Fall 2013
August 30 - No CLS Meeting
September 6 - Melinda Fricke (UC Berkeley, Department of Linguistics) : Language in Time: Phonetic Duration and the Sequential Nature of Phonological Encoding
Many studies have shown that word duration is reliably correlated with contextual probability in conversational speech (Jurafsky et al., 2001; Bell et al., 2003, 2009; Aylett & Turk, 2004, among others). Most explanations of this correlation focus on the lexical level of representation; the predictability of a word in a given context is typically hypothesized to affect its phonetic realization, either through listener-driven or speaker-driven processes. Listener-driven accounts focus on the amount of phonetic detail needed to recognize a word in its context, while speaker-driven accounts posit that the speed of lexical access has consequences for a word’s phonetic realization.
It seems intuitive that faster lexical access would be related to shorter word duration, but one aspect missing from existing speaker-driven accounts is an explicit mechanism linking higher contextual predictability to shorter articulatory duration. In this talk, I present three studies that help to lay the groundwork for such a mechanism. Results from a word-learning study with children indicate that difficulty in phonological encoding is reliably associated with longer articulatory duration, and results from studies of adult single-word production and conversational, connected speech are consistent with a model of language production that incorporates both lexical-phonological feedback and sequential encoding of segments (e.g. Sevald & Dell, 1994). I argue that the speed of phonological encoding, combined with the fact that encoding proceeds from left to right, can account for the present results and may provide the link between contextual predictability and articulatory duration.
September 13 - Ashley Roccamo (Penn State, Department of German) : Is Earlier Really Better? Comparing the Effectiveness of Pronunciation Training for Beginner and Intermediate Learners
Pronunciation in a second language (L2) is notoriously difficult to acquire. Even advanced L2 speakers often cannot acquire accurate pronunciation on their own (Grosser, 1997; Jilka, 1999; Munro & Derwing, 2008; Trofimovich & Baker, 2006). Despite its consequences for communication, however, L2 learners and teachers alike frequently ignore pronunciation during the stages of L2 acquisition. Yet pronunciation is a skill that can improve with focused training, as has been reported in a number of studies (Derwing, Munro & Wiebe, 1998; Elliott, 1997; Flege, 1989; Hardison, 2004; Saito & Lyster, 2011). Most of these training programs are introduced in more advanced stages of L2 proficiency (e.g., Counselman, 2010; Elliott, 1995, 1997; Lord, 2008), although a number of researchers have recently been calling for pronunciation training to begin as early as possible (e.g., Counselman, 2010; Derwing & Munro, 2013; Elliott, 1995, 1997; Eskenazi, 1999; Hardison, 2004; Neufeld & Schneiderman, 1980). Thus the question remains, at which stage in L2 acquisition a focus on pronunciation should begin.
The current study examines this issue further and tests whether an eight-week pronunciation training unit designed as a supplement to German language classrooms is more effective when implemented in the first or fourth semester. Training for the two experimental groups was divided into four two-week modules, each training a specific area of German pronunciation for just 10 minutes a day. Target areas were chosen both for their difficulty for L2 learners as well as their importance for communicating meaning, and included lexical stress, palatal and velar fricatives ([ç] and [x]), fricative and vocalized /r/, and the monophthongization of [e] and [o]. Identical pre- and post-tests were administered for all groups before and immediately after training. Randomly selected, matched samples from each participant’s pre- and posttests were rated by native-speaker listeners for accentedness and comprehensibility. Pre- and posttest ratings from the experimental and control groups are compared in order to compare the effectiveness of pronunciation training in early and intermediate semesters of German L2 learning. The results of these analyses will ascertain whether pronunciation training truly is more effective for beginner learners in the earliest stages of L2 acquisition.
September 20 - Merel Keijzer (University of Utrecht) : Cognitive and Language Control in Aging Dutch-English Late Bilinguals and Dutch Bilectals
Following the seminal work by Ellen Bialystok, the last decade has seen a host of studies on cognitive advantages as a result of early bilingualism in which early bilinguals are consistently reported as having enhanced cognitive control, notably inhibitory control (Fiszer, 2008). More recent work has sought to extend this effect to late bilinguals (cf. Fiszer, 2008; Tao et al., 2011; Luk et al., 2011), but with mixed results. This may be partly due to the large variability in exposure time to the L2, and, related to that, the lack of L1 and L2 proficiency measures. At the same time, the variable results are also – partially – ascribed to a difference in typological proximity of a bilingual’s languages under investigation. In that vein, questions have been raised concerning bilectals and whether the cognitive control advantage in these speakers is likely to be greater (due to increased demands on cognitive control with two closely related languages that have to be juggled) or smaller (due perhaps to the dialect not being perceived as a separate language cognitively).
The very few studies that have been done on the language and cognitive control of bilectals (cf. Kirk, 2012) have not been able to shed light on this matter. The aim of this study is to add to the growing research tradition of examining language and cognitive control in late bilinguals and well as bilectals. In order to do that four groups of speakers were examined: 1) Late bilinguals (L1 Dutch; L2 English) who were at least 15 when they started acquiring their L2. At the time of testing they had been immersed in an Anglophone environment for 42,6 years on average (n=69). 2) Bilectals (standard Dutch – Nedersaksisch, which is an eastern Dutch dialect spoken close to the German border)) (n=20). 3) Monolingual Dutch speakers were recruited to serve as controls (n = 60). 4) Since it has been pointed out that no true Dutch monolinguals exist with all Netherlands-based subjects having at least a basic command of foreign languages, an additional group of monolingual English speakers in Australia was recruited (n=60).
All participants were roughly subdivided over three age categories: a baseline group of middle-aged subjects (40-50), a ‘youngest old’ category (60-70) and an ‘oldest old’ group (71+). One exception was the bilectal group, which was solely made up of oldest old participants, i.e. aged 71 and older. Furthermore, participants in all groups were asked to fill in a (language) background and lifestyle questionnaire, and were subjected to various executive functioning (EF) and working memory (WM) measures in addition to standard processing speed and fluid IQ tests. Crucially, they completed various language proficiency tests in both their L1 and L2, varying from overall proficiency measures to a grammaticality judgment and receptive as well as productive vocabulary tests.
The results indicate a very large degree of variability in both cognitive and bilingual language control, in line with previous findings (Bedard et al., 2002; Borella et al., 2008). In addition, language and cognitive measures were correlated but this did not apply to all tests included in the battery. Predictor variables of success in control on both sides were L1 and L2 proficiency and educational level. Finally, on many occasions, the oldest old participants outperformed their younger old peers. Comparing the bilinguals and the bilectals, the two groups were found to be very distinct: the bilinguals were found to have a cognitive advantage (on some but not all the tasks of the test battery), whereas the bilectals performed very similar to the monolingual Dutch speakers with no cognitive advantage presenting itself. This itself leads to interesting questions regarding the cognitive status of a dialect. The results of a recently initiated continuous analysis, abandoning the idea of groups, but rather include all data in one larger regression model, will also be presented.
September 27 - Adele Goldberg (Princeton, Psychology) : Explain me something: how we learn what not to say
Although many constraints are motivated by general semantic or syntactic facts, in certain cases, formulations are semantically sensible and syntactically well-formed, and yet noticeably dispreferred (e.g., ??She explained me something; ??the afraid boy). Results from several experiments are reviewed that suggest that competition in context—statistical preemption--plays a key role in learning what not to say in these cases. I will also suggest a domain-general mechanism that may well underlie this process, and offer a speculative proposal as to why L2 learners may have more trouble avoiding these dispreferred utterances.
October 4 - MaryEllen MacDonald (UW-Madison, Psychology) : We Reap What We Sow: The Cascading Effects of Language Production on the Nature of Language and its Comprehension
Two critical questions in language research concerns why languages have the form that they have, which typically is a concern of language typologists and other linguists, and and why language comprehension works the way that it does, a major concern of psycholinguists. I will argue that we can find some important answers to both of these questions by investigating the central question of language production researchers: why people say things in certain ways and not others. Language production processes involve the conversion of communicative intents into spoken, signed, or written utterances, and this conversion requires an intricate interplay of attention, retrieval from long term memory, motor planning and temporary maintenance of an utterance plan. I'll present evidence that the nature of these processes (many of which are not language-specific) yield biases in production planning for certain kinds of word orders over others. These biases to produce certain utterance forms over other viable alternatives have downstream consequences for language statistics, language typology, and language comprehension processes. I'll trace these cascading effects from production biases and argue that a full account of comprehension and language typology will have to incorporate insights from language production.
October 11 - David Green (University College London, Division of Psychology and Language Sciences) : Language control
Communities differ in the way they use the languages at their disposal. In some, speakers code-switch between their languages within a conversational turn whereas in others they do not. Instead different languages are used in different realms of life. Might these different conversational demands shape the processes of language control within the individual speaker? I will explore and evaluate this possibility. Understanding the dynamics of language control contributes to our understanding of the mind as a control system, to our understanding of individual differences in executive control and to our understanding the patterns of speech recovery post-stroke.
October 18 - Fred Genesee (McGill University, Psychology) : Myths and Misunderstandings about Dual Language Acquisition in Young Learners
There has been growing interest in children who learn language in diverse contexts and under diverse circumstances. In particular, dual language acquisition has become the focus of much research attention, arguably as a reflection of the growing awareness that dual language learning is common in children. A deeper understanding of dual language learning under different circumstances is important to ensure the formulation of theories of language learning that encompass all language learners and to provide critical information for clinical and other practical decisions that touch the lives of all language learners. This talk will review research findings on dual language learning in both school and non-school settings, among simultaneous and sequential bilinguals, and in typically-developing learners and those with an impaired capacity for language learning. Key findings with respect to common myths and misunderstandings that surround dual language acquisition in young learners will be reviewed and discussed and their implications for both theoretical and practical matters will be considered.
October 25 - Ben Zinszer (Penn State, Psychology) : Age of L2 Onset and Left MTG Specialization for L1 Lexical Tones
Recent neuroimaging studies have revealed distinct functional roles of left and right temporal lobe structures in the processing of lexical tones in Chinese. In an evoked response potential paradigm, Xi et al (2010) elicited a greater mis-match negativity (MMN) to acoustically varying sets of intonated Chinese syllables when tone variants represented linguistically salient contrasts. Zhang et al (2011) further localized processing of Chinese lexical tones to the left middle temporal gyrus (lMTG) for linguistic processing and the right superior temporal gyrus (rSTG) for acoustic processing of tonal contrasts for native Chinese speakers. In the present study, we ask whether knowledge of a second language (English) modulates this pattern of activation in the perception of tonal contrasts. Twenty-five native Chinese speakers were recruited from undergraduate and graduate students at Beijing Normal University, China. Participants watched a silent film and listened to blocks of computationally manipulated /ba/ syllables which were varied to form within- and between-category deviants at equal acoustic intervals from a standard tone. Oxygenated hemoglobin levels in participants' temporal cortices were measured by functional near-infrared spectroscopy (fNIRS). Three block conditions were presented: standard falling tones, standard falling tones randomly ordered with deviant falling tones (Within Category), and standard falling tones randomly ordered with rising tones (Across Category). Deviant blocks were alternated 10 times each with standard blocks and rest periods of equal duration. Blocks were analyzed for peak oxygenated hemoglobin levels, and a mixed-effects model was fit to these data, including effects of Category (Standard, Within, or Across), age of earliest exposure to English (spoken), and proficiency in English. Functional changes in oxygenated hemoglobin levels indicated a significantly greater response to Within Category contrasts in right STG, consistent with previous findings. However, the effect of Category in left MTG was significantly modulated by the age of participants' earliest English exposure: Across Category activation exceeded Within Category activation only for participants exposed to English after 13 years of age. While previous research has established the importance of left MTG in the categorical perception of lexical tones, our findings suggest that the functional specialization of this region is sensitive to second language experience, even in the processing of native language.
November 1 - PIRE undergraduate presentations
November 8 - Greg Guy (NYU, Linguistics) : The Role of Lexical Frequency in Linguistic Variation
Lexical frequency has been argued (by, for example, Bybee 2007, Pierrehumbert 2001) to be a significant factor in the organization of linguistic structure and the operation of phonological processes, such that words that are used a lot are expected to behave differently in certain respects from those that occur more rarely. Variable processes in language are especially implicated in such models. This paper considers empirical evidence from a number of studies of linguistic variation that have looked at lexical frequency as a potential predictor, to investigate the validity and extent of frequency effects. The results are mixed: frequency does appear to affect some variable lenition processes, like -t,d deletion in English, but is unreliable in others, e.g. –s deletion in Spanish. It interacts significantly with morphological constraints: -t,d deletion increases with frequency in monomorphemes, but not in derived forms (regular past tense forms). And frequency effects can be found beyond the phonology: in morphological variation, for example, synthetic comparatives and superlatives in English are favored with frequent adjectival roots. In syntax, Spanish pro-drop shows no general frequency effect, but frequency systematically interacts with all other constraints on the process, magnifying them in high-frequency forms and attenuating them in low frequency forms. And there are cases where frequency fails to have any significant effect, or makes the wrong predictions. The data therefore suggest that frequency is a characteristic of entries in the mental lexicon that is available to speakers for constructing generalizations about linguistic operations, but frequency does not regularly predict or entail specific linguistic outcomes. This has implications for formal linguistic theories as well as for usage-based theories that emphasize frequency as an explanatory factor.
November 15 - No CLS Meeting
November 22 - Tamara Swaab (UC Davis, Psychology) : Understanding Individual Differences in Language Comprehension
Spoken language comprehension involves managing a set of interrelated cognitive tasks, including activation of stored phonological and semantic representations of words, activation or construction of syntactic structure representations, determination of how newly activated words relate to previously introduced information, and ultimately the construction of a representation of the meaning of the message. Whereas the processing of individual words and syntactic structures in isolation can proceed relatively automatically, the construction of a coherent representation of the overall message may rely more on controlled processing, requiring maintenance of previous context and rapid integration of incoming input in Working Memory (WM). This is especially the case for spoken language comprehension since listeners have no control over the rate of input, nor can they ‘‘re-experience’’ parts of the speech signal. I will present evidence from healthy adults and schizophrenia patients indicating that individual differences or impairments in the controlled maintenance of context predict which kind of language information is prioritized or processed during spoken language comprehension: the meanings of individual words or the integrated representation of the language context.
November 29 - No CLS Meeting
December 6 - Megan Zirnstein (Penn State, Psychology) : Investigating Semantic Prediction in Second Language Processing
The ability to predict upcoming words in a sentence or text is an important aspect of the language comprehension system. When predictions are correct, behavioral and ERP research has shown that processing load is reduced (Federmeier, 2007; Van Verkum, 2008). However, when predictions are false and do not fit with the sentence or discourse context, there are processing costs, often in the form of longer reading times (Van Berkum et al., 2005) or higher amplitude ERP effects (i.e., the N400 and a frontally-distributed positivity; Federmeier et al., 2007). In this talk, I will present an overview of what we currently know about how readers predict in their native language, and discuss implications for processing in a second language. I will also present data from a recent ERP study investigating prediction effects when bilinguals read in their second language.
December 13 - Angela Grant (Penn State, Psychology) : Working hard really does pay off: An fMRI investigation of lexical access in L2 learners
This study uses functional magnetic resonance imaging (fMRI) to investigate how the development of lexical access in second language (L2) learners may be influenced by individual differences in working memory (WM) and inhibitory control (IC). Models of bilingual processing suggest that (a) bilinguals must consistently use IC in comprehension and production and (b) highly proficient learners access concepts directly while less proficient learners access concepts only through the L1 (Green, 1998; Kroll & Stewart, 1994). The neural implication of these models is that less proficient bilinguals, compared with highly proficient bilinguals, will require more effort to inhibit their L1 in order to successfully retrieve words in the L2.
Our hypothesis based on the current neuroimaging literature is that lower proficiency learners will more strongly activate inhibitory control areas, such as the left inferior frontal gyrus (LIFG) and anterior cingulate cortex (ACC), in addition to areas associated with semantic retrieval, such as the left middle temporal gyrus (LMTG). Higher proficiency learners, by contrast, should utilize more efficient networks where the LMTG acts as a hub of semantic retrieval rather than in the LIPFC or ACC for cognitive control (Abutalebi, 2008; Yokoyama, 2009). Participants in our study completed measures of proficiency (TVIP; Dunn et al. 1986) WM (phonological letter number sequencing; Wechsler, 1997) and IC (flanker task; Emmorey et al. 2008) before an fMRI experiment in which participants were asked to make a language-specific lexical decision on L1 words, L2 words, and homographs (e.g.. pie, the Spanish translation of foot).
Behavioral results did not show any significant correlations between proficiency, WM, or IC. Although these measures did not correlate significantly with each other, when included as covariates in the fMRI data analysis, we observe a positive correlation between greater activation in our pre-specified regions of interest, working memory, and proficiency. This suggests that, contrary to our predictions, L2 learners with higher WM and proficiency are calling on the LIFG, ACC, and LMTG more when accessing L2 words, rather than less. Specifically, learners are utilizing the LIFG under high conflict conditions (when identifying homographs), and the ACC under normal conflict conditions (when identifying unambiguous Spanish words). Participants also utilized the LMTG and left precuneus more when accessing Spanish words compared with English words, areas associated with semantic and episodic retrieval, respectively. Results are discussed in the context of current neuroimaging models of second language acquisition and bilingualism.