Max Planck Institute for Psycholinguistics

Disfluencies in speech frequently occur before the production of longer and more complex speech content. Listeners are thought to use the distribution of disfluencies in the comprehension of speech to inform their predictions. Here, we investigated whether the presence of disfluencies in speech affects word processing also in naturalistic listening conditions. Participants (n = 36) listened to the spoken recall of the events of a television series while undergoing fMRI. We modelled word processing effort using parametric modulations for word length, frequency, entropy, as well as surprisal and presence/absence of a disfluency. To investigate the effects of disfluencies on word processing, we tested the interaction between disfluency and frequency, and disfluency and surprisal. Words preceded by a disfluency were associated with increased activity in the left and right superior temporal gyrus (STG). Lower word frequency was associated with an increase in activity in the left mid STG. Increased word surprisal elicited a similar distribution of activity, with bilateral superior temporal activation. The effect of surprisal was reduced after a disfluency in a cluster in the left posterior temporal lobe, while the effect of frequency increased following disfluencies in the left superior temporal gyrus and the left inferior frontal cortex. Therefore, the presence of a disfluency affects the response to upcoming input, suggesting that it prepares the listener for higher complexity in the upcoming speech, by potentially allocating increased attention resources that facilitate integration in context.

Brain responses to sensory stimuli depend on the specific task performed by the observer. Previously, fMRI measurements in ventral temporal cortex (VTC) showed that BOLD responses to visual input scale with task demands. This scaling of BOLD responses is thought to be driven by the engagement of cognitive functions required for the task. To better understand how these cognitive task demands influence neural activity in VTC, we measured electrocorticography (ECoG) responses to images of varying contrast in two human participants during either a fixation or image categorization task. ECoG high frequency broadband activity (>70 Hz) showed that local neuronal responses increased with image contrast as expected for sensory encoding, and were scaled by task demands approximately 200 ms after stimulus onset. In contrast, ECoG low frequency activity in the alpha/beta range (8-28 Hz) was insensitive to image contrast and showed larger decreases with increasing task demands. These results indicate that high frequencies in VTC reflect both the encoding of sensory inputs and task demands, similar to prior BOLD response measurements, whereas low frequency oscillations represent task demands, not sensory inputs per se. In line with the interpretation that low frequency oscillations reflect pulsed inhibition, we speculate that a power decrease in low frequency oscillations amplifies neural activity in VTC in support of visual task demands.

How does the brain turn a physical signal like speech into meaning? It draws on two key sources: linguistic structure (e.g., phonemes, syntax) and statistical regularities from experience. Yet how these jointly shape neural representations of language remains unclear. We used MEG to track phonemic and acoustic encoding during spoken language comprehension in native Dutch, Mandarin-Chinese, and Turkish speakers. Phoneme-level encoding is stronger during sentence comprehension than in word lists, and more robust within words than random syllables. Surprisingly, similar encoding emerges even in an uncomprehended language but only with prior exposure. In contrast, acoustic edges are briefly suppressed early in comprehension. This suggests that the brain’s alignment to speech (in phase and power) is robustly tuned by structure and by learned statistical patterns. Our findings show how structured knowledge and experience-based learning interact to shape neural responses to language, offering insight into how the brain processes complex, meaningful signals.

This workshop proposes looking at errors in humans and robots. The workshop will focus along three axes: Human, Robot and Interaction induced errors. Errors in Human-Robot Interaction (HRI) pose a crucial challenge for the deployment of robots. Errors can affect the safety and trust of users, therefore it is important for empowering society that these robots can handle errors robustly. The workshop intends to foster interdisciplinary discussion among researchers in robotics, HCI, cognitive science, and social sciences, to encourage the development of robust, user-centered, and socially responsible HRI systems. The invited speakers, paper presentations and discussions will focus around topics of detecting errors, handling errors, adapting to errors, restoring trust, using intentional errors, communication strategies, issues of expectation as well as look at design, interdisciplinary and ethical approaches to research. In this way we will help inform research into errors, and develop robotic systems capable of robust interaction and collaboration.

This study investigates whether and, if so, how cognates facilitate lexical-semantic processing during early bilingual development. Additionally, we examine the interaction between the cognate facilitation effect (CFE) and bilingual experience factors, such as language proficiency, exposure, and age. We investigated language backgrounds and recorded event-related potentials during a semantic priming task in Dutch-German bilingual children. Most participants were Dutch-dominant, characterized by higher exposure and proficiency in Dutch. We compared the N400 response to target words preceded by semantically related cognate versus non-cognate primes. We found a reduced N400 effect (indexing cognate facilitation) only in the non-dominant language (nDL; German). Individual difference analyses further revealed that higher proficiency of nDL and increasing age attenuated the CFE. In contrast, higher cumulative exposure was associated with an amplified CFE. These findings suggest that cross-linguistic activation in lexical-semantic processing may benefit younger children with either lower proficiency or higher exposure to their non-dominant language during language processing. Together, the study offers direct neural evidence for bilingual cognate facilitation effects and highlights the importance of investigating interactions with external factors in early bilingualism. Future longitudinal research should examine whether cognate reliance serves as a temporary scaffolding mechanism for the acquisition of the non-dominant language.

Abstract Normative Modelling (‘brain growth charting’) is now a well-established method for computational psychiatry and involves charting centiles of variation across a population in terms of mappings between biology and behavior, providing statistical inferences at the level of the individual. These models have helped the field to move away from case-control analysis toward individual-level analysis. Correspondingly, normative modelling has now been applied to chart brain development and ageing in many populations and has been used to quantify individual deviations across various neurological and psychiatric conditions. This has been supported by large-scale models that are openly accessible for diverse brain imaging modalities. As normative modelling continues to grow, several recent methodological developments, such as non-Gaussian models, longitudinal models, and federated learning, have been implemented in different software tools, including the Predictive Clinical Neuroscience toolkit (PCNtoolkit). In this protocol update, we provide: (i) a revised overview of this methodological landscape; (ii) an update to our 2022 standardised analytical protocol for normative modelling of neuroimaging data, including options for federated and longitudinal normative models; (iii) practical guidance suited to both novice and experienced practitioners supported by open-source code examples implemented in the refactored version of PCNtoolkit; and (iv) updated models for cortical thickness, volumetric data, diffusion-weighted imaging and longitudinal data for use by the community.

When speakers refer to objects in the world, they frequently overinform. Contrary to classical theories in linguistics, we hypothesise that overinformativeness is an efficient means of facilitating listener comprehension: speakers use redundancy to provide their listeners with search-efficient perceptual information. In Experiment One (N = 72), we operationalise search efficiency as the ease or difficulty of perceptual discrimination. We borrow methods from psychophysics to manipulate discriminability across attributes (material, colour) and across sensory modalities (audition, vision). We show that across both attributes and modalities, speakers overinform more often when the redundant information helps the listener perceptually discriminate the referent, thus aiding listener-search. Contrary to our expectations, we also find that speakers disproportionately overinform with colour information (relative to material). In Experiment Two (N = 97), we investigate the disproportionate use of colour directly, addressing explanations that appeal to colour’s perceptual salience by 1) dampening colour’s perceptual distinctiveness, and 2) controlling linguistic complexity both in terms of language production and retrieval effort. Contrary to widespread explanations, these factors cannot explain the disproportionate use of colour: colour is privileged in reference.

Common ground is the information that the participants in a conversation treat as background information for the purposes of their interaction. We review two traditions of research on common ground. The formal tradition, consisting mainly of theoretical linguists and philosophers of language, has developed increasingly sophisticated formal models of common ground to generate predictions about an expanding range of empirical phenomena. Meanwhile, the psycholinguistic tradition has focused on a narrower range of phenomena while developing more realistic theories of the psychological mechanisms that allow us to select and represent common ground. After summarizing these two traditions, we consider several reasons why they should be reintegrated, and we argue that the best way to bring them back together would be to adopt a cognitive-pluralist approach, whereby language users have access to a variety of mechanisms for managing background information, which are more or less available and efficient depending on the communicative situation and the kind of information mentally represented as well as the cognitive demands of each mechanism.

Abstract The emergence of human language is a complex process, and to investigate the role of iconicity in this, researchers have combined insights from computational models with empirical observations from laboratory experiments. This chapter provides an overview of the most important insights on the interaction between iconicity and other linguistic properties such as combinatoriality and systematicity. In the experimental and computational work reported, it is shown how iconicity can affect the way in which emerging languages are learned and used. The chapter also discusses how computational methods can help to better understand the gradient and subjective nature of iconicity.

<title>Abstract</title> Neural tracking, how brain activity synchronizes with speech, remains contentious regarding whether it's driven by lexical statistical information or linguistic structures. Previous research provided evidence for both accounts, but methodological limitations made direct comparisons difficult. We used isochronous Turkish stimuli with head-final phrase structure to eliminate confounds of word category repetition at the phrase level. By controlling acoustic features, word repetition, meaning-relatedness, transitional probabilities, and syntactic structure, we isolated distinct contributions of lexical statistics and syntactic structure to neural tracking. Results showed both factors independently shape neural responses, with complex linguistic structures engaging broader brain networks. Computational modeling revealed that hierarchically-coupled oscillations explained neural data better than statistical models alone, supporting structured linguistic representation building alongside word-level statistics. Our model proposes meaning emerges through flexible coupling among neural populations encoding lexical categories. These findings demonstrate that hierarchical processing crucially contributes to neural speech tracking beyond statistical repetitions.