Publications

Abstract

Studies of perception have long shown that the brain adds information to its sensory analysis of the physical environment. A touchstone example for humans is language use: to comprehend a physical signal like speech, the brain must add linguistic knowledge, including syntax. Yet, syntactic rules and representations are widely assumed to be atemporal (i.e., abstract and not bound by time), so they must be translated into time-varying signals for speech comprehension and production. Here, we test 3 different models of the temporal spell-out of syntactic structure against brain activity of people listening to Dutch stories: an integratory bottom-up parser, a predictive top-down parser, and a mildly predictive left-corner parser. These models build exactly the same structure but differ in when syntactic information is added by the brain—this difference is captured in the (temporal distribution of the) complexity metric “incremental node count.” Using temporal response function models with both acoustic and information-theoretic control predictors, node counts were regressed against source-reconstructed delta-band activity acquired with magnetoencephalography. Neural dynamics in left frontal and temporal regions most strongly reflect node counts derived by the top-down method, which postulates syntax early in time, suggesting that predictive structure building is an important component of Dutch sentence comprehension. The absence of strong effects of the left-corner model further suggests that its mildly predictive strategy does not represent Dutch language comprehension well, in contrast to what has been found for English. Understanding when the brain projects its knowledge of syntax onto speech, and whether this is done in language-specific ways, will inform and constrain the development of mechanistic models of syntactic structure building in the brain.

Abstract

Face-to-face communication is not only about ‘what’ is said but also ‘how’ it is said, both in speech and bodily signals. Beat gestures are rhythmic hand movements that typically accompany prosodic prominence in con-versation. Yet, it is still unclear how beat gestures influence language comprehension. On the one hand, beat gestures may share the same functional role of focus markers as prosodic prominence. Accordingly, they would drive attention towards the concurrent speech and highlight its content. On the other hand, beat gestures may trigger inferences of high speaker confidence, generate the expectation that the sentence content is correct and thereby elicit the commitment to the truth of the statement. This study directly disentangled the two hypotheses by evaluating additive and interactive effects of prosodic prominence and beat gestures on language comprehension. Participants watched videos of a speaker uttering sentences and judged whether each sentence was true or false. Sentences sometimes contained a world knowledge violation that may go unnoticed (‘semantic illusion’). Combining beat gestures with prosodic prominence led to a higher degree of semantic illusion, making more world knowledge violations go unnoticed during language comprehension. These results challenge current theories proposing that beat gestures are visual focus markers. To the contrary, they suggest that beat gestures automatically trigger inferences of high speaker confidence and thereby elicit the commitment to the truth of the statement, in line with Grice’s cooperative principle in conversation. More broadly, our findings also highlight the influence of metacognition on language comprehension in face-to-face ommunication.

Abstract

The increased interest in developing next-gen social robots has raised questions about the factors affecting the perception of robot emotions. This study investigates the impact of robot appearances (human-like, mechanical) and face regions (full-face, eye-region) on human perception of robot emotions. A between-subjects user study (N = 305) was conducted where participants were asked to identify the emotions being displayed in videos of robot faces, as well as a human baseline. Our findings reveal three important insights for effective social robot face design in Human-Robot Interaction (HRI): Firstly, robots equipped with a back-projected, fully animated face – regardless of whether they are more human-like or more mechanical-looking – demonstrate a capacity for emotional expression comparable to that of humans. Secondly, the recognition accuracy of emotional expressions in both humans and robots declines when only the eye region is visible. Lastly, within the constraint of only the eye region being visible, robots with more human-like features significantly enhance emotion recognition.

Abstract

To date, research has reliably shown that language can engage and modify perceptual processes in a top-down manner. However, our understanding of the cognitive and neural mechanisms underlying such top-down influences is still under debate. In this review, we provide an overview of findings from literature investigating the organization of semantic networks in the brain (spontaneous engagement of the visual system while processing linguistic information), and linguistic cueing studies (looking at the immediate effects of language on the perception of a visual target), in an effort to isolate such mechanisms. Additionally, we connect the findings from linguistic cueing studies to those reported in (nonlinguistic) literature on priors in perception, in order to find commonalities in neural processes allowing for top-down influences on perception. In doing so, we discuss the effects of language on perception in the context of broader, general cognitive and neural principles. Finally, we propose a way forward in the study of linguistic influences on perception.

Abstract

Prosody underpins various linguistic domains ranging from semantics and syntax to discourse. For instance, prosodic information in the form of lexical stress modifies meanings and, as such, syntactic contexts of words as in Turkish kaz-má "pickaxe" (noun) versus káz-ma "do not dig" (imperative). Likewise, prosody indicates the focused constituent of an utterance as the noun phrase filling the wh-spot in a dialogue like What did you eat? I ate----. In the present study, we investigated the relevance of such prosodic variations for discourse comprehension in Turkish. We aimed at answering how lexical stress and prosodic focus mismatches on critical noun phrases-resulting in grammatical anomalies involving both semantics and syntax and discourse-level anomalies, respectively-affect the perceived correctness of an answer to a question in a given context. To that end, 80 native speakers of Turkish, 40 participating in a psychometric experiment and 40 participating in an EEG experiment, were asked to judge the acceptability of prosodic mismatches that occur either separately or concurrently. Psychometric results indicated that lexical stress mismatch led to a lower correctness score than prosodic focus mismatch, and combined mismatch received the lowest score. Consistent with the psychometric data, EEG results revealed an N400 effect to combined mismatch, and this effect was followed by a P600 response to lexical stress mismatch. Conjointly, these results suggest that every source of prosodic information is immediately available and codetermines the interpretation of an utterance; however, semantically and syntactically relevant lexical stress information is assigned more significance by the language comprehension system compared with prosodic focus information.

Abstract

It has long been recognised that phrases and sentences are organised hierarchically, but many computational models of language treat them as sequences of words without computing constituent structure. Against this background, we conducted two experiments which showed that participants interpret ambiguous noun phrases, such as second blue ball, in terms of their abstract hierarchical structure rather than their linear surface order. When a neural network model was tested on this task, it could simulate such “hierarchical” behaviour. However, when we changed the training data such that they were not entirely unambiguous anymore, the model stopped generalising in a human-like way. It did not systematically generalise to novel items, and when it was trained on ambiguous trials, it strongly favoured the linear interpretation. We argue that these models should be endowed with a bias to make generalisations over hierarchical structure in order to be cognitively adequate models of human language.

Abstract

Recent research has established that cortical activity “tracks” the presentation rate of syntactic phrases in continuous speech, even though phrases are abstract units that do not have direct correlates in the acoustic signal. We investigated whether cortical tracking of phrase structures is modulated by the extent to which these structures compositionally determine meaning. To this end, we recorded electroencephalography (EEG) of 38 native speakers who listened to naturally spoken Dutch stimuli in different conditions, which parametrically modulated the degree to which syntactic structure and lexical semantics determine sentence meaning. Tracking was quantified through mutual information between the EEG data and either the speech envelopes or abstract annotations of syntax, all of which were filtered in the frequency band corresponding to the presentation rate of phrases (1.1–2.1 Hz). Overall, these mutual information analyses showed stronger tracking of phrases in regular sentences than in stimuli whose lexical-syntactic content is reduced, but no consistent differences in tracking between sentences and stimuli that contain a combination of syntactic structure and lexical content. While there were no effects of compositional meaning on the degree of phrase-structure tracking, analyses of event-related potentials elicited by sentence-final words did reveal meaning-induced differences between conditions. Our findings suggest that cortical tracking of structure in sentences indexes the internal generation of this structure, a process that is modulated by the properties of its input, but not by the compositional interpretation of its output.

Abstract

Listening to speech is difficult in noisy environments, and is even harder when the interfering noise consists of intelligible speech as compared to unintelligible sounds. This suggests that the competing linguistic information interferes with the neural processing of target speech. Interference could either arise from a degradation of the neural representation of the target speech, or from increased representation of distracting speech that enters in competition with the target speech. We tested these alternative hypotheses using magnetoencephalography (MEG) while participants listened to a target clear speech in the presence of distracting noise-vocoded speech. Crucially, the distractors were initially unintelligible but became more intelligible after a short training session. Results showed that the comprehension of the target speech was poorer after training than before training. The neural tracking of target speech in the delta range (1–4 Hz) reduced in strength in the presence of a more intelligible distractor. In contrast, the neural tracking of distracting signals was not significantly modulated by intelligibility. These results suggest that the presence of distracting speech signals degrades the linguistic representation of target speech carried by delta oscillations.

Abstract

The neurobiology of sentence production has been largely understudied compared to the neurobiology of sentence comprehension, due to difficulties with experimental control and motion-related artifacts in neuroimaging. We studied the neural response to constituents of increasing size and specifically focused on the similarities and differences in the production and comprehension of the same stimuli. Participants had to either produce or listen to stimuli in a gradient of constituent size based on a visual prompt. Larger constituent sizes engaged the left inferior frontal gyrus (LIFG) and middle temporal gyrus (LMTG) extending to inferior parietal areas in both production and comprehension, confirming that the neural resources for syntactic encoding and decoding are largely overlapping. An ROI analysis in LIFG and LMTG also showed that production elicited larger responses to constituent size than comprehension and that the LMTG was more engaged in comprehension than production, while the LIFG was more engaged in production than comprehension. Finally, increasing constituent size was characterized by later BOLD peaks in comprehension but earlier peaks in production. These results show that syntactic encoding and parsing engage overlapping areas, but there are asymmetries in the engagement of the language network due to the specific requirements of production and comprehension.

Abstract

Understanding spoken language requires transforming ambiguous acoustic streams into a hierarchy of representations, from phonemes to meaning. It has been suggested that the brain uses prediction to guide the interpretation of incoming input. However, the role of prediction in language processing remains disputed, with disagreement about both the ubiquity and representational nature of predictions. Here, we address both issues by analyzing brain recordings of participants listening to audiobooks, and using a deep neural network (GPT-2) to precisely quantify contextual predictions. First, we establish that brain responses to words are modulated by ubiquitous predictions. Next, we disentangle model-based predictions into distinct dimensions, revealing dissociable neural signatures of predictions about syntactic category (parts of speech), phonemes, and semantics. Finally, we show that high-level (word) predictions inform low-level (phoneme) predictions, supporting hierarchical predictive processing. Together, these results underscore the ubiquity of prediction in language processing, showing that the brain spontaneously predicts upcoming language at multiple levels of abstraction.

Abstract

Typical adults read remarkably quickly. Such fast reading is facilitated by brain processes that are sensitive to both word frequency and contextual constraints. It is debated as to whether these attributes have additive or interactive effects on language processing in the brain. We investigated this issue by analysing existing magnetoencephalography data from 99 participants reading intact and scrambled sentences. Using a cross-validated model comparison scheme, we found that lexical frequency predicted the word-by-word elicited MEG signal in a widespread cortical network, irrespective of sentential context. In contrast, index (ordinal word position) was more strongly encoded in sentence words, in left front-temporal areas. This confirms that frequency influences word processing independently of predictability, and that contextual constraints affect word-by-word brain responses. With a conservative multiple comparisons correction, only the interaction between lexical frequency and surprisal survived, in anterior temporal and frontal cortex, and not between lexical frequency and entropy, nor between lexical frequency and index. However, interestingly, the uncorrected index*frequency interaction revealed an effect in left frontal and temporal cortex that reversed in time and space for intact compared to scrambled sentences. Finally, we provide evidence to suggest that, in sentences, lexical frequency and predictability may independently influence early (<150ms) and late stages of word processing, but interact during later stages of word processing (>150-250ms), thus helping to converge previous contradictory eye-tracking and electrophysiological literature. Current neuro-cognitive models of reading would benefit from accounting for these differing effects of lexical frequency and predictability on different stages of word processing.

Abstract

Traditional experiments indicate that prediction is important for efficient speech processing. In three virtual reality visual world paradigm experiments, we tested whether such findings hold in naturalistic settings (Experiment 1) and provided novel insights into whether disfluencies in speech (repairs/hesitations) inform one’s predictions in rich environments (Experiments 2–3). Experiment 1 supports that listeners predict upcoming speech in naturalistic environments, with higher proportions of anticipatory target fixations in predictable compared to unpredictable trials. In Experiments 2–3, disfluencies reduced anticipatory fixations towards predicted referents, compared to conjunction (Experiment 2) and fluent (Experiment 3) sentences. Unexpectedly, Experiment 2 provided no evidence that participants made new predictions from a repaired verb. Experiment 3 provided novel findings that fixations towards the speaker increase upon hearing a hesitation, supporting current theories of how hesitations influence sentence processing. Together, these findings unpack listeners’ use of visual (objects/speaker) and auditory (speech/disfluencies) information when predicting upcoming words.

Abstract

Introduction: Mood is a constant in our daily life and can permeate all levels of cognition. We examined whether and how mood influences the processing of discourse content that is relatively neutral and not loaded with emotion. During discourse processing, readers have to constantly strike a balance between what they know in long term memory and what the current discourse is about. Our general hypothesis is that mood states would affect this balance. We hypothesized that readers in a positive mood would rely more on default world knowledge, whereas readers in a negative mood would be more inclined to analyze the details in the current discourse.

Methods: Participants were put in a positive and a negative mood via film clips, one week apart. In each session, after mood manipulation, they were presented with sentences in discourse materials. We created sentences such as “With the lights on you can see...” that end with critical words (CWs) “more” or “less”, where general knowledge supports “more”, not “less”. We then embedded each of these sentences in a wider discourse that does/does not support the CWs (a story about driving in the night vs. stargazing). EEG was recorded throughout.

Results: The results showed that first, mood manipulation was successful in that there was a significant mood difference between sessions. Second, mood did not modulate the N400 effects. Participants in both moods detected outright semantic violations and allowed world knowledge to be overridden by discourse context. Third, mood modulated the LPC (Late Positive Component) effects, distributed in the frontal region. In negative moods, the LPC was sensitive to one-level violation. That is, CWs that were supported by only world knowledge, only discourse, and neither, elicited larger frontal LPCs, in comparison to the condition where CWs were supported by both world knowledge and discourse.

Discussion: These results suggest that mood does not influence all processes involved in discourse processing. Specifically, mood does not influence lexical-semantic retrieval (N400), but it does influence elaborative processes for sensemaking (P600) during discourse processing. These results advance our understanding of the impact and time course of mood on discourse.

Abstract

The ability to comprehend phrases is an essential integrative property of the brain. Here we evaluate the neural processes that enable the transition from single word processing to a minimal compositional scheme. Previous research has reported conflicting timing effects of composition, and disagreement persists with respect to inferior frontal and posterior temporal contributions. To address these issues, 19 patients (10 male, 19 female) implanted with penetrating depth or surface subdural intracranial electrodes heard auditory recordings of adjective-noun, pseudoword-noun and adjective-pseudoword phrases and judged whether the phrase matched a picture. Stimulus-dependent alterations in broadband gamma activity, low frequency power and phase-locking values across the language-dominant left hemisphere were derived. This revealed a mosaic located on the lower bank of the posterior superior temporal sulcus (pSTS), in which closely neighboring cortical sites displayed exclusive sensitivity to either lexicality or phrase structure, but not both. Distinct timings were found for effects of phrase composition (210–300 ms) and pseudoword processing (approximately 300–700 ms), and these were localized to neighboring electrodes in pSTS. The pars triangularis and temporal pole encoded anticipation of composition in broadband low frequencies, and both regions exhibited greater functional connectivity with pSTS during phrase composition. Our results suggest that the pSTS is a highly specialized region comprised of sparsely interwoven heterogeneous constituents that encodes both lower and higher level linguistic features. This hub in pSTS for minimal phrase processing may form the neural basis for the human-specific computational capacity for forming hierarchically organized linguistic structures.

Abstract

This study investigated two questions. One is: To what degree is sentence processing beyond single words independent of the input modality (speech vs. reading)? The second question is: Which parts of the network recruited by both modalities is sensitive to syntactic complexity? These questions were investigated by having more than 200 participants read or listen to well-formed sentences or series of unconnected words. A largely left-hemisphere frontotemporoparietal network was found to be supramodal in nature, i.e., independent of input modality. In addition, the left inferior frontal gyrus (LIFG) and the left posterior middle temporal gyrus (LpMTG) were most clearly associated with left-branching complexity. The left anterior temporal lobe (LaTL) showed the greatest sensitivity to sentences that differed in right-branching complexity. Moreover, activity in LIFG and LpMTG increased from sentence onset to end, in parallel with an increase of the left-branching complexity. While LIFG, bilateral anterior temporal lobe, posterior MTG, and left inferior parietal lobe (LIPL) all contribute to the supramodal unification processes, the results suggest that these regions differ in their respective contributions to syntactic complexity related processing. The consequences of these findings for neurobiological models of language processing are discussed.

Abstract

Vocal learning, the ability to produce modified vocalizations via learning from acoustic signals, is a key trait in the evolution of speech. While extensively studied in songbirds, mammalian models for vocal learning are rare. Bats present a promising study system given their gregarious natures, small size, and the ability of some species to be maintained in captive colonies. We utilize the pale spear-nosed bat (Phyllostomus discolor) and report advances in establishing this species as a tractable model for understanding vocal learning. We have taken an interdisciplinary approach, aiming to provide an integrated understanding across genomics (Part I), neurobiology (Part II), and transgenics (Part III). In Part I, we generated new, high-quality genome annotations of coding genes and noncoding microRNAs to facilitate functional and evolutionary studies. In Part II, we traced connections between auditory-related brain regions and reported neuroimaging to explore the structure of the brain and gene expression patterns to highlight brain regions. In Part III, we created the first successful transgenic bats by manipulating the expression of FoxP2, a speech-related gene. These interdisciplinary approaches are facilitating a mechanistic and evolutionary understanding of mammalian vocal learning and can also contribute to other areas of investigation that utilize P. discolor or bats as study species.