Publications

Abstract

Speaker diarization (answering 'who spoke when') is a widely researched subject within speech technology. Numerous experiments have been run on datasets built from broadcast news, meeting data, and call centers—the task sometimes appears close to being solved. Much less work has begun to tackle the hardest diarization task of all: spontaneous conversations in real-world settings. Such diarization would be particularly useful for studies of language acquisition, where researchers investigate the speech children produce and hear in their daily lives. In this paper, we study audio gathered with a recorder worn by small children as they went about their normal days. As a result, each child was exposed to different acoustic environments with a multitude of background noises and a varying number of adults and peers. The inconsistency of speech and noise within and across samples poses a challenging task for speaker diarization systems, which we tackled via retraining and data augmentation techniques. We further studied sources of structured variation across raw audio files, including the impact of speaker type distribution, proportion of speech from children, and child age on diarization performance. We discuss the extent to which these findings might generalize to other samples of speech in the wild.

Abstract

A large amount of variability exists across human brains; revealed initially on a small scale by postmortem studies and,
more recently, on a larger scale with the advent of neuroimaging. Here we compared structural variability between human
and macaque monkey brains using grey and white matter magnetic resonance imaging measures. The monkey brain was
overall structurally as variable as the human brain, but variability had a distinct distribution pattern, with some key areas
showing high variability. We also report the first evidence of a relationship between anatomical variability and evolutionary
expansion in the primate brain. This suggests a relationship between variability and stability, where areas of low variability
may have evolved less recently and have more stability, while areas of high variability may have evolved more recently and
be less similar across individuals. We showed specific differences between the species in key areas, including the amount of
hemispheric asymmetry in variability, which was left-lateralized in the human brain across several phylogenetically recent
regions. This suggests that cerebral variability may be another useful measure for comparison between species and may add
another dimension to our understanding of evolutionary mechanisms.

Abstract

Assimilation of place of articulation across a nasal and a following stop consonant is obligatory in Japanese, but not in Dutch. In four experiments the processing of assimilated forms by speakers of Japanese and Dutch was compared, using a task in which listeners blended pseudo-word pairs such as ranga-serupa. An assimilated blend of this pair would be rampa, an unassimilated blend rangpa. Japanese listeners produced significantly more assimilated than unassimilated forms, both with pseudo-Japanese and pseudo-Dutch materials, while Dutch listeners produced significantly more unassimilated than assimilated forms in each materials set. This suggests that Japanese listeners, whose native-language phonology involves obligatory assimilation constraints, represent the assimilated nasals in nasal-stop sequences as unmarked for place of articulation, while Dutch listeners, who are accustomed to hearing unassimilated forms, represent the same nasal segments as marked for place of articulation.

Abstract

In two experiments, Mandarin listeners resolved potential syntactic ambiguities in spoken utterances in (a) their native language (L1) and (b) English which they had learned as a second language (L2). A new disambiguation task was used, requiring speeded responses to select the correct meaning for structurally ambiguous sentences. Importantly, the ambiguities used in the study are identical in Mandarin and in English, and production data show that prosodic disambiguation of this type of ambiguity is also realised very similarly in the two languages. The perceptual results here showed however that listeners’ response patterns differed for L1 and L2, although there was a significant increase in similarity between the two response patterns with increasing exposure to the L2. Thus identical ambiguity and comparable disambiguation patterns in L1 and L2 do not lead to immediate application of the appropriate L1 listening strategy to L2; instead, it appears that such a strategy may have to be learned anew for the L2.

Abstract

Using a phoneme detection task, the present series of
experiments examines whether listeners can entrain to
different combinations of prosodic cues to predict where focus
will fall in an utterance. The stimuli were recorded by four
female native speakers of Australian English who happened to
have used different prosodic cues to produce sentences with
prosodic focus: a combination of duration cues, mean and
maximum F0, F0 range, and longer pre-target interval before
the focused word onset, only mean F0 cues, only pre-target
interval, and only duration cues. Results revealed that listeners
can entrain in almost every condition except for where
duration was the only reliable cue. Our findings suggest that
listeners are flexible in the cues they use for focus processing.

Abstract

Previous research has shown that, in a phoneme detection task, vowels produce longer reaction times than consonants, suggesting that they are harder to perceive. One possible explanation for this difference is based upon their respective acoustic/articulatory characteristics. Another way of accounting for the findings would be to relate them to the differential functioning of vowels and consonants in the syllabic structure of words. In this experiment, we examined the second possibility. Targets were two pairs of phonemes, each containing a vowel and a consonant with similar phonetic characteristics. Subjects heard lists of English words had to press a response key upon detecting the occurrence of a pre-specified target. This time, the phonemes which functioned as vowels in syllabic structure yielded shorter reaction times than those which functioned as consonants. This rules out an explanation for response time difference between vowels and consonants in terms of function in syllable structure. Instead, we propose that consonantal and vocalic segments differ with respect to variability of tokens, both in the acoustic realisation of targets and in the representation of targets by listeners.

Abstract

Listeners adapt rapidly to previously unheard talkers by
adjusting phoneme categories using lexical knowledge, in a
process termed lexically-guided perceptual learning. Although
this is firmly established for listening in the native language
(L1), perceptual flexibility in second languages (L2) is as yet
less well understood. We report two experiments examining L1
and L2 perceptual learning, the first in Mandarin-English late
bilinguals, the second in Australian learners of Mandarin. Both
studies showed stronger learning in L1; in L2, however,
learning appeared for the English-L1 group but not for the
Mandarin-L1 group. Phonological mapping differences from
the L1 to the L2 are suggested as the reason for this result.

Abstract

Languages contain tens of thousands of words, but these are constructed from a tiny handful of phonetic elements. Consequently, words resemble one another, or can be embedded within one another, a coup stick snot with standing. me process of spoken-word recognition by human listeners involves activation of multiple word candidates consistent with the input, and direct competition between activated candidate words. Further, human listeners are sensitive, at an early, prelexical, stage of speeeh processing, to constraints on what could potentially be a word of the language.

Abstract

In three experiments, the processing of lexical tone in Cantonese was examined. Cantonese listeners more often accepted a nonword as a word when the only difference between the nonword and the word was in tone, especially when the F0 onset difference between correct and erroneous tone was small. Same–different judgments by these listeners were also slower and less accurate when the only difference between two syllables was in tone, and this was true whether the F0 onset difference between the two tones was large or small. Listeners with no knowledge of Cantonese produced essentially the same same-different judgment pattern as that produced by the native listeners, suggesting that the results display the effects of simple perceptual processing rather than of linguistic knowledge. It is argued that the processing of lexical tone distinctions may be slowed, relative to the processing of segmental distinctions, and that, in speeded-response tasks, tone is thus more likely to be misprocessed than is segmental structure.

Abstract

English and French listeners performed two tasks - click location and speeded click detection - with both English and French sentences, closely matched for syntactic and phonological structure. Clicks were located more accurately in open- than in closed-class words in both English and French; they were detected more rapidly in open- than in closed-class words in English, but not in French. The two listener groups produced the same pattern of responses, suggesting that higher-level linguistic processing was not involved in these tasks.

Abstract

Listeners' recognition of spoken language involves complex decoding processes: The continuous speech stream must be segmented into its component words, and words must be recognized despite great variability in their pronunciation (due to talker differences, or to influence of phonetic context, or to speech register) and despite competition from many spuriously present forms supported by the speech signal. L1 listeners deal more readily with all levels of this complexity than L2 listeners. Fortunately, the decoding processes necessary for competent L2 listening can be taught in the classroom. Evidence-based methodologies targeted at the development of efficient speech decoding include teaching of minimal pairs, of phonotactic constraints, and of reduction processes, as well as the use of dictation and L2 video captions.

Abstract

The phoneme detection task is widely used in spoken word recognition research. Alphabetically literate participants, however, are more used to explicit representations of letters than of phonemes. The present study explored whether phoneme detection is sensitive to how target phonemes are, or may be, orthographically realised. Listeners detected the target sounds [b,m,t,f,s,k] in word-initial position in sequences of isolated English words. Response times were faster to the targets [b,m,t], which have consistent word-initial spelling, than to the targets [f,s,k], which are inconsistently spelled, but only when listeners’ attention was drawn to spelling by the presence in the experiment of many irregularly spelled fillers. Within the inconsistent targets [f,s,k], there was no significant difference between responses to targets in words with majority and minority spellings. We conclude that performance in the phoneme detection task is not necessarily sensitive to orthographic effects, but that salient orthographic manipulation can induce such sensitivity.

Abstract

Research on the exploitation of prosodic information in the recognition of spoken language is reviewed. The research falls into three main areas: the use of prosody in the recognition of spoken words, in which most attention has been paid to the question of whether the prosodic structure of a word plays a role in initial contact with stored lexical representations; the use of prosody in the computation of syntactic structure, in which the resolution of global and local ambiguities has formed the central focus; and the role of prosody in the processing of discourse structure, in which there has been a preponderance of work on the contribution of accentuation and deaccentuation to integration of concepts with an existing discourse model. The review reveals that in each area progress has been made towards new conceptions of prosody's role in processing, and in particular this has involved abandonment of previously held deterministic views of the relationship between prosodic structure and other aspects of linguistic structure

Abstract

Psycholinguists strive to construct a model of human language processing in general. But this does not imply that they should confine their research to universal aspects of linguistic structure, and avoid research on language-specific phenomena. First, even universal characteristics of language structure can only be accurately observed cross-linguistically. This point is illustrated here by research on the role of the syllable in spoken-word recognition, on the perceptual processing of vowels versus consonants, and on the contribution of phonetic assimilation phonemena to phoneme identification. In each case, it is only by looking at the pattern of effects across languages that it is possible to understand the general principle. Second, language-specific processing can certainly shed light on the universal model of language comprehension. This second point is illustrated by studies of the exploitation of vowel harmony in the lexical segmentation of Finnish, of the recognition of Dutch words with and without vowel epenthesis, and of the contribution of different kinds of lexical prosodic structure (tone, pitch accent, stress) to the initial activation of candidate words in lexical access. In each case, aspects of the universal processing model are revealed by analysis of these language-specific effects. In short, the study of spoken-language processing by human listeners requires cross-linguistic comparison.

Abstract

The performance of automatic speech recognition systems is usually assessed in terms of error rate. Human speech recognition produces few errors, but relative difficulty of processing can be assessed via response time techniques. We report the construction of a measure analogous to response time in a machine recognition system. This measure may be compared directly with human response times. We conducted a trial comparison of this type at the phoneme level, including both tense and lax vowels and a variety of consonant classes. The results suggested similarities between human and machine processing in the case of consonants, but differences in the case of vowels.

Abstract

Segmentation of continuous speech into its component words is a nontrivial task for listeners. Previous work has suggested that listeners develop heuristic segmentation procedures based on experience with the structure of their language; for English, the heuristic is that strong syllables (containing full vowels) are most likely to be the initial syllables of lexical words, whereas weak syllables (containing central, or reduced, vowels) are nonword-initial, or, if word-initial, are grammatical words. This hypothesis is here tested against natural and laboratory-induced missegmentations of continuous speech. Precisely the expected pattern is found: listeners erroneously insert boundaries before strong syllables but delete them before weak syllables; boundaries inserted before strong syllables produce lexical words, while boundaries inserted before weak syllables produce grammatical words.

Abstract

Monolingual French speakers employ a syllable-based procedure in speech segmentation; monolingual English speakers use a stress-based segmentation procedure and do not use the syllable-based procedure. In the present study French-English bilinguals participated in segmentation experiments with English and French materials. Their results as a group did not simply mimic the performance of English monolinguals with English language materials and of French monolinguals with French language materials. Instead, the bilinguals formed two groups, defined by forced choice of a dominant language. Only the French-dominant group showed syllabic segmentation and only with French language materials. The English-dominant group showed no syllabic segmentation in either language. However, the English-dominant group showed stress-based segmentation with English language materials; the French-dominant group did not. We argue that rhythmically based segmentation procedures are mutually exclusive, as a consequence of which speech segmentation by bilinguals is, in one respect at least, functionally monolingual.

Abstract

The neural mechanism for selectively tuning in to a target speaker while tuning out the others in a multi-speaker situation (i.e., the cocktail-party effect) remains elusive. Here we addressed this issue by measuring brain activity simultaneously from a listener and from multiple speakers while they were involved in naturalistic conversations. Results consistently show selectively enhanced interpersonal neural synchronization (INS) between the listener and the attended speaker at left temporal–parietal junction, compared with that between the listener and the unattended speaker across different multi-speaker situations. Moreover, INS increases significantly prior to the occurrence of verbal responses, and even when the listener’s brain activity precedes that of the speaker. The INS increase is independent of brain-to-speech synchronization in both the anatomical location and frequency range. These findings suggest that INS underlies the selective process in a multi-speaker situation through neural predictions at the content level but not the sensory level of speech.

Abstract

One of the best-known types of non-independence between languages is caused by genealogical relationships due to descent from a common ancestor. These can be represented by (more or less resolved and controversial) language family trees. In theory, one can argue that language families should be built through the strict application of the comparative method of historical linguistics, but in practice this is not always the case, and there are several proposed classifications of languages into language families, each with its own advantages and disadvantages. A major stumbling block shared by most of them is that they are relatively difficult to use with computational methods, and in particular with phylogenetics. This is due to their lack of standardization, coupled with the general non-availability of branch length information, which encapsulates the amount of evolution taking place on the family tree. In this paper I introduce a method (and its implementation in R) that converts the language classifications provided by four widely-used databases (Ethnologue, WALS, AUTOTYP and Glottolog) intothe de facto Newick standard generally used in phylogenetics, aligns the four most used conventions for unique identifiers of linguistic entities (ISO 639-3, WALS, AUTOTYP and Glottocode), and adds branch length information from a variety of sources (the tree's own topology, an externally given numeric constant, or a distance matrix). The R scripts, input data and resulting Newick trees are available under liberal open-source licenses in a GitHub repository (https://github.com/ddediu/lgfam-newick), to encourage and promote the use of phylogenetic methods to investigate linguistic diversity and its temporal dynamics.

Abstract

Here we re-evaluate our 2013 paper on the antiquity of language (Dediu and Levinson, 2013) in the light of a surge of new information on human evolution in the last half million years. Although new genetic data suggest the existence of some cognitive differences between Neanderthals and modern humans — fully expected after hundreds of thousands of years of partially separate evolution, overall our claims that Neanderthals were fully articulate beings and that language evolution was gradual are further substantiated by the wealth of new genetic, paleontological and archeological evidence briefly reviewed here.

Abstract

In this article we investigate the relation between discourse markers and turn-transition strategies in face-to-face conversations and Instant Messaging (IM), that is, unplanned, real-time, text-based, computer-mediated communication. By means of a quantitative corpus study of utterances containing a discourse marker, we show that utterance-final discourse markers are used more often in IM than in face-to-face conversations. Moreover, utterance-final discourse markers are shown to occur more often at points of turn-transition compared with points of turn-maintenance in both types of conversation. From our results we conclude that the discourse markers in utterance-final position can function as a turn-transition mechanism, signaling that the turn is over and the floor is open to the hearer. We argue that this linguistic turn-taking strategy is essentially similar in face-to-face and IM communication. Our results add to the evidence that communication in IM is more like speech than like writing.

Abstract

Recurrent de novo variants in the TBR1 transcription factor are implicated in the etiology of sporadic autism spectrum disorders (ASD). Disruptions include missense variants located in the T-box DNA-binding domain and previous work has demonstrated that they disrupt TBR1 protein function. Recent screens of thousands of simplex families with sporadic ASD cases uncovered additional T-box variants in TBR1 but their etiological relevance is unclear. We performed detailed functional analyses of de novo missense TBR1 variants found in the T-box of ASD cases, assessing many aspects of protein function, including subcellular localization, transcriptional activity and protein-interactions. Only two of the three tested variants severely disrupted TBR1 protein function, despite in silico predictions that all would be deleterious. Furthermore, we characterized a putative interaction with BCL11A, a transcription factor that was recently implicated in a neurodevelopmental syndrome involving developmental delay and language deficits. Our findings enhance understanding of molecular functions of TBR1, as well as highlighting the importance of functional testing of variants that emerge from next-generation sequencing, to decipher their contributions to neurodevelopmental disorders like ASD.

Abstract

Intellectual disability (ID) is a severe neurodevelopmental disorder with genetically heterogeneous causes. Large-scale sequencing has led to the identification of many gene-disrupting mutations; however, a substantial proportion of cases lack a molecular diagnosis. As such, there remains much to uncover for a complete understanding of the genetic underpinnings of ID. Genetic variants present in non-coding regions of the genome have been highlighted as potential contributors to neurodevelopmental disorders given their role in regulating gene expression. Nevertheless the functional characterization of non-coding variants remains challenging. We describe the identification and characterization of de novo non-coding variation in 3′UTR regulatory regions within an ID cohort of 50 patients. This cohort was previously screened for structural and coding pathogenic variants via CNV, whole exome and whole genome analysis. We identified 44 high-confidence single nucleotide non-coding variants within the 3′UTR regions of these 50 genomes. Four of these variants were located within predicted miRNA binding sites and were thus hypothesised to have regulatory consequences. Functional testing showed that two of the variants interfered with miRNA-mediated regulation of their target genes, AMD1 and FAIM. Both these variants were found in the same individual and their functional consequences may point to a potential role for such variants in intellectual disability.

Abstract

Understanding the genetic factors underlying neurodevelopmental and neuropsychiatric disorders is a major challenge given their prevalence and potential severity for quality of life. While large-scale genomic screens have made major advances in this area, for many disorders the genetic underpinnings are complex and poorly understood. To date the field has focused predominantly on protein coding variation, but given the importance of tightly controlled gene expression for normal brain development and disorder, variation that affects non-coding regulatory regions of the genome is likely to play an important role in these phenotypes. Herein we show the importance of 3 prime untranslated region (3'UTR) non-coding regulatory variants across neurodevelopmental and neuropsychiatric disorders. We devised a pipeline for identifying and functionally validating putatively pathogenic variants from next generation sequencing (NGS) data. We applied this pipeline to a cohort of children with severe specific language impairment (SLI) and identified a functional, SLI-associated variant affecting gene regulation in cells and post-mortem human brain. This variant and the affected gene (ARHGEF39) represent new putative risk factors for SLI. Furthermore, we identified 3′UTR regulatory variants across autism, schizophrenia and bipolar disorder NGS cohorts demonstrating their impact on neurodevelopmental and neuropsychiatric disorders. Our findings show the importance of investigating non-coding regulatory variants when determining risk factors contributing to neurodevelopmental and neuropsychiatric disorders. In the future, integration of such regulatory variation with protein coding changes will be essential for uncovering the genetic causes of complex neurological disorders and the fundamental mechanisms underlying health and disease

Abstract

In conversation, people regularly deal with problems of speaking, hearing, and understanding. We report on a cross-linguistic investigation of the conversational structure of other-initiated repair (also known as collaborative repair, feedback, requests for clarification, or grounding sequences). We take stock of formats for initiating repair across languages (comparable to English huh?, who?, y’mean X?, etc.) and find that different languages make available a wide but remarkably similar range of linguistic resources for this function. We exploit the patterned variation as evidence for several underlying concerns addressed by repair initiation: characterising trouble, managing responsibility, and handling knowledge. The concerns do not always point in the same direction and thus provide participants in interaction with alternative principles for selecting one format over possible others. By comparing conversational structures across languages, this paper contributes to pragmatic typology: the typology of systems of language use and the principles that shape them.

Abstract

Ideophones (also known as expressives or mimetics, and including onomatopoeia) have been systematically studied in linguistics since the 1850s, when they were first described as a lexical class of vivid sensory words in West-African languages. This paper surveys the research history of ideophones, from its roots in African linguistics to its fruits in general linguistics and typology around the globe. It shows that despite a recurrent narrative of marginalisation, work on ideophones has made an impact in many areas of linguistics, from theories of phonological features to typologies of manner and motion, and from sound symbolism to sensory language. Due to their hybrid nature as gradient vocal gestures that grow roots in discrete linguistic systems, ideophones provide opportunities to reframe typological questions, reconsider the role of language ideology in linguistic scholarship, and rethink the margins of language. With ideophones increasingly being brought into the fold of the language sciences, this review synthesises past theoretical insights and empirical findings in order to enable future work to build on them.

Abstract

How a system represents information tightly constrains the kinds of problems it can solve. Humans routinely solve problems that appear to require structured representations of stimulus properties and the relations between them. An account of how we might acquire such representations has central importance for theories of human cognition. We describe how a system can learn structured relational representations from initially unstructured inputs using comparison, sensitivity to time, and a modified Hebbian learning algorithm. We summarize how the model DORA (Discovery of Relations by Analogy) instantiates this approach, which we call predicate learning, as well as how the model captures several phenomena from cognitive development, relational reasoning, and language processing in the human brain. Predicate learning offers a link between models based on formal languages and models which learn from experience and provides an existence proof for how structured representations might be learned in the first place.

Abstract

A commentary on
Transcranial Magnetic Stimulation over Left Inferior Frontal and Posterior Temporal Cortex Disrupts Gesture-Speech Integration

by Zhao, W., Riggs, K., Schindler, I., and Holle, H. (2018). J. Neurosci. 10, 1748–1717. doi: 10.1523/JNEUROSCI.1748-17.2017

Abstract

Previous work revealed that visual semantic information conveyed by gestures can enhance degraded speech comprehension, but the mechanisms underlying these integration processes under adverse listening conditions remain poorly understood. We used MEG to investigate how oscillatory dynamics support speech–gesture integration when integration load is manipulated by auditory (e.g., speech degradation) and visual semantic (e.g., gesture congruency) factors. Participants were presented with videos of an actress uttering an action verb in clear or degraded speech, accompanied by a matching (mixing gesture + “mixing”) or mismatching (drinking gesture + “walking”) gesture. In clear speech, alpha/beta power was more suppressed in the left inferior frontal gyrus and motor and visual cortices when integration load increased in response to mismatching versus matching gestures. In degraded speech, beta power was less suppressed over posterior STS and medial temporal lobe for mismatching compared with matching gestures, showing that integration load was lowest when speech was degraded and mismatching gestures could not be integrated and disambiguate the degraded signal. Our results thus provide novel insights on how low-frequency oscillatory modulations in different parts of the cortex support the semantic audiovisual integration of gestures in clear and degraded speech: When speech is clear, the left inferior frontal gyrus and motor and visual cortices engage because higher-level semantic information increases semantic integration load. When speech is degraded, posterior STS/middle temporal gyrus and medial temporal lobe are less engaged because integration load is lowest when visual semantic information does not aid lexical retrieval and speech and gestures cannot be integrated.

Abstract

During face-to-face communication, listeners integrate speech with gestures. The semantic information conveyed by iconic gestures (e.g., a drinking gesture) can aid speech comprehension in adverse listening conditions. In this magnetoencephalography (MEG) study, we investigated the spatiotemporal neural oscillatory activity associated with gestural enhancement of degraded speech comprehension. Participants watched videos of an actress uttering clear or degraded speech, accompanied by a gesture or not and completed a cued-recall task after watching every video. When gestures semantically disambiguated degraded speech comprehension, an alpha and beta power suppression and a gamma power increase revealed engagement and active processing in the hand-area of the motor cortex, the extended language network (LIFG/pSTS/STG/MTG), medial temporal lobe, and occipital regions. These observed low- and high-frequency oscillatory modulations in these areas support general unification, integration and lexical access processes during online language comprehension, and simulation of and increased visual attention to manual gestures over time. All individual oscillatory power modulations associated with gestural enhancement of degraded speech comprehension predicted a listener's correct disambiguation of the degraded verb after watching the videos. Our results thus go beyond the previously proposed role of oscillatory dynamics in unimodal degraded speech comprehension and provide first evidence for the role of low- and high-frequency oscillations in predicting the integration of auditory and visual information at a semantic level.

Abstract

Native listeners neurally integrate iconic gestures with speech, which can enhance degraded speech comprehension. However, it is unknown how non-native listeners neurally integrate speech and gestures, as they might process visual semantic context differently than natives. We recorded EEG while native and highly-proficient non-native listeners watched videos of an actress uttering an action verb in clear or degraded speech, accompanied by a matching ('to drive'+driving gesture) or mismatching gesture ('to drink'+mixing gesture). Degraded speech elicited an enhanced N400 amplitude compared to clear speech in both groups, revealing an increase in neural resources needed to resolve the spoken input. A larger N400 effect was found in clear speech for non-natives compared to natives, but in degraded speech only for natives. Non-native listeners might thus process gesture more strongly than natives when speech is clear, but need more auditory cues to facilitate access to gestural semantic information when speech is degraded.

Abstract

This paper summarizes the results of a descriptive syntactic category analysis of child English no which reveals that young children use and represent no as a determiner and negatives like no pen as NPs, contra standard analyses.

Abstract

Biologically inspired spiking networks are an important tool to study the nature of computation and cognition in neural systems. In this work, we investigate the representational capacity of spiking networks engaged in an identity mapping task. We compare two schemes for encoding symbolic input, one in which input is injected as a direct current and one where input is delivered as a spatio-temporal spike pattern. We test the ability of networks to discriminate their input as a function of the number of distinct input symbols. We also compare performance using either membrane potentials or filtered spike trains as state variable. Furthermore, we investigate how the circuit behavior depends on the balance between excitation and inhibition, and the degree of synchrony and regularity in its internal dynamics. Finally, we compare different linear methods of decoding population activity onto desired target labels. Overall, our results suggest that even this simple mapping task is strongly influenced by design choices on input encoding, state-variables, circuit characteristics and decoding methods, and these factors can interact in complex ways. This work highlights the importance of constraining computational network models of behavior by available neurobiological evidence.

Abstract

Numerous studies in psychology, cognitive neuroscience and psycholinguistics have used pictures of objects as stimulus materials. Currently, authors engaged in cross-linguistic work or wishing to run parallel studies at multiple sites where different languages are spoken must rely on rather small sets of black-and-white or colored line drawings. These sets are increasingly experienced as being too limited. Therefore, we constructed a new set of 750 colored pictures of concrete concepts. This set, MultiPic, constitutes a new valuable tool for cognitive scientists investigating language, visual perception, memory and/or attention in monolingual or multilingual populations. Importantly, the MultiPic databank has been normed in six different European languages (British English, Spanish, French, Dutch, Italian and German). All stimuli and norms are freely available at http://www.bcbl.eu/databases/multipic

Abstract

While the importance of mental simulation during literary reading has long been recognized, we know little about the factors that determine when, what, and how much readers mentally simulate. Here we investigate the influence of a specific text characteristic, namely verb tense (present vs. past), on mental simulation during literary reading. Verbs usually denote the actions and events that take place in narratives and hence it is hypothesized that verb tense will influence the amount of mental simulation elicited in readers. Although the present tense is traditionally considered to be more “vivid”, this study is one of the first to experimentally assess this claim. We recorded eye-movements while subjects read stories in the past or present tense and collected data regarding self-reported levels of mental simulation, transportation and appreciation. We found no influence of tense on any of the offline measures. The eye-tracking data showed a slightly more complex pattern. Although we did not find a main effect of sensorimotor simulation content on reading times, we were able to link the degree to which subjects slowed down when reading simulation eliciting content to offline measures of attention and transportation, but this effect did not interact with the tense of the story. Unexpectedly, we found a main effect of tense on reading times per word, with past tense stories eliciting longer first fixation durations and gaze durations. However, we were unable to link this effect to any of the offline measures. In sum, this study suggests that tense does not play a substantial role in the process of mental simulation elicited by literary stories.

Abstract

Predictive language processing is often studied by measuring eye movements as participants look at objects on a computer screen while they listen to spoken sentences. The use of this variant of the visual world paradigm has shown that information encountered by a listener at a spoken verb can give rise to anticipatory eye movements to a target object, which is taken to indicate that people predict upcoming words. The ecological validity of such findings remains questionable, however, because these computer experiments used two-dimensional (2D) stimuli that are mere abstractions of real world objects. Here we present a visual world paradigm study in a three-dimensional (3D) immersive virtual reality environment. Despite significant changes in the stimulus material and the different mode of stimulus presentation, language-mediated anticipatory eye movements were observed. These findings thus indicate prediction of upcoming words in language comprehension in a more naturalistic setting where natural depth cues are preserved. Moreover, the results confirm the feasibility of using eye-tracking in rich and multimodal 3D virtual environments.

Abstract

This chapter reviews the computational processes that are responsible for recognizing word forms in the speech stream. We outline the different stages in a processing hierarchy from the extraction of general acoustic features, through speech‐specific prelexical processes, to the retrieval and selection of lexical representations. We argue that two recurring properties of the system as a whole are abstraction and adaptability. We also present evidence for parallel processing of information on different timescales, more specifically that segmental material in the speech stream (its consonants and vowels) is processed in parallel with suprasegmental material (the prosodic structures of spoken words). We consider evidence from both psycholinguistics and neurobiology wherever possible, and discuss how the two fields are beginning to address common computational problems. The challenge for future research in speech perception will be to build an account that links these computational problems, through functional mechanisms that address them, to neurobiological implementation.

Abstract

One of the fundamental issues for a language is its capacity to express
argument structure unambiguously. This study presents evidence
for the emergence and the incremental development of these
basic mechanisms in a newly developing language, Central Taurus
Sign Language. Our analyses identify universal patterns in both the
emergence and development of these mechanisms and in languagespecific
trajectories.

Abstract

FOXP transcription factors play important roles in neurodevelopment, but little is known about how their transcriptional activity is regulated. FOXP proteins cooperatively regulate gene expression by forming homo- and hetero-dimers with each other. Physical associations with other transcription factors might also modulate the functions of FOXP proteins. However, few FOXP-interacting transcription factors have been identified so far. Therefore, we sought to discover additional transcription factors that interact with the brain-expressed FOXP proteins, FOXP1, FOXP2 and FOXP4, through affinity-purifications of protein complexes followed by mass spectrometry. We identified seven novel FOXP-interacting transcription factors (NR2F1, NR2F2, SATB1, SATB2, SOX5, YY1 and ZMYM2), five of which have well-established roles in cortical development. Accordingly, we found that these transcription factors are co-expressed with FoxP2 in the deep layers of the cerebral cortex and also in the Purkinje cells of the cerebellum, suggesting that they may cooperate with the FoxPs to regulate neural gene expression in vivo. Moreover, we demonstrated that etiological mutations of FOXP1 and FOXP2, known to cause neurodevelopmental disorders, severely disrupted the interactions with FOXP-interacting transcription factors. Additionally, we pinpointed specific regions within FOXP2 sequence involved in mediating these interactions. Thus, by expanding the FOXP interactome we have uncovered part of a broader neural transcription factor network involved in cortical development, providing novel molecular insights into the transcriptional architecture underlying brain development and neurodevelopmental disorders.

Abstract

Human language offers rich ways to track, compare, and engage the attentional and epistemic states of interlocutors. While this task is central to everyday communication, our knowledge of the cross-linguistic grammatical means that target such intersubjective coordination has remained basic. In two serialised papers, we introduce the term ‘engagement’ to refer to grammaticalised means for encoding the relative mental directedness of speaker and addressee towards an entity or state of affairs, and describe examples of engagement systems from around the world. Engagement systems express the speaker’s assumptions about the degree to which their attention or knowledge is shared (or not shared) by the addressee. Engagement categories can operate at the level of entities in the here-and-now (deixis), in the unfolding discourse (definiteness vs indefiniteness), entire event-depicting propositions (through markers with clausal scope), and even metapropositions (potentially scoping over evidential values). In this first paper, we introduce engagement and situate it with respect to existing work on intersubjectivity in language. We then explore the key role of deixis in coordinating attention and expressing engagement, moving through increasingly intercognitive deictic systems from those that focus on the the location of the speaker, to those that encode the attentional state of the addressee.

Abstract

Engagement systems encode the relative accessibility of an entity or state of affairs to the speaker and addressee, and are thus underpinned by our social cognitive capacities. In our first foray into engagement (Part 1), we focused on specialised semantic contrasts as found in entity-level deictic systems, tailored to the primal scenario for establishing joint attention. This second paper broadens out to an exploration of engagement at the level of events and even metapropositions, and comments on how such systems may evolve. The languages Andoke and Kogi demonstrate what a canonical system of engagement with clausal scope looks like, symmetrically assigning ‘knowing’ and ‘unknowing’ values to speaker and addressee. Engagement is also found cross-cutting other epistemic categories such as evidentiality, for example where a complex assessment of relative speaker and addressee awareness concerns the source of information rather than the proposition itself. Data from the language Abui reveal that one way in which engagement systems can develop is by upscoping demonstratives, which normally denote entities, to apply at the level of events. We conclude by stressing the need for studies that focus on what difference it makes, in terms of communicative behaviour, for intersubjective coordination to be managed by engagement systems as opposed to other, non-grammaticalised means.

Abstract

A principal components analysis of the ten scales of the Worry Questionnaire revealed the existence of major worry factors or domains of social evaluation and physical threat, and these factors were confirmed in a subsequent item analysis. Those high in trait anxiety had much higher scores on the Worry Questionnaire than those low in trait anxiety, especially on those scales relating to social evaluation. Scores on the Marlowe-Crowne Social Desirability Scale were negatively related to worry frequency. However, groups of low-anxious and repressed individucores did not differ in worry. It was concluded that worry, especals formed on the basis of their trait anxiety and social desirability sially in the social evaluation domain, is of fundamental importance to trait anxiety.

Abstract

In contrast to the large amount of dual-task research investigating the coordination of a linguistic and a nonlinguistic
task, little research has investigated how two linguistic tasks are coordinated. However, such research
would greatly contribute to our understanding of how interlocutors combine speech planning and listening in
conversation. In three dual-task experiments we studied how participants coordinated the processing of an
auditory stimulus (S1), which was either a syllable or a tone, with selecting a name for a picture (S2). Two SOAs,
of 0 ms and 1000 ms, were used. To vary the time required for lexical selection and to determine when lexical
selection took place, the pictures were presented with categorically related or unrelated distractor words. In
Experiment 1 participants responded overtly to both stimuli. In Experiments 2 and 3, S1 was not responded to
overtly, but determined how to respond to S2, by naming the picture or reading the distractor aloud. Experiment
1 yielded additive effects of SOA and distractor type on the picture naming latencies. The presence of semantic
interference at both SOAs indicated that lexical selection occurred after response selection for S1. With respect to
the coordination of S1 and S2 processing, Experiments 2 and 3 yielded inconclusive results. In all experiments,
syllables interfered more with picture naming than tones. This is likely because the syllables activated phonological
representations also implicated in picture naming. The theoretical and methodological implications of the
findings are discussed.

Abstract

Objectives: Our goal was to assess MR image uniformity by investigating aspects influencing said uniformity via a method laid out by the National Electrical Manufacturers Association (NEMA).
Methods: Six metallic materials embedded in a glass phantom were scanned (i.e. Au, Ag, Al, Au-Ag-Pd alloy, Ti and Co-Cr alloy) as well as a reference image. Sequences included spin echo (SE) and gradient echo (GRE) scanned in three planes (i.e. axial, coronal, and sagittal). Moreover, three surface coil types (i.e. head and neck, Brain, and temporomandibular joint coils) and two image correction methods (i.e. surface coil intensity correction or SCIC, phased array uniformity enhancement or PURE) were employed to evaluate their effectiveness on image uniformity. Image uniformity was assessed using the National Electrical Manufacturers Association peak-deviation non-uniformity method.
Results: Results showed that temporomandibular joint coils elicited the least uniform image and brain coils outperformed head and neck coils when metallic materials were present. Additionally, when metallic materials were present, spin echo outperformed gradient echo especially for Co-Cr (particularly in the axial plane). Furthermore, both SCIC and PURE improved image uniformity compared to uncorrected images, and SCIC slightly surpassed PURE when metallic metals were present. Lastly, Co-Cr elicited the least uniform image while other metallic materials generally showed similar patterns (i.e. no significant deviation from images without metallic metals).
Conclusions: Overall, a quantitative understanding of the factors influencing MR image uniformity (e.g. coil type, imaging method, metal susceptibility, and post-hoc correction method) is advantageous to optimize image quality, assists clinical interpretation, and may result in improved medical and dental care.

Abstract

This article presents the ventriloquist paradigm, an innovative method for studying speech processing in dialogue whereby participants interact face-to-face with a confederate who, unbeknownst to them, communicates by playing pre-recorded speech. Results show that the paradigm convinces more participants that the speech is live than a setup without the face-to-face element, and it elicits more interactive conversation than a setup in which participants believe their partner is a computer. By reconciling the ecological validity of a conversational context with full experimental control over phonetic exposure, the paradigm offers a wealth of new possibilities for studying speech processing in interaction.

Abstract

Between 2 and 5% of children who are otherwise unimpaired have significant difficulties in acquiring expressive and/or receptive language, despite adequate intelligence and opportunity. While twin studies indicate a significant role for genetic factors in developmental disorders of speech and language, the majority of families segregating such disorders show complex patterns of inheritance, and are thus not amenable for conventional linkage analysis. A rare exception is the KE family, a large three-generation pedigree in which approximately half of the members are affected with a severe speech and language disorder which appears to be transmitted as an autosomal dominant monogenic trait. This family has been widely publicised as suffering primarily from a defect in the use of grammatical suffixation rules, thus supposedly supporting the existence of genes specific to grammar. The phenotype, however, is broader in nature, with virtually every aspect of grammar and of language affected. In addition, affected members have a severe orofacial dyspraxia, and their speech is largely incomprehensible to the naive listener. We initiated a genome-wide search for linkage in the KE family and have identified a region on chromosome 7 which co-segregates with the speech and language disorder (maximum lod score = 6.62 at theta = 0.0), confirming autosomal dominant inheritance with full penetrance. Further analysis of microsatellites from within the region enabled us to fine map the locus responsible (designated SPCH1) to a 5.6-cM interval in 7q31, thus providing an important step towards its identification. Isolation of SPCH1 may offer the first insight into the molecular genetics of the developmental process that culminates in speech and language.

Abstract

The human Xp11.23-p11.22 interval has been implicated in several inherited diseases including Wiskott-Aldrich syndrome; three forms of X-linked hypercalciuric nephrolithiaisis; and the eye disorders retinitis pigmentosa 2, congenital stationary night blindness, and Aland Island eye disease. In constructing YAC contigs spanning Xp11. 23-p11.22, we have previously shown that the region around the synaptophysin (SYP) gene is refractory to cloning in YACs, but highly stable in cosmids. Preliminary analysis of the latter suggested that this might reflect a high density of coding sequences and we therefore undertook the complete sequencing of a SYP-containing cosmid. Sequence data were extensively analyzed using computer programs such as CENSOR (to mask repeats), BLAST (for homology searches), and GRAIL and GENE-ID (to predict exons). This revealed the presence of 29 putative exons, organized into three genes, in addition to the 7 exons of the complete SYP coding region, all mapping within a 44-kb interval. Two genes are novel, one (CACNA1F) showing high homology to alpha1 subunits of calcium channels, the other (LMO6) encoding a product with significant similarity to LIM-domain proteins. RT-PCR and Northern blot studies confirmed that these loci are indeed transcribed. The third locus is the previously described, but not previously localized, A4 differentiation-dependent gene. Given that the intron-exon boundaries predicted by the analysis are consistent with previous information where available, we have been able to suggest the genomic organization of the novel genes with some confidence. The region has an elevated GC content (>53%), and we identified CpG islands associated with the 5' ends of SYP, A4, and LMO6. The order of loci was Xpter-A4-LMO6-SYP-CACNA1F-Xcen, with intergenic distances ranging from approximately 300 bp to approximately 5 kb. The density of transcribed sequences in this area (>80%) is comparable to that found in the highly gene-rich chromosomal band Xq28. Further studies may aid our understanding of the long-range organization surrounding such gene-enriched regions.

Abstract

The Cha’palaa language of Ecuador (Barbacoan) features verbal morphology for marking knowledge-based categories that, in usage, show a variant of the cross-linguistically recurrent pattern of ‘egophoric distribution': specific forms associate with speakers in contrast to others in statements and with addressees in contrast to others in questions. These are not person markers, but rather are used by speakers to portray their involvement in states of affairs as active, agentive participants (ego) versus other types of involvement (non-ego). They interact with person and argument structure, but through pragmatic ‘person sensitivities’ rather than through grammatical agreement. Not only does this pattern appear in verbal morphology, it also can be observed in alternations of predicate construction types and case alignment, helping to show how egophoric marking is a pervasive element of Cha'palaa's linguistic system. This chapter gives a first account of egophoricity in Cha’palaa, beginning with a discussion of person sensitivity, egophoric distribution, and issues of flexibility of marking with respect to degree of volition or control. It then focuses on a set of intransitive experiencer (or ‘endopathic') predicates that refer to internal states which mark egophoric values for the undergoer role, not the actor role, showing ‘quirky’ accusative marking instead of nominative case. It concludes with a summary of how egophoricity in Cha'palaa interacts with issues of argument structure in comparison to a language with person agreement, here represented by examples from Cha’palaa’s neighbor Ecuadorian Highland Quechua.

Abstract

It has long been claimed that there is no lexical field of smell, and that smell is of too little validity to be expressed in grammar. We demonstrate both claims are false. The Cha'palaa language (Ecuador) has at least 15 abstract smell terms, each of which is formed using a type of classifier previously thought not to exist. Moreover, using conversational corpora we show that Cha'palaa speakers also talk about smell more than Imbabura Quechua and English speakers. Together, this shows how language and social interaction may jointly reflect distinct cultural orientations towards sensory experience in general and olfaction in particular.

Abstract

Gratitude is argued to have evolved to motivate and maintain social reciprocity among people, and to be linked to a wide range of positive effects — social, psychological, and even physical. But is socially reciprocal behaviour dependent on the expression of gratitude, for example by saying "thank you" as in English? Current research has not included cross-cultural elements, and has tended to conflate gratitude as an emotion with gratitude as a linguistic practice, as might appear to be the case in English. Here we ask to what extent people actually express gratitude in different societies by focussing on episodes of everyday life where someone obtains a good, service, or support from another, and comparing these episodes across eight languages from five continents. What we find is that expressions of gratitude in these episodes are remarkably rare, suggesting that social reciprocity in everyday life relies on tacit understandings of people’s rights and duties surrounding mutual assistance and collaboration. At the same time, we also find minor cross-cultural variation, with slightly higher rates in Western European languages English and Italian, showing that universal tendencies of social reciprocity should not be conflated with more culturally variable practices of expressing gratitude. Our study complements previous experimental and culture-specific research on social reciprocity with a systematic comparison of audiovisual corpora of naturally occurring social interaction from different cultures from around the world.

Abstract

Patients with stroke offer a unique window into understanding human brain function. Mapping stroke lesions poses several challenges due to the complexity of the lesion anatomy and the mechanisms causing local and remote disruption on brain networks. In this prospective longitudinal study, we compare standard and advanced approaches to white matter lesion mapping applied to acute stroke patients with aphasia. Eighteen patients with acute left hemisphere stroke were recruited and scanned within two weeks from symptom onset. Aphasia assessment was performed at baseline and six-month follow-up. Structural and diffusion MRI contrasts indicated an area of maximum overlap in the anterior external/extreme capsule with diffusion images showing a larger overlap extending into posterior perisylvian regions. Anatomical predictors of recovery included damage to ipsilesional tracts (as shown by both structural and diffusion images) and contralesional tracts (as shown by diffusion images only). These findings indicate converging results from structural and diffusion lesion mapping methods but also clear differences between the two approaches in their ability to identify predictors of recovery outside the lesioned regions.

Abstract

Structural imaging based on computerized tomography (CT) and magnetic resonance imaging (MRI) has progressively replaced traditional post‐mortem studies in the process of identifying the neuroanatomical basis of language. In the clinical setting, the information provided by structural imaging has been used to confirm the exact diagnosis and formulate an individualized treatment plan. In the research arena, neuroimaging has permitted to understand neuroanatomy at the individual and group level. The possibility to obtain quantitative measures of lesions has improved correlation analyses between severity of symptoms, lesion load, and lesion location. More recently, the development of structural imaging based on diffusion MRI has provided valid solutions to two major limitations of more conventional imaging. In stroke patients, diffusion can visualize early changes due to a stroke that are otherwise not detectable with more conventional structural imaging, with important implications for the clinical management of acute stroke patients. Beyond the sensitivity to early changes, diffusion imaging tractography presents the possibility of visualizing the trajectories of individual white matter pathways connecting distant regions. A pathway analysis based on tractography is offering a new perspective in neurolinguistics. First, it permits to formulate new anatomical models of language function in the healthy brain and allows to directly test these models in the human population without any reliance on animal models. Second, by defining the exact location of the damage to specific white matter connections we can understand the contribution of different mechanisms to the emergence of language deficits (e.g., cortical versus disconnection mechanisms). Finally, a better understanding of the anatomical variability of different language networks is helping to identify new anatomical predictors of language recovery. In this chapter we will focus on the principles of structural MRI and, in particular, diffusion imaging and tractography and present examples of how these methods have informed our understanding of variance in language performances in the healthy brain and language deficits in patient populations.

Abstract

The information we obtain from how speakers sound—for example their accent—affects how we interpret the messages they convey. A clear example is foreign accented speech, where reduced intelligibility and speaker's social categorization (out-group member) affect memory and the credibility of the message (e.g., less trustworthiness). In the present study, we go one step further and ask whether evaluations of messages are also affected by regional accents—accents from a different region than the listener. In the current study, we report results from three experiments on immediate memory recognition and immediate credibility assessments as well as the illusory truth effect. These revealed no differences between messages conveyed in local—from the same region as the participant—and regional accents—from native speakers of a different country than the participants. Our results suggest that when the accent of a speaker has high intelligibility, social categorization by accent does not seem to negatively affect how we treat the speakers' messages.

Abstract

The information we obtain from how speakers sound—for example their accent—affects how we interpret the
messages they convey. A clear example is foreign accented speech, where reduced intelligibility and speaker's
social categorization (out-group member) affect memory and the credibility of the message (e.g., less trust-
worthiness). In the present study, we go one step further and ask whether evaluations of messages are also
affected by regional accents—accents from a different region than the listener. In the current study, we report
results from three experiments on immediate memory recognition and immediate credibility assessments as well
as the illusory truth effect. These revealed no differences between messages conveyed in local—from the same
region as the participant—and regional accents—from native speakers of a different country than the partici-
pants. Our results suggest that when the accent of a speaker has high intelligibility, social categorization by
accent does not seem to negatively affect how we treat the speakers' messages.

Abstract

The aim of the present fMRI study was to investigate whether typical and dyslexic adult readers differed in the neural correlates of audiovisual speech processing. We tested for Blood Oxygen-Level Dependent (BOLD) activity differences between these two groups in a 1-back task, as they processed written (word, illegal consonant strings) and spoken (auditory, visual and audiovisual) stimuli. When processing written stimuli, dyslexic readers showed reduced activity in the supramarginal gyrus, a region suggested to play an important role in phonological processing, but only when they processed strings of consonants, not when they read words. During the speech perception tasks, dyslexic readers were only slower than typical readers in their behavioral responses in the visual speech condition. Additionally, dyslexic readers presented reduced neural activation in the auditory, the visual, and the audiovisual speech conditions. The groups also differed in terms of superadditivity, with dyslexic readers showing decreased neural activation in the regions of interest. An additional analysis focusing on vision-related processing during the audiovisual condition showed diminished activation for the dyslexic readers in a fusiform gyrus cluster. Our results thus suggest that there are differences in audiovisual speech processing between dyslexic and normal readers. These differences might be explained by difficulties in processing the unisensory components of audiovisual speech, more specifically, dyslexic readers may benefit less from visual information during audiovisual speech processing than typical readers. Given that visual speech processing supports the development of phonological skills fundamental in reading, differences in processing of visual speech could contribute to differences in reading ability between typical and dyslexic readers.

Abstract

Results from a recent neuroimaging study on spoken sentence comprehension have been interpreted as evidence for cortical entrainment to hierarchical syntactic structure. We present a simple computational model that predicts the power spectra from this study, even
though the model's linguistic knowledge is restricted to the lexical level, and word-level representations are not combined into higher-level units (phrases or sentences). Hence, the
cortical entrainment results can also be explained from the lexical properties of the stimuli, without recourse to hierarchical syntax.

Abstract

Speaking and listening are complex tasks that we perform on a daily basis, almost without conscious effort. Interestingly, speaking almost never occurs without listening: whenever we speak, we at least hear our own speech. The research in this thesis is concerned with how the perception of our own speech influences our speaking behavior. We show that unconsciously, we actively monitor this auditory feedback of our own speech. This way, we can efficiently take action and adapt articulation when an error occurs and auditory feedback does not correspond to our expectation. Processing the auditory feedback of our speech does not, however, automatically affect speech production. It is subject to a number of constraints. For example, we do not just track auditory feedback, but also its consistency. If auditory feedback is more consistent over time, it has a stronger influence on speech production. In addition, we investigated how auditory feedback during speech is processed in the brain, using magnetoencephalography (MEG). The results suggest the involvement of a broad cortical network including both auditory and motor-related regions. This is consistent with the view that the auditory center of the brain is involved in comparing auditory feedback to our expectation of auditory feedback. If this comparison yields a mismatch, motor-related regions of the brain can be recruited to alter the ongoing articulations.

Abstract

When talking, speakers continuously monitor and use the auditory feedback of their own voice to control and inform speech production processes. When speakers are provided with auditory feedback that is perturbed in real time, most of them compensate for this by opposing the feedback perturbation. But some speakers follow the perturbation. In the current study, we investigated whether the state of the speech production system at perturbation onset may determine what type of response (opposing or following) is given. The results suggest that whether a perturbation-related response is opposing or following depends on ongoing fluctuations of the production system: It initially responds by doing the opposite of what it was doing. This effect and the non-trivial proportion of following responses suggest that current production models are inadequate: They need to account for why responses to unexpected sensory feedback depend on the production-system’s state at the time of perturbation.

Abstract

Speaking is a complex motor skill which requires near instantaneous integration of sensory and motor-related information. Current theory hypothesizes a complex interplay between motor and auditory processes during speech production, involving the online comparison of the speech output with an internally generated forward model. To examine the neural correlates of this intricate interplay between sensory and motor processes, the current study uses altered auditory feedback (AAF) in combination with magnetoencephalography (MEG). Participants vocalized the vowel/e/and heard auditory feedback that was temporarily pitch-shifted by only 25 cents, while neural activity was recorded with MEG. As a control condition, participants also heard the recordings of the same auditory feedback that they heard in the first half of the experiment, now without vocalizing. The participants were not aware of any perturbation of the auditory feedback. We found auditory cortical areas responded more strongly to the pitch shifts during vocalization. In addition, auditory feedback perturbation resulted in spectral power increases in the θ and lower β bands, predominantly in sensorimotor areas. These results are in line with current models of speech production, suggesting auditory cortical areas are involved in an active comparison between a forward model's prediction and the actual sensory input. Subsequently, these areas interact with motor areas to generate a motor response. Furthermore, the results suggest that θ and β power increases support auditory-motor interaction, motor error detection and/or sensory prediction processing.

Abstract

We analyze the problem of response suggestion in a closed domain along a real-world scenario of a digital library. We present a text-processing pipeline to generate question-answer pairs from chat transcripts. On this limited amount of training data, we compare retrieval-based, conditioned-generation, and dedicated representation learning approaches for response suggestion. Our results show that retrieval-based methods that strive to find similar, known contexts are preferable over parametric approaches from the conditioned-generation family, when the training data is limited. We, however, identify a specific representation learning approach that is competitive to the retrieval-based approaches despite the training data limitation.

Abstract

We present multi-modal adversarial autoencoders for recommendation and evaluate them on two different tasks: citation recommendation and subject label recommendation. We analyze the effects of adversarial regularization, sparsity, and different input modalities. By conducting 408 experiments, we show that adversarial regularization consistently improves the performance of autoencoders for recommendation. We demonstrate, however, that the two tasks differ in the semantics of item co-occurrence in the sense that item co-occurrence resembles relatedness in case of citations, yet implies diversity in case of subject labels. Our results reveal that supplying the partial item set as input is only helpful, when item co-occurrence resembles relatedness. When facing a new recommendation task it is therefore crucial to consider the semantics of item co-occurrence for the choice of an appropriate model.

Abstract

Resource-constrained models of language processing predict that perceptual simulation during language understanding would be compromised by sensory limitations (such as reading text in unfamiliar/difficult font), whereas strong versions of embodied theories of language would predict that simulating perceptual symbols in language would not be impaired even under sensory-constrained situations. In 2 experiments, sensory decoding difficulty was manipulated by using easy and hard fonts to study perceptual simulation during sentence reading (Zwaan, Stanfield, & Yaxley, 2002). Results indicated that simulating perceptual symbols in language was not compromised by surface-form decoding challenges such as difficult font, suggesting relative resilience of embodied language processing in the face of certain sensory constraints. Further implications for learning from text and individual differences in language processing will be discussed