“Understanding 'prediction' in the cloze task: Data and model”

Abstract:
In the very large literature investigating effects of lexical predictability on language comprehension, the predictability of a word is generally operationalized as its cloze probability, i.e., the proportion of subjects who provide the word in question as the likely next word of the sentence, given the preceding context. The present work is motivated by the observation that without a model of how subjects perform the cloze task itself, we lack a clear interpretation of the cloze probability variable. To develop such a model, we first collected cloze responses in a speeded version of the cloze task. Subjects (N = 33) read sentences in RSVP format, and at some point in the sentence were prompted to speak the next word within 3 seconds. Each subject completed 375 critical trials. Responses were highly consistent with previous paper-and-pencil cloze norming for these items. The critical findings were that RT decreased linearly with (a) the probability of the subject’s response, and (b) the item’s level of constraint. These two effects were both highly significant, and were additive; higher item constraint resulted in reduced RT at each level of response probability. We show that a simple race model of response selection in the cloze task, which we call the Cloze Race Model (CRM), generates the same patterns. In the model, each lexical item advances stochastically and independently toward a response criterion. Sentence context is modeled as influencing the distribution of starting points of the potential responses. Across a wide range of parameter settings, RTs are faster when a response has higher probability, and when the item is more constraining; as in the data, these effects are additive. Based on the data and model, we suggest that variation in cloze responses may result from the stochastic nature of the response process, not from between-subject variability. The data and model also suggest that cloze probability may be a very imperfect measure of lexical expectation or activation.

Maria Piñango, Yale University, Linguistics
Dissociations in Broca's Area and the Structure of Language Processing

Abstract:
Even though the connection between Broca's area and language use has been well-established, work from neuroimaging and neuroanatomy from the past fifteen years or so has shown that Broca's area is not homogeneously recruited even for language specific processes such as sentence comprehension (e.g., Amunts et al., 1999, Bookheimer, 2002, Price, 2010). Otherwise competing neurocognitive models (e.g., Friederici, 2002/2012, Arbib, 2005, Hagoort, 2005, Rogallsky & Hickok, 2009) converge on the observation that Broca's area must be further articulated separating the function of the pars opercularis (BA44) from that of the pars triangularis (BA45) and that only the latter be connected to the function of the pars orbitalis (BA47). The talk will focus on this articulation by bringing to bear evidence from our lab which not only supports such a differentiation, but also points to a carving of function that constrains in some respects our traditional views of how language structure gets implemented by the brain, and how more generally we should be understanding the architecture of the system for processing purposes.

A novel framework for a neural architecture of language

In this talk I will argue that high-level linguistic processing is accomplished by the joint engagement of two functionally and computationally distinct brain systems: i) the classic “language regions” on the lateral surfaces of left frontal and temporal lobes that appear to be functionally specialized for linguistic processing, showing little or no response to arithmetic, general working memory, cognitive control or music (e.g., Fedorenko et al., 2011; Monti et al., 2009), and ii) the fronto-parietal network, a set of regions that are engaged across a wide range of cognitive demands (e.g., Duncan, 2001, 2010). Most past neuroimaging work on language processing has not explicitly distinguished between these two systems, especially in the frontal lobes, where subsets of each system reside side by side within the region referred to as “Broca’s area”. Using methods which surpass traditional methods in sensitivity and functional resolution - i.e., identifying key brain regions functionally in each individual brain (Fedorenko et al., 2010; Saxe et al., 2006) - we are beginning to characterize the important roles played by both domain-specific and domain-general brain regions in linguistic processing.

"Preliminary comparison of spectral versus ultrasound biofeedback: Implications for the causes of residual speech errors"

Speech motor system is thought to be linked to speech perception. Recent studies have suggested that motor areas somatotopically related to individual speech organs such as the lips and tongue can be coactivated in a phoneme-dependent manner during speech perception. However, the direct contribution of articulatory movements to speech perception has been little studied. We examined the intelligibility of plosive-initial syllables when listeners simultaneously and silently articulated discordant plosive-initial syllables, and compared with those when listeners simultaneously watched a face producing discordant syllables. The results revealed an interesting contrast between auditory-visual and auditory-articulatory interactions. In the case of auditory-visual interaction, larger divergence of the visible articulatory information from the auditory stimulus had a heavier effect. On the contrary, auditory-articulatory interaction occurred only WITHIN the same primary articulator, and did not ACROSS different articulators.

We usually think of well-learned daily movements, such as reaching and walking, as being fully controlled by our action intentions. Typically, the sensory outcomes of our movements would be recognized as self-actions (as opposed to movements of the external world) due to an internal model prediction using 'efference copy' of our motor commands. Here we describe an exception to this rule, which is caused by an implicit motor process. When we step onto a stationary escalator, we frequently suffer clumsy body and leg movements and associated odd sensation. Even though we completely understand the environmental state of the stopped escalator, we cannot avoid behaving clumsily, and feeling an odd sensation before adaptation. From experimental results of behavioral measures and subjective odd sensation scores, we suggested that this odd sensation is triggered by the implicit expectation that the escalator is in fact moving, rather than the changes of the center of gravity by irregular steps or entrance slope. Further experiments involving the systematic manipulation of the surrounding visual motion indicated that exogenous postural changes elicited by visual motion can induce a similar odd sensation. This result suggests that the implicit behavioral change itself, rather than endogenous motor program, is a 'necessary' factor of the odd sensation in perception of movement.

This talk is part of the Speaker Series: THE HASKINS LEGACY - The Science of the Spoken and Written Word.

Abstract
"The purpose of this presentation is to summarize the key findings from two studies that investigated the role of vocabulary knowledge in early literacy development, and discuss the implications of these findings for the structure of the simple view of reading (SVR) theoretical framework. The aim of the first study (Tunmer & Chapman, Scientific Studies of Reading, 2011) was to investigate the hypothesis that, in addition to having a direct relation to future reading comprehension performance, early vocabulary knowledge contributes to the development of both decoding and word recognition skills indirectly through "set for variability", the ability to determine the correct pronunciation of approximations to spoken English words. The aim of the second study (Tunmer & Chapman, Journal of Learning Disabilities, in press) was to investigate the hypothesis that the contributions of oral language comprehension (C) and word recognition (D) to reading comprehension (R) in the SVR model are not entirely independent because a component of C (vocabulary knowledge) directly contributes to variance in D. The findings of the two studies have both theoretical and practical implications. Regarding theoretical issues, the findings provide the basis for resolving differences between the lexical quality and phonological processing accounts of reading acquisition and reading disabilities. Regarding implications for educational practice, the findings suggest that prevention programs for children at risk of reading failure should focus as much attention on improving these children's oral language skills, especially vocabulary knowledge, as on improving their phonological and alphabetic coding skills. Greater attention may also need to be given to promoting the development of set for variability in beginning readers."

Supplementary reading includes two papers, which are attached:

1. William E. Tunmer and James W. Chapman. "The Simple View of Reading Redux: Vocabulary Knowledge and the Independent Components Hypothesis." (Tunmer_JLD.pdf)

2. William E. Tunmer and James W. Chapman. "Does Set for Variability Mediate the Influence of Vocabulary Knowledge on the Development of Word Recognition Skills?" (TunmerSSSR.pdf)

Download supplemental reading in PDF

• NEL_514.pdf
• JSL5305_Sidtis.pdf
• c17.pdf

MORTEN CHRISTIANSEN (Cornell University):
"You Are What You Read: The Impact of Experience on Individual Differences in Language Processing"

Stream the two part talk
Part 1 [.mov]
Part 2 [.mov]

Abstract: Individuals differ considerably in their ability to process language. Traditionally, such differences have been attributed to variations in a biologically-determined working memory capacity. In this talk, I propose an alternative account that emphasizes the importance of variations in linguistic experience as a cause of individual differences in language processing. Focusing on relative clause processing, I present results from studies manipulating language exposure in both connectionist networks and human subjects to test predictions from this usage-based perspective. Further experimental data suggest that individual differences in basic abilities for sequential learning and processing may affect individuals' ability to learn from linguistic experience. Together, these results underscore the role of experience in sentence comprehension and challenge the importance of a fixed working memory in accounting for individual differences in language processing.

Morten will be visiting UConn/Haskins for 3 days and will give 3 talks (abstracts below):

Title: "Brains, Genes and Language Evolution"
Place: Bousfield 160, Storrs
Time: 4pm, September 7, 2011

Title: "You Are What You Read: The Impact of Experience on Individual Differences in Language Processing"
Place: Haskins Laboratories
Time: 12:30pm, September 8, 2011

Title: "The Sound of Syntax: The Importance of Phonology in Syntactic Acquisition and Processing"
Place: Bousfield 109, Storrs
Time: 12:00pm, September 9, 2011

Abstract:

Previous work suggests that concurrent changes in jaw displacement and formant frequencies may manifest rhythmical characteristics of spoken American English utterances [e.g., Erickson et al, 2011. Rhythm and Emphasis in American English : Comparison of native and non-native speakers' productions. Proceedings of International Seminar of Speech Production, Montreal, Canada, pp.345-352]. This paper compares jaw displacement and corresponding formant frequencies of monosyllabic American English words produced on low vowels as spoken by three native speakers and three Japanese speakers of English. Preliminary findings for the native speakers show an alternating pattern of strong-weak jaw displacement along with corresponding changes in F1. Interestingly, this articulatory/acoustic pattern seems to correspond with metrical grids for these utterances. For the non-native speakers, this pattern was not consistently seen, and seemed to vary as a function of the speakers' skill in spoken English. (This work was supported in part by the Japanese Ministry of Education, Science, Sport, & Culture, Grant-in-Aid for Scientific Research (C) #22520412.)

Maturation of neural correlates of speech processing

Abstract
Behavioral research has established that infants initially can discriminate many of the phonemic contrasts found cross-linguistically during the first few months of life. However, during the latter half of the first year of life the pattern of discrimination shifts, with infants maintaining discrimination of phonemic contrasts in the ambient language, but failing to show discrimination of many non-native contrasts that are not part of the child's environment. These findings implied that native language speech perception was fairly mature early on. However, more recent research has demonstrated failure to discriminate native language contrasts in more complex tasks for toddlers between 12- and 20-months of age, and this failure has been linked to the development of the lexicon, task difficulty, and "developmental level" (Curtin, Byers-Heinlein & Werker, 2011). A missing explanation for differences in performance on speech perception tasks during the first few years of life is that the neural mechanisms supporting these tasks are highly immature. Both experience and neural maturation will contribute to a child's "developmental level"; however, it is very difficult to determine the contribution of each using behavioral methods. Neurophysiological methods, in contrast, are ideal for examining the relationship between neural maturation, language experience and speech perception growth. In this talk I will demonstrate how event-related potential (ERP) correlates of speech sound encoding, discrimination and attention can be used to inform our understanding of speech perception development in infants and children exposed monolingually to English and bilingually to Spanish and English.

Abstract:
In most of our conversations, we typically hear and see a speaker talk. In this talk, I will illustrate some of the diverse benefits that audiovisual communication provides the listener with across the lifespan, such as in learning the meaning of novel words and in the recognition of known words. For example, seeing the speaker can aid in the acquisition of referential meaning for novel words. We have shown that in adult-to-child interactions, the motion imposed on a referent object by adult speakers is temporally aligned to the prosodic structure of their accompanying utterances. Adult perceivers are sensitive to this prosodic alignment in resolving the referent of a novel word. Two-year-old children can use the gesture-speech alignment to deduce and learn the meaning of novel words.Seeing a speaker also aids in the recognition of already known words, even under possibly cognitively-demanding listening conditions. Older adults are more affected in their comprehension of a speaker when a competing speaker is also audible than are young adults. This seems to be due to age-related hearing loss but also possibly due to cognitive decline found with aging. We have shown that older listeners benefit, nevertheless, in this listening situation from also seeing the speaker they intend to listen to. Older and young adults benefitted equally in their speed at which they identified phonemes, but young adults benefitted more than older adults in response accuracy. Visual speech aids listeners here by providing segmental information about a phoneme, but also by providing more global contextual information. Possible processing levels at which visual speech benefits listeners as well as factors potentially explaining individual differences in the size of the obtained audiovisual benefit will be discussed.

Abstract: It has been shown several times that the speech recognition accuracy improves if the direct measurement of speech articulation is used in addition to the speech acoustics from the talker. However, access to such direct speech articulation data during speech recognition is not feasible in practice. In my presentation, I shall show that the speech production/articulation features can be estimated from the speech signal of any arbitrary talker although these features are not directly available from the talker. For the estimation of production-oriented features, I shall present a talker-independent acoustic-to-articulatory inversion framework using generalized smoothness criterion, which requires parallel articulatory and acoustic data from a single subject only (exemplar) and this exemplar need not be any of the talkers. Use of these estimated features improves the acoustic-feature based recognition accuracy by ~4% (absolute) in a phonetic recognition experiment on TIMIT corpus. Interestingly, when the exemplar is interpreted as a listener, the production-oriented features and, hence, the speech recognition can be interpreted as listener-specific. We will see that such a listener-specific framework to speech recognition provides a production-oriented explanation of the variability in recognition accuracy by non-native listeners. We will also see that the listener-specific framework acts as a bridge between the scientific and technological viewpoints towards the role of speech production in speech perception in the human speech communication.

BRIAN FIDALI (Yale '11):
"Perceiving Perturbations in Polyrhythms"

ALLEN ZHANG (Yale '11):
"Synchronization with Auditory or Visual Sequences in the Presence of Cross-Modal Distracters"

EVAN ZHAO (Yale '11):
"Does Note Spacing Play Any Role in Music Reading?"

Abstract:
Segmenting words from continuous speech is not a trivial task to master as the signal itself does not contain unique phonetic markers of word boundaries. Nevertheless, infants master this task within the early phase of language acquisition typically starting their language production with utterances roughly corresponding to words of the target language. I will provide an overview on studies from our lab showing that from early on word segmentation skills are affected by the specific phonological properties of the target language. A special focus will be laid on children learning German as a stress-based and on children learning Turkish considered as a syllable-based language with vowel harmony as a cue to word boundaries.

Abstract:
Are black-white differences in reading achievement related to differences in language background? Many African American children speak a dialect of English that differs from the mainstream dialect emphasized in school. We examined how knowledge of alternative dialects affects generating pronunciations from print, an important component of early reading and marker of reading skill. Computational models show that learning is slower for words with different pronunciations in the two dialects. Behavioral data show that knowledge of alternative pronunciations affects adult performance, with larger effects for readers who use more African American English. The results indicate that the achievement gap may be due in part to differences in task complexity: bidialectal children fall behind because their task is more difficult but they are given the same amount of time to reach developmental milestones.

"Articulatory Characterization of Complex Consonants - Liquids and Fricatives"

"Do you know who that is? An ERP examination of ambiguous anaphors with real-world referents"

Abstract:

Establishing coreference between anaphors and their antecedents allows comprehenders to track “who did what to whom”, facilitating construction of an accurate discourse model. Anaphors used in research rarely refer to actual people, objects or events and typically have very minimal semantic detail. In contrast, readers frequently possess substantial information about discourse entities, and recent findings suggest that processing difficulty can arise when potential antecedents are semantically similar. In the current study event-related potentials were recorded as participants read sentences containing ambiguous pronominal anaphors. Such anaphors contain semantic information about the sex of their antecedent, and a P600 effect typically results when they mismatch their antecedents (e.g., “she” referring to a male). As such, sentential pronouns either matched or mismatched their referents. In order to investigate the influence of additional real-world information on coreferential processing, antecedent entities were either semantically “empty” or were known, famous individuals. Mismatching anaphors elicited a P600 effect whose amplitude was significantly greater when sentence characters were real. Moreover, matching pronouns elicited a P600-like effect when their antecedents were fictional. Finally, sentence-final words elicited an N400 effect following referentially failing pronouns relative to those with matching pronouns – but only in "empty" reference conditions. Our results suggest that the presence of high-quality representations in a discourse model facilitates the processing of coherent coreference.

"The Speech-To-Song Illusion revisited"

Abstract:

Speech-to-song illusion refers to a perceptual effect (first described by Diana Deutsch in 1995) when a spoken phrase shifts to being heard as sung without changing any acoustic characteristics of the signal. This is achieved by a simple repeating of the phrase several times. This study investigates the boundaries of speech and song from acoustic-perceptual perspective. Using the speech-to-song illusion as a method, we tested rhythmic and tonal hypotheses in order to find out whether acoustic characteristics can cue the perceptual classification of a sentence by German listeners as sung or spoken. First, our results show that, despite individual differences, the speech-to-song illusion is a robust perceptual phenomenon comparable to those known in visual perception. Second, the experiment revealed that acoustic parameters – especially tonal structure – facilitate the perceptual shift from speech to song pointing to an acoustically guided decoding strategy for speech- vs. song-like signals.

Abstract

Relationship between auditory acuity and the use of motor equivalence in production of the sibilant “sh”.
J. Brunner, S. Ghosh, P. Hoole, M. Matthies, M. Tiede & J. Perkell

Motor-equivalent trading relations (i.e. different articulations with essentially the same acoustic output) reflect strategies for compensating for perturbations. Previous studies have shown that some speakers use motor equivalence but others do not. In the present study we test the hypothesis that the use of these strategies is linked to the speakers' auditory acuity: Speakers with high auditory acuity will use motor equivalence to a greater extent and concomitantly produce clearer phoneme contrasts than low acuity speakers.

In a first experiment, seven speakers were recorded via electromagnetic articulography while adapting over the course of two weeks to an articulatory perturbation consisting of an artificial palate. Motor equivalence in production of the sibilant “sh”, i.e., more lip protrusion when the tongue tip is fronted and less if it is retracted, was found for five of the seven speakers but not for the remaining two.

In a second experiment, the speakers' auditory acuity was assessed by testing their ability to hear small differences along a synthetic s-sh continuum. The results indicate that higher-acuity speakers used motor equivalence to a greater extent when adapting to a perturbation than lower-acuity speakers. Additionally, higher-acuity speakers produced greater acoustic contrasts than lower-acuity speakers.

Speech rate had an influence on the use of motor equivalence: slow speakers used motor equivalence to a lesser degree than fast speakers.

Abstract:

Coreferential processing is thought to be modulated by an antecedent’s accessibility in a comprehender’s discourse model. Many theories of coreferential processing suggest that an antecedent’s availability as a referent is the result of being in linguistic focus, as when an entity is mentioned first or most frequently in a discourse, or when an entity occupies a prominent position in a sentence’s syntactic structure. However, it may also be possible to enhance an entity’s referential accessibility by virtue of domain-general principles of memory (such as primacy, frequency, or levels-of-processing effects) without appealing to higher-level linguistic operations such as syntactic processing. Because these memory and language effects are typically confounded, the question of whether referential accessibility may be attributed to non-linguistic rather than linguistic processes has never been directly tested. The current study used both event-related potentials (ERPs) and behavioral measures in a series of experiments in order to investigate the effects of linguistic and non-linguistic focus on coreference. We examined this question against the backdrop of a well-known finding in the coreferential processing literature, the repeated name penalty (Gordon, Grosz, & Gilliom, 1993); this penalty results when an entity that is prominent in a sentence’s syntactic structure is referred to by a repeated name anaphor. By using a generation task (analogous to a levels-of-processing manipulation) we modulated antecedent accessibility without altering the entity’s status in sentence/discourse structure. Experiment 1 showed that, while both generation and syntactic focus improved recognition memory performance, only the syntactic manipulation elicited online discourse processing effects – an N400 effect, the electrophysiological correlate of the repeated name penalty (Swaab, Camblin & Gordon, 2004). When syntactically-enhanced access was removed as a factor in Experiment 2, generation interacted with repetition in the offline recognition memory for antecedent names, while no discourse processing consequences in the ERP waveforms were evident. In Experiment 3, memory enhancement by means of generation and repetition was examined for proper names in word lists. Lexical repetition effects (reduced N400 and increased LPC amplitudes) were found regardless of generation condition; however, repetition and generation elicited different patterns of recognition memory performance. These results suggest that general memory enhancement alone is insufficient to produce discourse processing effects, and that sentence comprehension engages recognition memory processes that can be distinguished from those elicited by word lists.

Abstract:
Developmental dyslexia (DD) is one of the most prominent specific learning disabilities, affecting 5-17% of children. Heritability of DD has been supported by strong evidence from molecular-genetic studies and from studies of twins and families with DD. Substantial evidence suggests that neurological abnormalities underlie DD in children and adults. Multiple functional neuroimaging studies comparing adults and children with DD to typical controls have observed dysfunctions within various brain regions. Differences in the neural correlates of language and auditory processing have also been observed in infants with a family history of reading or language impairments compared to those without. Additionally, it has been suggested that the functional differences observed in children with DD may be related to morphological brain differences. Measures of brain volume have revealed structural abnormalities in children and adults with a diagnosis of DD in regions that broadly overlap with observed functional differences.
A number of key questions remain unanswered. We do not know whether observed functional and structural brain differences associated with DD are present /prior /to reading onset in those who will later receive a diagnosis. We do not know whether these functional and structural differences could be exploited to predict later reading outcome (DD in particular). It also remains unclear how key skills that are impaired in those with DD (e.g., phonological processing and rapid auditory processing) develop in children as they move from the pre-reading to the skilled reading stage. In this talk, I will present preliminary results from our ongoing longitudinal study which investigates possible neural and behavioral pre-markers of DD in pre-reading children (age 4-5) with and without a family history of DD. All children have been examined with standardized tests sampling language and cognitive functions, psychophysical measures, functional magnetic resonance imaging (using a phonological processing task, a rapid auditory processing task and a executive functioning task) and whole brain structural imaging prior to reading onset and after one year of reading instruction. I will further present some recent data looking at the neural correlates of reading fluency in an adult population.
**

Talk title TBA

Susie Levi, New York University
"Talker perception by children with SLI: A possible window into the disorder?"

Christina Reuterskiöld, New York University
"tba"

Abstract: Precise vocal pitch regulation is crucial for prosody in speech and accurate production of notes and melodies in singing. The integration of auditory feedback with the vocal motor system, known as audio-vocal
integration, is essential for vocal pitch regulation; however, its neural substrate is not well understood. We used singing tasks in three functional magnetic resonance imaging (fMRI) experiments designed to: 1) investigate the neural mechanisms underlying vocal pitch regulation, and 2) evaluate the effects of training on vocal performance and neural activity involved in vocal pitch regulation. In Experiment 1, we tested non-musicians and experienced singers with particular singing tasks to
find experience-dependent neural substrates of audio-vocal integration. To target brain activity relevant for audio-vocal integration, we used pitch-shifted auditory feedback and instructed subjects either to adjust
or ignore the shifted feedback. Experiment 2 used a similar fMRI paradigm with large and small pitch shifts to distinguish between voluntary and involuntary vocal pitch regulation and to elucidate the neural mechanisms governing each type of vocal pitch regulation. In
Experiment 3, we trained non-musicians in an auditory discrimination task and administered perceptual discrimination and singing tasks to determine whether training-enhanced auditory discrimination would result
in improved vocal accuracy. We also used fMRI to evaluate any training-induced modulations within brain regions involved in singing and audio-vocal integration. Collectively, the behavioral data showed that correcting for pitch-shifted feedback is an automatic process,
whereas ignoring shifted feedback requires training. The imaging data confirmed the existence of a singing network, comprised of auditory cortex, anterior cingulate cortex, insula, motor and premotor areas, thalamus, and cerebellum; within this network, experienced singers
recruit additional activity in the superior temporal sulcus, rostral cingulate zone, and anterior insula as a network for audio-vocal integration that specifically underlies voluntary vocal pitch regulation. However, auditory training alone is not sufficient to improve vocal accuracy or engage this audio-vocal integration network in
non-musicians.

Abstract:
It is generally assumed that the early stages of speech perception involve the extraction of some kind of generalized auditory patterns or properties from the peripheral input. The auditory representation that results is of considerable interest, since it serves as the input
to higher-level, speech specific processes of phonetic perception. The current research examines this auditory representation using a priming paradigm, in which perception of vowel targets is facilitated when the targets are preceded by acoustically-matched nonspeech stimuli. By manipulating the acoustic parameters of these nonspeech "prime" tones, it is possible to determine the role of these parameters in the auditory stages of vowel processing. Previous results [Wallace and Blumstein, J. Acoust. Soc. Am. 119, 3245 (2006)] suggest a short
window of analysis of no more than 25 ms. In Experiment 1, frequency of nonspeech primes was varied, with results suggesting broad frequency tuning. In Experiment 2, primes matched to both F1 and F2 of the target vowels were found to elicit a greater priming effect than would be predicted by summing the response to separately-presented F1 and F2 primes, suggesting that the auditory representation of vowels encodes combinations of formant frequencies.

Abstract
How do we understand the actions of others as they are described to us through language? Embodied theories of conceptual knowledge suggest that sensory-motor representations of actions similar to those involved in the performance of the action described are recruited during language comprehension. The extent of this recruitment, however, and the brain mechanisms supporting how word meanings are combined remain unknown. In this talk, I will discuss three neuroimaging studies investigating these issues. One study compared the activity elicited by sentences that conveyed different degrees of physical effort (push the piano vs. push the chair vs. see the piano). A graded BOLD response to these sentences, reflecting the graded effort required to perform these actions, was found in the same motoric areas that were activated by physically stressing a ball. A similar response was also found in a part of Broca’s area that was not active in action performance. A subsequent study suggests that the motoric component of such findings is likely confined to brain areas implicated in planning actions, rather than executing them. The fact that Broca’s area is sensitive to the real-world effort entailed by executing the described actions suggests that language comprehension involves a complex interplay between abstract representations of action and linguistically based representations.

Abstract
The world about us changes at an extraordinary pace. If language is to have any influence on what we attend to, that influence has to be exerted at a pace that can keep up. In this talk I shall focus on two aspects of this requirement: The speed with which language can mediate visual attention, and the fact that the cognitive system can very efficiently make up for the fact that, to be expedient (i.e. to keep up with the changing world) we do not in fact refer to all the changes that are associated with, or entailed, by an event. Rather, we infer aspects of those changes. One example of this is through elaborative inference, and another is through the manner in which we track (often unstated) changes in the states of objects as those objects undergo change. The talk will conclude with data suggesting that multiple representations of the same object in different event-dependent states may in fact compete with one another, and that this competitive process may bring both costs and benefits.

Abstract:

This work focuses on speech development in children with genetic duplication at 7q11.23 ("DUP7"), the same region where deletion results in Williams syndrome ("WS"). Findings include: (1) phonological and apraxia-like motor speech deficits are present in the majority of preschool/school-aged children with DUP7; (2) a comparison of matched pairs of 21- and 24-month-old children with DUP7 versus WS demonstrates reduced volubility and consonant inventories in toddlers with DUP7.

Abstract:
Researchers have long noted that children's early word productions are variable in form, yet the nature and extent of this variability has not been systematically examined. Furthermore, much of our understanding of these issues comes from orthographic and/or IPA transcriptions that may be unable to capture systematic covert contrasts being made. As a result, we still know little about children's early phonological and morphological representations, and how these develop over time. However, our research has begun to show that some of children's early variability in morpheme production is due to contextual effects. Other work finds that children provide acoustic cues to "missing" segments. These results suggest that early phonological and morphological representations may be more robust than often assumed. The theoretical and applied implications are discussed.

Abstract:
MIMICRI is a signal processing algorithm that has been shown to blindly infer and invert memoryless nonlinear functions of unobservable bandlimited signals, such as the mapping from the unobservable positions of the speech articulators to observable speech sounds. This blind
inversion can be accomplished because the bandwidth of signals almost always increases when transformed by a nonlinear function. Thus, if we transform the observable signals to have the same pass-band as the unobservable signals, then we are within an affine transform of the
unobservable signals. We review results of using MIMICRI on toy problems and on human speech data. We note that MIMICRI requires that the user specifies two parameters, the dimensionality and pass-band of the unobservable signals, but the user may not know the best values to
use. We show how to use cross-validation techniques with MIMICRI to help estimate parameters that previously needed to be specified. An unexpected consequence of this work is that it helps separate signals that have different frequency characteristics. For example, since lip motions tend to be slower than tongue motions, we may be able to
separate the acoustic effects of lips from the acoustic effects of the tongue, or possibly add noise robustness to speech recognition.

Abstract

The property of duality of patterning Ð the existence of two levels of structure, a meaningful level of words and sentences alongside a meaningless level of sounds Ð has been characterized as a basic design feature of human language (Hockett 1960). Some have also argued that a meaningless level, i.e., phonology, must have existed prior to hierarchical syntactic structure in the evolution of language (Pinker & Jackendoff 2005). Sign languages were admitted to the 'bona fide language club' only after Stokoe (1960) demonstrated that they do exhibit duality. But is it possible for a conventionalized language to exist without a fully developed phonological system—without duality? Using evidence from a sign language that has emerged over the past 75 years in a small, insular community, I will show that phonology cannot be taken for granted. The Al-Sayyid Bedouins have a conventionalized language with certain syntactic and morphological regularities (Sandler et al 2005, Aronoff et al 2008), but the language is apparently still in the process of developing a level of structure with discrete meaningless units that behave systematically. In other words, we don't find evidence for a full-blown phonological system in this language. Can a language go on like this? Data from children and from families with several deaf people help to pinpoint emerging regularities and complexity at the level of meaningless formational elements in ABSL. While phonology in language cannot be taken for granted, then, its existence in all older languages, spoken and signed, suggests that it is inevitable. Rather than assume that phonology is somehow 'given' or hard-wired, this work leads us to ask, Why and how does it arise?

Abstract:

Studying the complex and rapid processing of spoken language in typical and impaired populations requires a sophisticated set of scientific tools. I will describe a combination of experimental, statistical, and computational modeling tools developed for the study of typical spoken language processing and how these tools can be used to study the nature of spoken language impairments in aphasia. The visual world eye-tracking paradigm provides an experimental tool that reveals the precise time course of spoken language processing in a naturalistic and (relatively) simple task. Growth curve analysis provides powerful quantification of fixation time course, including quantification of individual differences. Simulation studies of computational models provide a concrete method for testing theoretical interpretations of the experimental results. When applied to the study of spoken word recognition in aphasia, these techniques suggest an integrated "dynamic balance" theory of Broca's and Wernicke's aphasia.

Abstract:
Language production is sometimes described as a process in which conceptual semantic messages drive the selection of linguistic information. Such a description is probably appropriate for some aspects of the process (e.g. noun selection), yet it is clearly incomplete.

Consider for instance the fact that in numerous languages determiner forms depend not only on semantics, but also on several other kinds of information. In Germanic, Slavic, and Romance languages, the retrieval of the determiners (and other closed- class words, such as pronouns) also depends on a property of the nouns called "grammatical gender." For instance, in Dutch, nouns belong to the so-called "neuter" gender or to the "common" gender. The definite determiners accompanying the nouns belonging to the two sets are respectively het (e.g. het huis, 'the house') and de (e.g. de appel, the apple). In English, consonant-initial nouns and vowel-initial noun can require different indefinite article forms (e.g. a pear vs. an apple).

Such properties of determiners surely impose constraints on how these lexical items can be retrieved. For this very reason, determiners provide a broad testing ground for contrasting psycholinguistic hypothesis of lexical processing and grammatical encoding. In my talk, I will review the cross-linguistic research I have been conducting on determiner retrieval. One important question that will be asked, and only tentatively answered, concerns the extent to which open-class words such as nouns and closed-class words such as determiners are processed and selected by similar mechanisms.

*Early literacy development: Tracking the influences of genes, homes, teachers, schools and countries *

Brian is a long-time Haskins collaborator who has been a leader in
behavior genetic studies of reading and learning.

Abstract:
Comprehension and production deficits associated with
Broca's agrammatism have traditionally received non-converging
generalizations despite having in common the idea that the deficit has a syntactic source. Such a situation seems to suggest that the locus of the problem may not be in the linguistic representation itself but in its dynamic instantiation. Capitalizing on this insight, I present an alternative view whereby the production and comprehension patterns of impairment manifested in the syndrome stem from a unified processing-based source. Crucially, this view is built upon a linguistically-based processing model that takes syntactic representation to be a dynamic manifestation of sentence formation. This analysis and its implications will be discussed in the context not only of theories of cortical distribution of language, given the localizing value of Broca's aphasia agrammatism, but also of theories of language processing and representation.

MEIJIN BRUTTOMESSO:
Effects of Metrical Subdivision on the Continuation, Reproduction, and Perception of Beat Tempo"

JACKIE THOMPSON:
Context Effects in the Tritone Paradox

HAITHAM JENDOUBI:
Effects of Phase-Shifted Subdivisions on Synchronization with a Beat

Animal models of early cortical disruption: possible relevance tso higher-order functional disruptions in humans

March 27, 12:30 PM
Haskins

March 28, 4:00PM
UConn
Bousfield 160, STORRS

Ken McRae, University of Western Ontario

How Knowledge of Real-world Events Influences Language Comprehension

Abstract
A significant proportion of everyday utterances concern real-word events. Thus, people's knowledge of everyday events, including their common participants, is an important component of language comprehension. In many theories of language comprehension, event knowledge is outside of "the lexicon", is accessed slowly, and influences comprehension only after an architecturally-determined time delay. In contrast, I present results from semantic priming and sentence comprehension studies that strongly support a view in which event knowledge is organized efficiently so that it is computed immediately from words (and combinations of words). In addition, our studies show that event knowledge is an important source of information that is used to generate expectancies about upcoming concepts and syntactic structure during on-line sentence comprehension.

Abstract:
Variation set structure --- partial alignment of successive utterances in child-directed speech --- has been shown to correlate with progress in the acquisition of syntax by children. The present study demonstrates that arranging a certain proportion of utterances in a training corpus in variation sets facilitates word segmentation and phrase structure learning in the acquisition of miniature artificial languages by adult subjects.

The positive effects of variation sets in the reported experiments suggest that learners can reuse the same learning procedures --- alignment, and comparison --- at different levels of linguistic structure (here, lexical and phrasal units). We are presently extending
our approach to investigate whether variation sets also facilitate the learning of other core features of language, such as lexical categorization, long-distance dependencies, and recursion.

Beyond having implications for understanding the course of L1 acquisition by children, this work contributes to the development of more efficient algorithms for automatic language acquisition, as well as better methods for L2 instruction.

Also visiting will be Albert Zhang and Dan Hudock from East Carolina University. (PowerPoint presentation)

"Hearing lips and seeing voices: How cortical areas supporting speech production mediate audiovisual speech perception."

Len Katz. Two statistical issues journal editors are concerned about:
(1) the appropriate analysis when your experiment includes items or tokens, and
(2) the death of null hypothesis testing.

Fermin Moscoso Del Prado Martin. MRC Cognition and Brain Sciences Unit, Cambridge, UK. A sketch of an information-theoretical approach to lexical processing.