SATURDAY, May 19th
Location: Room W245, South College, UMass
SESSION CHAIR: Joe Pater
9:15 – 9:30: Chuck Clifton, Welcome and opening remarks
9:30 – 10:15: Jesse Harris (UCLA), Alternatives on demand and locality: Resolving discourse-linked wh-phrases in sluices
Previous studies have observed a tendency to associate the remnant (e.g., who) of ambiguous sluicing ellipsis with the closest / most local correlate (someone) in the matrix clause, as in Somebody said Fred fired someone, but I don’t know who (Frazier & Clifton, 1998; Carlson et al., 2009; Harris, 2015). I present the results of two experiments investigating the interplay between locality and the discourse status of potential correlates. The studies exploit the discourse-linking property of which-phrases in ambiguous sluiced sentences, like A teacher scolded Max or Dotty, but I can’t remember which one, to explore whether the preference for more local correlates is modulated by the discourse status of the potential correlates. I propose a discourse economy constraint (Alternatives on Demand: Avoid positing new discourse alternatives without evidence), which interacts with structural constraints like locality. Evidence from several questionnaire studies, as well as two online self-paced reading studies, supports the predictions of a sentence processing model in which the discourse status of items in memory immediately impacts the retrieval of a correlate for the remnant of sluicing ellipsis.
10:15 – 11:00: Amanda Rysling (UCSC) and John Kingston (UMass), Regressive spectral assimilation bias in ambiguous speech sound perception
When a listener comprehends speech, she must segment and recognise the phones in the acoustic signal. Segmentation requires her to bind the acoustic components associated with each phone, while recognition requires her to determine those phones’ identities. Prominent accounts hold that segmentation is performed either by identifying that a change has occurred in the signal, or by knowing how different phones affect each other and un-doing their blending. These same accounts hold that recognition occurs when listeners hear a phone as maximally different from those around it, either because of low-level auditory contrast effects, or because of the action of a dedicated compensation for coarticulation mechanism. This is circular: segmentation requires the listener to have recognised the phones, and recognition requires the listener to have segmented them. In an attempt to break out of this explanatory loop, we begin from two insights of Lyn Frazier’s work: (i) considering the incremental unfolding of a stimulus in time is key to understanding the processing of that stimulus (ii) the action of a parsing system under conditions of ambiguity reveals that system’s default behavior.
We present the results of a series of speech sound categorisation studies in which listeners were faced with ambiguity about the identity of the first of two successive phones. In these contexts, listeners productively heard the first sound as spectrally similar to the second sound in a manner suggesting that they construe the transitions between the two as evidence about the identity of the first. Listeners seem to default to construing the acoustic properties of the input as evidence about the phone they have already begun processing, rather than positing a new phone. Moreover, they do so until they encounter acoustics that are clearly inconsistent with that first phone. These effects go unaccounted for in the two prominent models of speech perception. Given parallels between this effect and several known domain-general effects in perceptual processing, we argue that this default is likely a consequence of the structure of the human auditory system. If this physiological basis is correct, then we can account for the predominance of regressive place of articulation assimilation in the world’s languages by appealing to a perceptual predisposition rather than a grammatical one.
11:00 – 11:15: Coffee Break
SESSION CHAIR: John Kingston
11:15 – 12:00: Amy Schafer (Hawai’i at M?noa), Amber Camp (Hawai’i at M?noa), Hannah Rohde (Edinburgh), and Theres Grüter (Hawai’i at M?noa), Contrastive prosody and the subsequent mention of alternatives during discourse processing
Linguistic research has long viewed prosody as an important indicator of information structure in intonationally rich languages like English. Correspondingly, numerous psycholinguistic studies have shown significant effects of prosody, particularly with respect to the immediate processing of a prosodically prominent phrase. Although referential decisions are known to be influenced by information structure, it has been less clear whether prosodic prominence can affect decisions about next mention in a discourse, and if so, how. We present results from an open-ended story continuation task, conducted as part of a series of experiments that examine how prosody influences the anticipation and resolution of co-reference. Overall results from the project suggest that prosodic prominence can increase or decrease reference to a saliently pitch-accented phrase, depending on additional circumstances of the referential decision. We argue that an adequate account of prosody’s role in co-reference requires consideration of how the processing system interfaces with multiple levels of linguistic representation.
12:00 – 12:45: Shayne Sloggett, (Northwestern), Logophlexivity: When refelxives behave like logophoric pronouns
12:45 – 2:45: Catered lunch and poster session, South College Atrium
SESSION CHAIR: Peggy Speas
2:45 – 3:30: Brian Dillon, Christopher Hammerly, and Adrian Staub (UMass Amherst), Under no illusion: Reëvaluating illusions of grammaticality
Just as visual illusions inform theories of how the visual system parses its input, ‘grammatical illusions’ help us understand the mechanisms that comprehenders use to parse and interpret linguistic input. Most commonly, grammatical illusions are taken to reflect ‘errors’ that arise due to noise inherent in the cognitive systems that are used to assemble grammatical structure in real time comprehension (Phillips, Wagers & Lau, 2011). For example, agreement attraction effects in sentences like ‘the key to the cabinets are rusty’ have been argued to reflect memory retrieval errors that arise when the parser searches for an agreement controller to license a dependent verb (Wagers, Lau & Phillips, 2009). In this talk we re-evaluate this broad perspective. I will suggest instead that grammatical illusions should not, in general, be characterized as processing ‘errors.’ To make this case, we will present evidence in favor of a representational account of agreement attraction. These findings suggest that grammatical illusions are not ‘errors’ or ‘illusions’ attributed to noisy memory mechanisms; instead, they are rooted in the syntactic representations that are created to support parsing and generation.
3:30 – 4:15: Katy Carlson (Morehead State), Parallelism and syntactic structure in ellipsis with and without “and”
A set of written and auditory studies explored the processing of ellipsis sentences with an embedded infinitive clause, as in Mika wanted to bake cupcakes, and Leah did too. In VP Ellipsis using the conjunction and as well as Gapping, the unavailability of a lower coordination structure led to almost universal choice of the higher interpretation. In comparatives and VP Ellipsis using subordinating conjunctions, though, there was a preference for the lower verb due to lower attachment being possible. When the VP Ellipsis sentences were modified to have a finite embedded clause, allowing two different possible structures (e.g., Mika hoped that Dad baked muffins, and/before Leah did (too)), both interpretations became possible. Interestingly, only the VPE with and showed an effect of prosodic parallelism (matching accents on contrastive subject pairs, like Mika/Leah or Dad/Leah), although parallelism has been shown to be active in a range of other ellipsis types not involving and.
4:15 – 5:00: Ted Gibson (MIT), Meaning explanations of two syntactic islands: Subject islands and Embedded-clause islands
Syntacticians have long proposed (a) that certain extractions from embedded positions are universally ungrammatical across languages and (b) that their ungrammaticality is not explainable in terms of meaning. These two ideas together imply the existence of syntactic universals in Chomsky’s Universal Grammar (e.g., Ross, 1967; Chomsky, 1973; and many more recent studies including Sprouse et al (2012) and Sprouse et al (2016). In this talk, I will present data from studies of two different kinds of syntactic islands that strongly suggest meaning explanations for both, without any need for syntactic universals. First, I report collaborative work with Anne Abeille, Barbara Hemforth and Elodie Winckel (CNRS, Paris), where we show that extractions out of subject position are actually easier to process than extractions from object position, in both English and French relative clauses, contrary to the claim of the universality of a so-called “subject island”. Second, I report collaborative work led by Yingtong Liu, in collaboration with Rachel Ryskin and Richard Futrell, in which we show that the grammaticality of extractions from embedded clauses as in (1)-(3) is best explained in terms of simple verb subcategorization frequency of the S-complement verb.
- “bridge” verb extractions: Who did Mary think / say that Bill saw _?
- “factive” verb extractions: ?* Who did Mary know / realize that Bill saw _?
- “manner” verb extractions: ?* Who did Mary mumble / stammer that Bill saw _?
We propose that the difficulty of these extractions is simply due to their plausibility in experience: the “bad” ones are just weird events. We can see the same effects in declarative versions; the extra bad ratings of the extracted versions are just scaled extra bad versions. And in context, the extracted versions get much better. The simple meaning-based explanation accounts for the ratings for extracted and unextracted versions, in and out of context. Furthermore, the interactions that others have observed (and that we observe) with respect to rating these types of materials (declarative / wh-question x easy / hard) is probably due to scaling issues in the acceptability scale: ratings are compressed towards the “good” end of the scale. Overall, we propose, following Ivan Sag and others, that perhaps all “islands” are meaning- and memory-based, contrary to the UG syntax claim.
Sunday, May 20th
Location: Room W245, South College, UMass
SESSION CHAIR: Brian Dillon
9:30 – 10:15: Adrian Staub (UMass Amherst), Avoiding filler-gap analyses
The Minimal Chain Principle (De Vincenzi, 1991) proposed that the parser follows two (related) principles when resolving ambiguities that can arise in processing filler-gap structures: “Avoid postulating unnecessary chain members at S-structure, but do not delay required chain members.” The second clause of the MCP has been investigated in great detail, under the heading of active gap-filling. Much less attention has been paid to the first clause. In the present talk, I will review both older and newer evidence favoring a strong version of the claim that the parser avoids an analysis that involves a filler-gap dependency when an alternative analysis is present. Both gap-first analyses (i.e., Heavy NP Shift) and filler-first analyses (i.e., relative clauses) are avoided, and neither structural frequency nor lexical bias appears to override this preference.
10:15 – 11:00: Markus Bader (Goethe University Frankfurt), The division of labor between structure building and feature checking
The paper will review evidence from syntactic ambiguity resolution and grammatical illusions addressing the proper place of syntactic features (case and number features) within a model of the human parsing mechanism. The paper will conclude that features do not guide phrase-structure building, but are used at a later stage of processing for evaluating and revising phrase-structural representations.
11:00 – 11:15: Coffee Break
SESSION CHAIR: Tom Roeper
11:15 – 12:00: Meg Grant (Humboldt University of Berlin) with collaborators Sonia Michniewicz and Jessica Rett, Real-time commitments in processing individual/degree polysemy
The field of semantic processing has had a focus on understanding when semantic commitments are made during real-time processing. In this talk, we examine the processing of individual/degree polysemy, which can be used to address this line of inquiry. Individual/degree polysemy is a phenomenon in which individual-denoting Determiner Phrases of any type can, in certain contexts, denote a degree corresponding to some salient measure of that individual (Rett 2014). For example, in certain contexts a Determiner Phrase such as the pizzas may be used to denote an amount of pizzas, rather than the salient plurality of pizzas. Like deferred reference (Nunberg 1995), individual/degree polysemy feeds agreement: compare Four pizzas are vegetarian to Four pizzas is more than Sue had asked for. We discuss the empirical phenomenon of individual/degree polysemy, and then report the results of one study of eye movements during reading and one study of self-paced reading. These experiments test whether there is immediate commitment to either an individual or a degree interpretation during on-line processing. Our results provide some evidence for immediate commitments, although the direction of the commitment depends on the internal properties of the Determiner Phrase (e.g., plain definites such as the pizzas versus number DPs like four pizzas).
12:00 – 12:45: Mako Hirotani (Carleton University), Neural substrates for conversation: A hyper-scanning functional MRI study
A hyper-scanning fMRI study was carried out, with the aim of identifying conversational features that lead to two speakers’ building a “shared” mental model and the neural substrates supporting the processes involved. Methods: Inside MRI scanners, 22 pairs of Japanese speaking participants played collaborative maze games similar to those in Garrod & Anderson (1987). Players played the maze game with the same players for the first three games and with new players for the last game. This allowed us to explore how mental model were built with the same partners (partner-specific effect) vs. with a new partner (general learning effect). Results: A mental model for discourse was likely established as the participants played the maze games. This is supported by the increased activation of the theory of mind network consisting of Dorsal Medial Pre-frontal Cortex and Precuneus/Posterior Cingulate Cortex, as well as the corpus data. The corpus analyses showed the overlap of the consecutive utterances by paired players. The partner-specific activation of the right Cerebellum and the left BA6/BA44 likely played a critical role in predicting the timing of players’ utterance turns. This effect was partner-specific probably because it became easier to attend to thesame partners and predict the timing of their utterance turns, compared to the unfamiliar partners. Further discussion on the results of the corpus data will be provided.
12:45 – 2:45: Lunch (on your own)
SESSION CHAIR: Rajesh Bhatt
2:45 – 3:30: Michael Dickey, Tessa Warren, and Michelle Colvin (University of Pittsburgh), Event knowledge and verb knowledge predict sensitivity to different aspects of semantic anomalies in aphasia
Neurotypical comprehenders show exquisite sensitivity to violations of the coarse-grained semantic restrictions that verbs place on their arguments (e.g., Dowty, 1991) — see Warren & McConnell (2007), Warren et al. (2015). However, they also show rapid sensitivity to violations of world knowledge regarding prototypical event participants (McRae & Matsuki, 2009), and many have argued that verb-related semantic constraints cannot meaningfully be distinguished from world knowledge about events (e.g., Jackendoff, 2002). We present evidence from aphasia suggesting that these two types of knowledge may be dissociated during real-time comprehension of verb-argument combinations. Specifically, degree of impairment in verb processing and in access to event knowledge predict different aspects of reading-time responses to anomalous verb-argument combinations by people with aphasia. These findings provide neuropsychological evidence that coarse-grained verb constraints cannot be reduced to fine-grained knowledge about events. Rather, these grammaticized abstractions over event-related experience may be differentially impaired, and they exert their own separable influence on real-time processing.
3:30 – 4:15: Florian Schwarz (University of Pennsylvania), Definites, Domain Restriction, and Discourse Structure in Online Processing
Definite descriptions are commonly assumed to be associated with a uniqueness requirement, which has to be evaluated relative to a contextually restricted domain. While there are a number of quite detailed theoretical proposals for modelling such effects, less is known about how interpreters choose a suitable domain in comprehension, and how domain selection unfolds in online processing. I report two visual world studies investigating these issues. The results show that domain restriction effects are reflected rapidly in eye movement patterns, and are shaped by the discourse structure created by preceding text. These findings are related to the theoretical options for modelling domain restriction in the literature.
4:15 – 5:00: Barbara Hemforth (Université Paris Diderot) and Lars Konieczny (University of Freiburg), All linguists did not go to the workshop; none of the Germans but some of the French did: The role of alternative constructions for quantifier scope
In this talk, we will present crosslinguistic data on the interpretation of negation over quantifier scope in sentences like “All children did not go to the zoo.” Questionnaire data show that English as well as German speakers prefer a linear scope interpretation of the quantifier and the negation, where it is true for all children that they did not go to the zoo. French speakers, however, strongly prefer the inverse scope interpretation where some but not all children did not go to the zoo. The preference for linear scope is moreover stronger for German speakers than for English speakers. It diminishes with age for French and English, but not for German speakers. We will argue that language differences result from two constraints: the availability of a “close” alternative in the language and the topicalty of a preverbal subject. An unambiguous alternative corresponding to inverse scope in the “all-not” construction can easily be achieved in English and German by fronting the negation as in “Not all children went to the zoo”. The corresponding construction is not available in standard French. The particularly strong preference for linear scope in German will be argued to be linked to the stronger topicality of preverbal subjects in German main clauses.