Faculty Lunch Talks
Every year, Winter Storm features lunch talks by a wide variety of faculty members involved in the Language Science community. This year, talks will occur 12:15pm - 1:30pm. Free lunch is provided. Speakers and titles are listed below:
Monday 1/12: Jared Linck (CASL)
The role of cognitive control during multilingual speech production: Evidence from trilingual language switching
Recent studies of bilingual speech production suggest that multiple executive functions (EFs) contribute to the cognitive control of language production. Although some studies have provided evidence that the cognitive demands on bilingual speakers change as proficiency in the second language (L2) increases, it is unclear how the contributions of various EFs to language control vary depending on the speaker’s unique proficiency profile – particularly when more than two languages are known. In this talk, I will report data from two studies in which multilinguals performed a language switching speech production task as well as measures of three EFs (working memory, inhibitory control, and task switching). The first study included a group of less proficient multilinguals, comprised of university students who are native English speakers with intermediate proficiency in French (L2) and Spanish (L3). The second study included a group of more proficient multilinguals, comprised of native English speakers with advanced proficiency in at least two foreign languages covering a broad array of language families.
Results from the two studies indicate differential contributions of the three EFs to successful language control during the multilingual language switching task. For the less proficient multilinguals, inhibitory control and working memory (but not task switching) were related to language switch costs. For the highly proficient trilinguals, working memory was again related to switch costs but in the opposite direction, and now task switching (but not inhibitory control) was related to language switch costs. I will discuss the implications of the results for theories of multilingual language control (e.g., the Inhibitory Control Model, Green, 1998; the Adaptive Control Hypothesis, Green & Abutalebi, 2013). In particular, I will consider the relative contributions of these three EFs and how their roles in supporting language control depend on the multilingual’s language proficiency profile.
Tuesday 1/13: Sandra Gordon-Salant (HESP)
Effects of listener age on the perception of Spanish-accented English
Among the many communication challenges that older listeners experience is difficulty understanding accented English. The problem may arise from a number of factors, including age-related hearing loss, cognitive decline, and central factors such as deficits in auditory temporal processing. The hypothesis is offered that Spanish-accented English alters the timing characteristics of speech at the segmental and suprasegmental levels, rendering the bottom-up task of processing the distorted signal through the older listener’s impaired auditory system quite difficult. Results of several recent studies will be presented in which effects of age and hearing loss were prominent in the perception of Spanish-accented words and sentences, presented in quiet and noise backgrounds. The principal listener, talker, and environmental factors contributing to the observed findings will be considered.
Wednesday 1/14: Richard Prather (HDQM)
Mathematical models as a tool for bringing together biology and behavior
The use of mathematical models in scientific research is an increasingly widespread practice covering a range of disciplines. I will briefly cover background regarding the use of such models in regard to brain & cognitive sciences. The question being: has the use of mathematical models advanced our understanding of human behavior more (or equally so) than non-mathematical models? Is there a clear advantage in the use of mathematical models or are they simply one of a variety of model types available? I will use my research area of cognitive development as an example, including my own work with neurocomputational models.
Thursday 1/15: Omer Preminger (LING)
Scales & Hierarchies as a Window into the Domain-Specificity Question
This talk is about domain-specificity: the question of whether the explanation for a given linguistic phenomenon lies with general cognitive mechanisms (domain-general), or must appeal to mechanisms that are specific to language (domain-specific). We will look at linguistic phenomena found in the Kichean languages (Mayan; Guatemala) and in Zulu (Bantu; South Africa, i.a.).
It is the case that, across different languages and constructions, one finds phenomena whose analysis has been stated in terms of a “scale” or “hierarchy”: a set of linguistic categories set in particular order. For example: human >> animate >> inanimate (one of the two scales considered to regulate Differential Object Marking; see Aissen 2003, and references therein). Against this backdrop, we can pose the following question:
Does a given “scale” of this sort reflect properties of the grammar per se, or does it arise via general cognitive mechanisms?
We will first see data from Kichean that have been claimed to adhere to a scale based in “cognitive salience.” We will then juxtapose these with data from Zulu, where one finds the same combinatorics seen in Kichean, but operating over an entirely different set of operands. Crucially, “cognitive salience” does not seem to have any purchase on the operands involved in the Zulu case. Thus, insofar as we find the parallels between the Kichean and Zulu cases compelling, we have an argument against an explanation of the Kichean facts based in general cognitive mechanisms.
The larger point concerns how “scales” of this sort may be leveraged to investigate the domain-specificity of our linguistic capacities. Since such scales are often excellent candidates for a domain-general etiology (e.g. “cognitive salience”), they stack the deck against a domain-specific explanation. If even here we can show that a domain-specific explanation is indispensable, we arrive at a particularly strong argument for the domain-specificity of linguistic knowledge.
Friday 1/16: Jonathan Fritz (ISR)
Attention and the cortical representation of phonemes - an overview of animal and human studies
Since the discovery that animals can discriminate human phonemes, and even demonstrate categorical perception, there has been great interest in using animal models to explore the neural encoding of phonetic information. I will describe some recent studies, from our lab and others, on the cortical representation of phonemes. I will also describe research on the neural basis of speech representation in humans, using the tools of ecog and functional imaging. I shall also summarize research on the neural basis of attention to specific acoustic features and to single speakers in a multispeaker environment.
Monday 1/19: No sessions (MLK Day)
Tuesday 1/20: Nan Ratner (HESP)
From Freud to fMRI: language processing in people who stutter
Although historically viewed as a personality or psychiatric disturbance, it has become increasingly evident that stuttering in children and adults involves atypicalities in language skills and processing (even when overt speech is not required). I will review behavioral, electrophysiological and imaging results which suggest that stuttering can be viewed, at least in part, as a disorder of speech production. Because very few of the classic psycholinguistic paradigms have been used with people who stutter (PWS), this population offers numerous opportunities for language science students seeking to work with a clinical population on a communication disorder that is still poorly understood and extremely handicapping.
Wednesday 1/21: Ana Taboada Barber (CHSE)
United States History for Engaged Reading (USHER): A Content-Area Literacy Intervention in Social Studies
Content-Area literacy has been discussed as a framework for middle school literacy which may lead to the more advanced thinking tools used in disciplinary literacy, in high school and beyond. In this presentation I make the case for the necessity of a content-area literacy framework for middle-school struggling readers, especially English Learners (ELs), who struggle with accessing content while developing oral English proficiency. Through empirical findings from our three-year USHER project I share my perspective on the usefulness of content-area literacy as a preliminary step for disciplinary literacy in the later, high school grades.
Thursday 1/22: Marine Carpuat (CS/UMIACS)
Second Languages as Noisy Supervision for Natural Language Processing
Translations in a second language are an attractive source of supervision for learning algorithms in Statistical Natural Language Processing (NLP): translated texts are available in large quantities in many languages and can replace costly annotation by experts. In this talk, I will discuss two scenarios where NLP can be guided by second language supervision.
First, translations can be directly used as labels that disambiguate the meaning of words in context. This lets us frame word sense disambiguation as a conventional supervised learning task, but it also raises new challenges: in particular, the space of possible labels grows very large and cannot be fully observed. I will show how we can address this issue by predicting when words observed in new contexts require previously unseen labels.
Second, translations can be used to project linguistic annotation from one language to another. I will show how translations can be used to perform discourse analysis on Chinese given annotation on English only, despite important divergences in discourse organization between the two languages.
Friday 1/23: Jonathan Simon (ECE/BIOL/ISR)
Neural Representations of Continuous Speech in Auditory Cortex
An auditory scene is perceived in terms of its constituent auditory objects, which in a “Cocktail Party” scenario correspond to individual speech streams. Here, we investigate how these speech streams are individually represented in human auditory cortex, using magnetoencephalography (MEG) to record the neural responses of listeners to speech streams in a variety of auditory scenes. In the acoustically richest example, subjects selectively listen to one of two competing speakers mixed into a single channel. Individual neural representations of the speech of each speaker are observed, with each being selectively phase locked to the rhythm of the corresponding speech stream, and from which can be exclusively reconstructed the temporal envelope of that speech stream. The neural representation of the attended speech, originating in posterior auditory cortex, dominates the responses. Overall, these results indicate that concurrent auditory objects, even if spectrally overlapping and not resolvable at the auditory periphery, are indeed neurally separated and encoded individually as objects, in higher order auditory cortex.
