan and MacKenzie 1999, a.o.) shows the following pattern: it always agrees with the highest accessible third person DP, skipping over first and second persons. In fact, it doesn’t ever agree with first or second person DPs at all. In order to capture this behavior, the probe in C needs to be relativized to a feature that third persons have but first and second persons lack.

What should this feature be? I want to have a discussion with the audience about the ramifications this has for our feature theory. Do we just want to add an extra privative feature [3] to the standard set of privative features [PART, AUTH, ADDR]? Or do we want binary features, with third persons specified [–PART]? Additionally, if we have third person features, how do we capture the defaultness of third person crosslinguistically? I am not an expert here—y’all’s input is warmly requested. Let’s think through these issues together!

LingLunch 2/24: Madeline Bossi (UC Berkeley)

Speaker: Madeline Bossi (UC Berkeley)

Date and time: Thursday February 24, 12:30-1:50pm

Location: 32-D461, https://mit.zoom.us/j/94474505971

Title: Negative bias and pragmatic reasoning in Kipsigis belief reports

Abstract: Much work has explored how belief reports of the form x V-att p function pragmatically not just as reports of x’s internal state, but as devices for indicating the status of p with respect to the Common Ground (CG). In addition to the well-studied case of factive verbs, which presuppose p, recent work has explored negatively biased belief verbs, which suggest that p cannot or should not be added to the CG (e.g. Kierstead 2013, Hsiao 2017, Anvari et al. 2019, Glass 2020). Drawing from original fieldwork, I show that the negatively biased belief verb pɑr ‘think’ in Kipsigis (Kalenjin; Kenya) is best modeled as contributing, in addition to its basic belief semantics, an instruction for CG management: p is not to be added to the CG. Together with context-sensitive pragmatic reasoning, this instruction explains the curious case of a verb that can be used both to suggest that p is false and to remind the addressee that p is true.