McGill Linguistics

01/07/2024

McGill linguists of past and present at the LSA

11/28/2022

This Friday we are having the last colloquium of this semester! The talk will be given by Dr. Alan Bale (Concordia University) on Friday, December 2nd at 3:30pm in the Education building, room 624.

Title: Ignorance and Distraction: A case study that uses experimental methodologies to help distinguish between different theories of inference.


Abstract: Ever since Grice (1975), there has been a vibrant literature on the nature of quantity implicatures (e.g., inferences from “some” to “not all” as well as inferences from “some” to “I don’t know whether all”). Some authors have defended the spirit of Grice’s original proposal, namely that these types of inferences stem from people rationally reasoning about basic principles of conversational cooperation (for a recent defence, see Geurts 2010). Others have amended Grice’s original proposal with various grammatical interventions, special algorithms, and/or default heuristics that sometimes run counter to rational reasoning (e.g., Levinson 2000). Still others have proposed that at least some types of quantity implicatures are purely grammatical and do not involve any type of strengthening outside of the linguistic faculty (e.g., Chierchia, Fox and Spector 2012; Meyer 2013).


In this talk, I will discuss how inducing states of ignorance and employing methods of distraction can allow us to gather experimental evidence that might be able to distinguish between these various theories. In particular, I will argue that the cumulative results from several recent studies, including those from an experiment conducted in my lab, are inconsistent with the grammatical account of quantity implicatures. Rather, such results are more consistent with Grice’s original proposal (with some relatively minor amendments).

11/06/2022

Our next colloquium in the series is happening this Friday! The talk will be given by Dr. Kyle Johnson (University of Massachusetts Amherst) on Friday, November 11th at 3:30pm at the Education building, room 624.

Title: Implicit arguments as Incorporated Theta-roles

Abstract:
Some verbs are capable of being used without an expression of their arguments. The direct and indirect objects of eat and throw are standard examples.
(1) a. Marlys ate cake. Marlys ate.
b. Marlys threw the ball to Sam. Marlys threw the ball.
The meanings of eat and throw preserve the θ-roles that cake and to Sam bear, even when those arguments are not present. Those θ-roles are understood to be existentially closed. They are said to be implicit when this happens. The ability for a θ-role to be implicit seems to be idiosyncratically controlled by the verb, but it does not extend to external arguments. To make an external θ-role implicit, a valency changing operation is required. An external/internal argument contrast of this sort is also found in many kinds of Noun Incorporation constructions. And the lexically idiosyncratic nature of making a θ-role implicit also seems to be a feature of some Noun Incorporation constructions. Martí (2015) argues that the syntax and semantics of Noun Incorporation underlies making a θ-role implicit. I will pursue that thesis in this talk. I will suggest that we should think of θ-roles as being kinds of nominals, and sketch a syntax that makes sense of that idea. One of its consequences is that θ-roles can undergo Incorporation, and this is how implicit arguments are achieved.

10/25/2022

*EDIT* Information about the room is added

Please join us Friday, October 28th at 3:30pm for our next talk of the 2021-2022 McGill Linguistics Colloquium Series.

The title and abstract for next week’s colloquium, which will be given by Beth MacLeod (Carleton University), are provided below. Room: Education 624

If you would like to meet with Beth before the talk on Friday, please email Connie Ting with your availability if you haven’t already.

Title: The other piece of the imitation puzzle: individual variation in the perception of phonetic imitation

Abstract: Phonetic imitation occurs when, during an interaction, a speaker’s pronunciation shifts to become more like that of the person to whom they are speaking. According to Communication Accommodation Theory (e.g., Giles, 1973), speakers imitate their interlocutors to minimize social distance between themselves and their interlocutors and to influence the interlocutor’s impression of them. Previous work has found that when speakers imitate, they and the interaction are evaluated more positively than when no imitation takes place (Chartrand & Bargh, 1999; Giles & Smith, 1979; Street, 1982). In order for this positive evaluation to occur, however, the interlocutor must perceive the imitation. While many studies on the production of imitation use a perceptual method (often an AXB task) to assess how much the talkers have imitated (e.g., Dias & Rosenblum, 2016; Goldinger, 1998), virtually no study has focused on the behavior of the listeners and whether they might vary in their ability to perceive imitation. If listeners do vary in this ability, it might predict that they would also vary in their ability to access social cues, both in their own conversations and in conversations they observe between others (e.g., Dias et al., 2021; Giles et al., 1991; Pardo et al., 2012; Shepard et al., 2001). As such, the perception of imitation is a critical piece of the puzzle in understanding the social motivations and consequences of imitation.

In this talk, I discuss my recent research focusing on the variability and consistency of listener performance in an AXB assessment of phonetic imitation. The results suggest that individuals differ in their ability to perceive imitation in this task, but that this variation reflects stable characteristics of the individual listener, rather than random fluctuations. I also discuss how this research will set the stage for future work exploring an individual-level connection between the perception and production of imitation.

02/17/2022

if you missed the public lecture by Anne Harper Charity Hudley, you can find the recording here:

On February 11th, as part of the programming for Black History Month, Dr. Anne H. Charity Hudley (Stanford University), gave a virtual public lecture titled ...

02/14/2022

Please join us Friday, February 18th at 3:30pm for our next talk of the 2021-2022 McGill Linguistics Colloquium Series. If you are planning to attend talks and have not yet registered, you can do so here (you only need to register once for the 2021-2022 year). After registering, you will receive a confirmation email containing information about joining the meeting.

Speaker: Gillian Ramchand (University of Tromsø)

Title: Verbal Symbols and Demonstrations Across Modalities

Abstract:

The evidence from syntax and morphology suggests that the extended projection of the verb is divided into typologically robustly attested zones (see Ramchand and Svenonius 2014). The lowest part of each extended projection (whether nominal or verbal) is associated with open class lexical items, and the meanings of those items are notoriously hard to define and integrate into a compositional treatment of meaning without rampant allosemy. In this talk I will argue that the truth conditional, or denotational view of lexical contents is misguided and that we should move to a more mentalistic view of the semantic content of open class lexical items. This will involve reifying the symbol and the demonstrative act of reference as a way of mediating between mentalistic contents and assertions which have worldly truthmakers. I show that the neo-quotational view of compositional semantics (as I will call it),allows for a more satisfying account of iconic and gestural meaning, and gives a new perspective on indexical shift.

02/06/2022

Public lecture February 11th! registration: https://mcgill.zoom.us/meeting/register/tZMocumqpjksGdIU0fEPrRE7IQpQRtng3Y8J

https://reporter.mcgill.ca/anne-h-charity-hudley-on-linguistic-justice/

On February 11, noted linguist will deliver virtual lecture titled "A Model for Linguistics Reparations” as part of Black History Month

01/17/2022

The next talk of the 2021-2022 McGill Linguistics Colloquium Series will be held on Friday, January 21st at 3:30pm. The talk will be given by Maribel Romero (University of Konstanz).

The title of the talk is "A unified semantic analysis of the Q-Particle in Sinhala across interrogative types". The abstract is below.

"In a wide variety of languages, Q(uestion)-particles are –optionally or mandatorily– used in the formation of some interrogative clause types(see e.g. Kamali 2015 for Turkish mI; Rudin et al. 1999 for Macedonian li; Hagstrom 1998 for Japanese ka; Kishimoto 2005, Cable 2010 and Slade 2011 for Sinhala də). The present talk concentrates on Sinhala, in which the Qparticle də appears in wh-questions (WhQs), alternative questions (AltQs) and polar questions (PolQs), as illustrated in (1)-(3):

(1) Chitra monəwa də gatte WhQ
Chitra what də bought.E
`What did Chitra buy?' [Slade 2011: (2) p. 19]
(2) oyaa maalu.də mas.də kanne? AltQ
you fish.də meat.də eat.E
`Did you eat meat or fish?’ [Weerasooriya 2019: (36) p. 12]
(3) Chitra [ee potə]F də kieuwe? PolQ- narrow
Chitra that book də read.E
`Was it that book that Chitra read?' [Kishimoto 2005: (21a) p. 11]

A prominent line of analysis, developed by Cable (2010) for WhQs and by Slade (2011) for AltQs and PolQs, posits that the Q-particle də introduces a choice function variable coindexed with the question operator in all three interrogative clause types.
Against this background, the goal of our talk is two-fold. First, we present novel data on the distribution of the Q-particle də in interrogatives containing islands which challenge the Cable-Slade choice function analysis. We will see that, while də in AltQs patterns like də in WhQs in being sensitive to islands, də is PolQs is island-insensitive. Second, we develop a first unified semantic analysis of the Q-particle də that accounts for its distribution in all three question types and for the disparity in island-(in)sensitivity between WhQs/AltQs and PolQs."

If you have not yet registered for the colloquium series, please contact Jing ( [email protected]) You only need to register once for the 2021-2022 year.

11/22/2021

Please join us Friday, November 26th at 3:30pm for our next talk of the 2021-2022 McGill Linguistics Colloquium Series.

Speaker: Ewan Dunbar (University of Toronto)

Title: Probing state-of-the-art speech representation models using experimental speech perception data from human listeners

Abstract:

The strong performance of neural network natural language processing has led to an explosion of research probing systems' linguistic knowledge (whether language models implicitly learn syntactic hierarchy, whether word embeddings understand quantifiers, and so on), in order to understand if the data-crunching power of these models can be harnessed as the basis for serious, theoretically-grounded models of grammatical learning and processing. Much of this "(psycho)linguistics for robots" work has focussed on textual models. Here, I show how we have applied this same approach to phonetics. In particular, we probe state-of-the-art unsupervised speech processing models and compare their behaviour to humans' in order to shed light on the traditionally hazy and ad hoc construct of "acoustic distance."

On the basis of a series of simple, broad-coverage speech perception experiments run on English- and French-speaking participants, I compare human listeners' behaviour (how well they discriminate sounds in the experiment) to the "behaviour" of representations (how well they separate those same stimuli) which come from models trained with the express purpose of building better representations to be used in automatic speech recognition. For example, Facebook AI's recent wav2vec 2.0 model takes large amounts of unlabelled speech as training data, and learns to extract a representation of the audio that is highly predictive of the surrounding context; it has now proven extraordinarily useful for replacing off-the-shelf audio features, to the point that some of the best-performing speech recognition systems today have switched to using these representations, which has substantially reduced the amount of labelled data needed to train high-quality speech recognizers.

We use the comparison with human behaviour to show that, for this and related systems, contrary to what many researchers may have *thought* these systems are doing, they are not really "learning representations of the sound inventory" of the training language, so much as learning good representations of the acoustics of speech - so good that they are very good models of "auditory distance" in human speech processing, but, notably, they lack the categorical effects on speech perception which are pervasive in human listening experiments, and they only show very weak effects of the language on which they are trained, unlike our human listeners. As well, I present new evidence that "speech is special" in human auditory processing, by comparing learned representations trained on speech data to the same models, trained on non-speech data. We show that representations trained on non-speech are very (very) poor predictors of human speech perception behaviour in experiments.