Poster Sessions

The brain undergoes a wide range of changes throughout the human lifespan. By the age of two, most axonal connections in the brain are established; however, maturation continues through early childhood and adolescence. Studies focusing on the cortex revealed a developmental pattern that follows a sensory-association axis, with unimodal areas developing first and association areas showing a prolonged maturation until early adulthood. This developmental trajectory parallels the emergence of higher cognitive abilities during childhood and adolescence. The cortex, however, is not an isolated system but closely interacts with the thalamic subnuclei from early development onwards. While studies have shown thalamocortical connectivity changes in the perinatal stage, it is yet unknown how the thalamus and its connectivity with the cortex matures in childhood and adolescence. However, this thalamocortical connectivity plays a critical role in cognition but also neuropsychological disorders. Here, we aim to study developmental changes of the thalamus and how thalamocortical connectivity profiles evolve from childhood to young adulthood (age range 5 to 21). Extending previous studies that treated the thalamus as a single unit, we here differentiate the thalamus at the subnuclei level. Using diffusion MRI, we first access the local microstructural changes of thalamic nuclei estimated by age-related fractional anisotropy. Next, we employ probabilistic tractography to study changes in the structural connectivity between thalamic nuclei and cortical subregions. Finally, we explore the development of structure-function coupling of thalamocortical connectivity. We hypothesize that unimodal thalamic areas saturate earlier, whereas higher-order nuclei and their connectivity to association areas exhibit a prolonged development, mirroring cortical trajectories.

An increase in adolescent obesity is directly implicated in the growing prevalence of health problems, such as diabetes and cardiovascular disease. One promising avenue of intervention and treatment is the strengthening of executive function (EF), as prior research indicates it may be relevant to breaking unhealthy habits. Previous studies have indicated that executive functioning is compromised in obesity. Progress in this area is currently limited, however, by a poor understanding of the precise nature of these impairments. Studies have been inconclusive as to whether EF impairment is present in adolescents living with obesity, as well as which EF subdomains show the most pronounced impairments. Here, I will present a systematic review and analysis of the literature to determine whether EF is generally impaired in adolescents with obesity, as well as exploring which subdomains show the most pronounced impairments. Comprehensively mapping out EF deficits in adolescents with obesity can open up new avenues for more targeted obesity treatments and early interventions.

Proper name anomia is a common language deficit observed in patients with dementia, impacting their ability to recall and retrieve names of familiar people. This study investigates the neural changes associated with a 6-week proper name anomia therapy program in patients diagnosed with dementia using magnetoencephalography (MEG). Methods: Thirteen patients diagnosed with mild to moderate dementia and exhibiting proper name anomia were enrolled in this prospective intervention study. The participants underwent comprehensive neuropsychological assessments, including language and memory evaluations, to confirm the presence of proper name anomia and establish a baseline level of cognitive function. Following the baseline assessment, patients underwent a structured 6-week proper name anomia therapy program, consisting of personalized name training, audio cueing, and spaced retrieval techniques. During the intervention period, participants received weekly therapy sessions focused on recalling names of familiar people. MEG recordings were obtained at two time points: pre-therapy (baseline), post-therapy (immediately after the 6-week intervention). MEG recordings were obtained during proper name retrieval tasks, with patients recalling both trained and untrained proper names. Results: Results revealed gamma activity changes in response to therapy for both trained and untrained proper names. Patients demonstrated increased/decreased gamma activity in the left temporal pole during retrieval of both trained and untrained items following therapy, suggesting enhanced neural synchronization and network connectivity associated with name retrieval. Interestingly, the gamma activity changes were more pronounced for trained proper names compared to untrained items. Trained proper names exhibited a more substantial increase in gamma activity during retrieval tasks, indicating that therapy-induced neuroplastic changes were particularly effective in enhances.

The dorsolateral prefrontal cortex (DLPFC) is a prominent stimulation target for the treatment of major depressive disorder (MDD) with repetitive transcranial magnetic stimulation (rTMS). The therapeutic effect of DLPFC stimulation likely stems from the DLPFC–subgenual anterior cingulate cortex (sgACC) connectivity. A recent study suggested that there is an overlap of the hubs in the depression network, including the DLPFC, sgACC, and vagus nerve (VN), with the heart-brain axis, and that stimulation of these hubs consequently leads to heart rate (HR) decelerations. A specific 10 Hz dash rTMS protocol (‘neuro-cardiac guided rTMS version 2.0’, NGC-2.0 rTMS) has been proposed to induce coupling of TMS and HR changes (heart-brain coupling; HBC). Hypotheses HBC effects can be observed as reported in [4] and are stable across three sessions. No effects are shown in the sham session. Experimental Design This study aims to replicate and corroborate previous findings by conducting NGC-2.0 rTMS sessions targeting the DLPFC, supplemented by sham TMS, across three sessions with a larger cohort of twenty healthy young participants. Our research employs a within-subject factorial design, considering three factors: Target (Beam F3, anterior F3, posterior F3, inferior F3, superior F3, 5CM method (left DLPFC), and Sham), Intensity (16 levels), and Session (first, second, third). This approach will yield 336 (7 targets * 16 intensities * 3 sessions) unique measurements per participant, involving 20 participants over three sessions. Our stimulation protocol is primarily based on the NGC-2.0 rTMS protocol as proposed by Dijkstra et al., with modifications that include adding resting periods before and after stimulation sessions for baseline measurements. Analysis Plan Electrocardiogram (ECG) data will be processed as described in [4]. We will then use a 3-way Bayesian analysis of variance to assess the main effects and interaction effects of target, intensity, and session on HBC, quantified as the mean power (mV2) at 0.0625 Hz (stimulation frequency). This design will yield 336 (7 Targets * 16 Intensities * 3 Sessions) measurements per participant. For the main effects, we will assess the individual impact of target, intensity, and session on HBC. For the interaction effects, we will explore how combinations of the three factors influence the outcome, e.g., how different targets will interact with intensity levels across different sessions.

The Somatic Marker Hypothesis, introduced by Antonio Damasio in 1994, emphasizes the essential role of perceiving 'body states' in decision-making processes. This theory links interoceptive processing to specific brain areas, such as the insula and somatosensory cortex, suggesting their primary function is emotional evaluation of action outcomes. Later, motor initiation was shown to be influenced by the phases of the cardiac cycle [Palser et al., 2021; Al et al., 2023], and even motor agency was suggested to depend on cardiac input [Herman & Tsakiris, 2020]. Here, we further explore this perspective on cardiac input for motor preparation, advocating for its possible incentive role in movement execution. We implemented the classical Libet’s paradigm with W- and M-experimental condition and 40 self-paced movements per each of the two conditions in naïve participants (n = 27, mean age 24) [Bredikhin et al., 2023]. Heart-evoked potential (HEP) was obtained by averaging all trials excluding the epochs which fell into the interval of < 500 millisecond prior to the button press. Behavioral data were aligned with the cardiac cycle. We observed increase in HEP amplitude in the W-condition concomitantly with the non-uniform distribution of button presses predominantly occurring during the diastolic phase of the cardiac cycle, which provides a link between cardiac interoception and motor control. These findings suggest that cardiac interoception relates to motor preparation under conditions of uncertainty, offering a novel interpretation of the W-condition in Libet’s task. Moreover, our results reinforce the association between cardiac interoception and the experience of volition in tasks that involve self-paced movements. Overall, our results challenge traditional interpretations of the W-condition and presents an alternative perspective on the 'urge to move' phenomenon.

Resting-state heartbeat-evoked potentials as a feature of the executive function The heartbeat-evoked potentials (HEP) measured as neural responses to heartbeats, have been studied a marker for cardiac interoception processing, emotions, perceptual awareness, and self-consciousness[1-3]. Interoceptive processing is considered an important source of sensory input for cognition. Among the cognitive functioning, executive functions have also been shown to predict memory decline and further global cognitive decline[4,5]. Nevertheless, no study has reliably shown so far that resting-state HEP (rsHEP) could be considered a neural trait by itself that would meaningfully link to cognition. In this study, we will test whether HEP may index inter-individual differences in executive function. We associate resting-state heartbeat evoked potential (rsHEP) with trail making test time and accuracy using the data of adult participants in the LIFE dataset (N = 1337). We will use general linear models (GLM) with cognitive task scores as the response variable and rsHEP as the variable of interest controlling for confounding variables such as age, gender, blood pressure, BMI, and heart-rate variability. If rsHEP can be used to capture the inter-individual differences in executive function, it helps to understand the heart-brain interaction in the domain of cognition.

Deficits in rhythm perception and production have been reported in a variety of psychiatric, neurodevelopmental and neurologic disorders. Since correlations between rhythmic abilities and cognitive functions have been demonstrated in neurotypical individuals, we here investigate whether and how rhythmic abilities are associated with cognitive functions in 35 participants with neurocognitive deficits due to acquired brain lesions. We systematically assessed a diverse set of rhythm perception and production abilities including time and beat perception and finger-tapping tasks. Neuropsychological tests were applied to assess separable cognitive functions. Using multiple regression analyses we show that lower variability in aligning movements to a pacing sequence was predicted by better inhibitory control and better working memory performance. Working memory performance also predicted lower variability of rhythmic movements in the absence of an external pacing sequence and better anticipatory timing to sequences with gradual tempo changes. Importantly, these predictors remained significant for all regression models when controlling for other cognitive variables (i.e., cognitive flexibility, information processing speed, and verbal learning ability) and potential confounders (i.e., age, symptom strength of depression, manual dexterity, duration of illness, severity of cognitive impairment, and musical experience). Thus, all rhythm production abilities were significantly predicted by measures of executive functions. In contrast, rhythm perception abilities (time perception / beat perception) were not predicted by executive functions in this study. Our results, enhancing the understanding of cognitive underpinnings of rhythmic abilities in individuals with neurocognitive deficits, may be a first mandatory step to further potential therapeutic implications of rhythm-based interventions in neuropsychological rehabilitation.

Yoga nidra is a form of meditation intended to regulate arousal. It involves awareness of the body, the breath, predominating emotions and their opposites, and thoughts. In small-scale case studies, yoga nidra has been found to reduce symptoms of PTSD and depression in veterans (Pence et al., 2014; Stankovic, 2011). Disordered sleep is a risk factor for development of PTSD and a critical component of PTSD symptomology (Gehrman et al., 2013; Mellman et al., 1995; Nguyen et al., 2023 ; Spoormaker & Montgomery, 2008). This study aims to use an analogue trauma (viewing of a ‘stressful film’) to explore the impact of yoga nidra as an early intervention on development of intrusive memories, explore the impact of self-reported sleep the week prior to film on intrusive memories, and explore potential mechanisms for yoga nidra: regulation of physiological arousal, mindfulness, and avoidance. To explore these aims, the following study procedure will be used: Participants will complete one week of sleep diaries prior to an in-person lab session. The in-person session will involve completion of baseline trait questionnaires, watching the ‘trauma film’, an explanation of the relaxation technique and 30 minutes of either yoga nidra or music listening, intrusive memory diaries, and a word-viewing eye-tracking task. During the session, heart rate, electrodermal activity, and blood pressure will be measured. After the in-person lab session, participants will complete 1 week of sleep diaries and intrusive memory diaries, along with additional questionnaires 1 week after the lab session. We aim to use the data to investigate both the effectiveness of yoga nidra meditation for the mitigation of intrusive memory formation and maintenance, as well as its potential mechanisms. In addition, we aim to examine the relationship between sleep and intrusive memories, to predict intrusive memory formation before analogue trauma and to investigate their relationship after analogue trauma.

Multiple Myeloma (MM) is an incurable haematological malignancy gaining increasing importance due to recent advances in treatment and improved survival rates in an ageing population. Yet to date, there has been no systematic reviews examining the impact on health-related quality of life from a cognitive lens across treatment regimes. The current study aimed to synthesise existing data on cognition in patients with MM. The research question posed was whether patients with MM experience cognitive deficits across treatment. A systematic database and grey literature search were conducted using four databases including PsycINFO, MEDLine, Embase, and Google Scholar yielding 119 abstracts which were then screened. 14 studies were included in the final review of which 5 studies (N= 988) have been retained for meta-analysis. The systematic research review revealed general cognitive deficiencies in MM patients. Furthermore, the need for more randomized controlled trials to establish the objective and specific components of cognition that are affected by MM, and its treatments, was identified.

The brain constantly receives interoceptive feedback about the internal state of the body. One form of feedback is the cardiac axis, where the neurophysiological indicator of the cardiac signal is the heartbeat-evoked potential (HEP). Various evidence suggests that the HEP represents a neuronal prediction for each heartbeat. If so, that raises important questions about its role in common forms of arrhythmia, such as the primarily benign, premature beats that occur outside the physiological heart rhythm known as extrasystoles. This type of arrhythmia can be subclassified into supraventricular extrasystoles (SVES) and ventricular extrasystoles (VES). We analysed the neural response to both types of extrasystoles as an interruption of the physiological heart rhythm. Using a multiverse approach, we compared the HEP associated with extrasystoles with the HEP associated with normal heartbeats in time and source space. We found that the HEP was significantly reduced for VES due to sources in the left and right insular areas. For the postextrasystolic beat of VES and SVES, the HEP amplitude was significantly larger in the time window of 130 ms to 200 ms after the R-peak, originating from sources in the left frontal orbital cortex and the anterior cingulate gyri. The results indicate that the reduced HEP response to the VES may be attributed to the absence of the baroreceptor reflex during the VES. The HEP of the postextrasystolic beat may indicate an interoceptive surprise response to the discrepancy between the predicted and actual heart rhythm or an early firing of the baroreceptor. These findings emphasise the significant role of the baroreceptor in the HEP pathway.

Epilepsy is increasingly regarded as a network disorder. Cortical thinning and white matter abnormalities, are common structural alterations in epilepsy patients and indicate disease progression. At the same time, sex differences in epilepsy are also widely recognized such as seizure susceptibility and treatment response. These differences could be associated with steroid hormones, neurotransmitter systems, and variations in regional morphology and neural circuit. However, to what extent these morphological abnormalities are related to sex differences remains unknown. Here, we study the relation between sex differences and MRI brain structural markers, using the multisite ENIGMA-Epilepsy dataset, including 896 adults with temporal lobe epilepsy (TLE), 113 adults with idiopathic generalized epilepsy (IGE), and 886 healthy controls from 19 international sites. Around this objective, the following analyses will be conducted: (1) Comparison of different cortical biomarkers distribution between sexes in TLE and IGE, including cortical thickness, subcortical volumes, fractional anisotropy (FA) and mean diffusivity (MD). (2) Comparison of clinical characteristics in TLE and IGE between sexes, such as disease duration, age of onset and therapeutic responses. The correlation between clinical variables and cortical features will be explored separately for each sex in TLE and IGE. (3) To investigate whether sex differences are related to epilepsy susceptibility in TLE and IGE from a network perspective, we will generate atrophy profiles by comparing cortical thickness and subcortical volumes between epilepsy patients and healthy controls in each brain region. Additionally, spatial correlation analysis will be performed between normative connectivity profiles and atrophy profiles, to identify the cortical and subcortical hub regions and disease epicenters. (4) To elucidate the contribution of sex difference in the cortical features and network factors, a linear model will be constructed to examine the effects of sex and morphological features in TLE and IGE. Furthermore, we will compare these effects patterns with degree centrality and disease epicenters maps we got from previous steps, by exploring their correlation patterns. We anticipate differences in biomarker distributions and network disruption between sexes in TLE and IGE. These differences might relate to disease progress and can be regarded as biomarkers for individualized therapy.

The locus coeruleus (LC), a key part of the brain's neuromodulatory system, is responsible for arousal by distributing noradrenaline throughout the brain. In doing so, the LC also regulates the autonomic nervous system, thereby influencing cardiovascular functions such as heart rate and blood pressure. In older adults, a reduction in neuromelanin-based contrast in structural MRI of the LC correlates with a higher prevalence of Alzheimer's disease and cognitive decline. Conversely, in younger adults, increased LC contrast is associated with lower heart rate variability (HRV), indicative of reduced physical and cognitive functioning. Interestingly, both Alzheimer's disease and impaired cognitive functioning are linked to poorer cardiovascular health. Therefore, to investigate the relationship between cardiovascular health and LC contrast, we will use HRV and baroreflex sensitivity as indicators of cardiovascular health and examine their associations with LC contrast in both younger and older adults. We hypothesize that lower LC contrast in younger adults will correlate with better cardiovascular health, while in older adults, higher LC contrast will be associated with better cardiovascular health. This study aims to enhance our understanding of the relationship between LC contrast and cardiovascular health, and their contributions to neurodegenerative diseases and cognitive functioning. Key words: Locus Coeruleus, Cardio vascular health, MRI, Heart rate variability

The expression of synaptic plasticity markers varies systematically across cortical areas in primates, being lower in limbic areas, of poor laminar elaboration, and higher in eulaminate areas, with six-well developed layers. In the temporal cortex, perirhinal and parahippocampal limbic areas are more vulnerable to epilepsy and Alzheimer’s disease (AD). This suggests that synaptic plasticity markers and selective vulnerability to neurological disorders are linked. Particularly, selective vulnerability of temporal mesocortical areas has been observed in AD, suggesting that simpler laminar architecture may be associated with the expression of factors that render mesocortical neurons more vulnerable than in eulaminate areas. Here, we present a simple manual method to quantify the intracortical content of myelin (a well-known inhibitor of synaptic plasticity) along the cortex of the temporal lobe. To this end, we used T1 MRI coronal slices of one individual from the open access data of Montreal Institute, which were loaded in Image J. Intracortical myelin content was quantified by drawing a ROI for each cortical type at 6 coronal levels of the temporal lobe, obtaining the optical density of myelin. Our data shows increased myelin density in eulaminate areas compared to limbic areas. The content of myelin also increased across eulaminate areas of progressively better laminar elaboration. These findings suggest that limbic areas of the human temporal cortex are more plastic than eulaminate temporal areas. Therefore, higher synaptic plasticity of limbic temporal areas may be related to the selective vulnerability to neurological disorders. Prospects for this project include the validation of these preliminary results within a more extensive sample from the Fundación CIEN Brain Bank. Our simple method will be compared with state-of-the-art automated methods, such as the Lesion Segmentation Toolbox (LST)-AI deep learning ensemble.

Psychosis progresses along a continuum. While clinical heterogeneity is evident across the continuum, it remains unknown whether this is also reflected in white matter (WM) heterogeneity and whether parsing WM heterogeneity may reveal subgroups with more pronounced clinical features. This analysis included 212 participants consisting of healthy controls (HC, n=59), individuals with high schizotypy (SPT, n=27), at-risk mental state (ARMS, n=35), and patients with first episode psychosis (FEP, n=50) and schizophrenia (SZ, n=41). Fractional anisotropy (FA), mean diffusivity (MD) and fiber density (FD) were derived from diffusion tensor imaging (DTI) and the Person-Based-Similarity Index (PBSI) and Coefficient of Variation Ratio (CVR) were computed to assess global and local heterogeneity. ANCOVAs were performed to determine whether people with deviating PBSIs exhibit more pronounced clinical features. Global heterogeneity for all DTI parameters significantly differed across groups, with greatest heterogeneity in SZ and SPT. Results further indicate that FA deviators exhibit lower global functioning and higher negative symptoms. Local FA heterogeneity was greater in FEP relative to ARMS and HC in almost all WM tracts, while SZ patients specifically showed greater heterogeneity in the right thalamic radiation and the left uncinate compared to HCs. Group differences in WM heterogeneity might be indicative of symptom specificity and duration. Future large-scale studies are warranted to test the robustness of DTI subtypes and their clinical relevance.

Aging is associated with declines in cognition and brain structural integrity. However, there is equivocality over 1) the specificity of affected domains in different people, 2) the location of associated patterns of brain structural deterioration, and 3) the sociodemographic factors contributing to ‘unhealthy’ cognition. We aimed to identify cognitive profiles displayed by older adults and determine brain and sociodemographic features potentially shaping these profiles. A sample of Southeast-Asian older adults (N = 386) participated in a multi-session study comprising cognitive testing, neuroimaging, and a structured interview. We used computational models to extract latent mechanisms underlying cognitive flexibility and response inhibition. Data-driven methods were used to construct cognitive profiles based on standard performance measures and model parameters. We also investigated grey matter volume and machine-learning derived ‘brain-ages’. A profile associated with poor set-shifting and rigid focusing was associated with widespread grey matter reduction in cognitive control regions. A slow responding profile was associated with advanced brain-age. Both profiles were correlated with poor socioeconomic standing and cognitive reserve. This study furthers understanding of how distinct aging profiles of cognitive impairment uniquely correspond to specific vs. global brain deterioration and the significance of socioeconomic factors in informing cognitive performance in older age.

Successful social interactions and building enduring relationships depend on the ability to adapt to social cues and learn from them. During adolescence, individuals undergo significant changes in their social cognitive abilities, including how they respond to social rewards and feedback. In our research, we examined how these abilities vary with age by studying 152 adolescents aged 10 to 24. Participants engaged in a trust game where they interacted with groups categorized as mostly trustworthy or untrustworthy. We found that across all ages, participants improved in trusting the trustworthy group over time. However, older adolescents showed more improvement in withholding trust from the untrustworthy group compared to younger adolescents. The most pronounced differences in learning between trustworthy and untrustworthy groups occurred in mid-adolescence. Using computational models, we analyzed how participants learned from positive and negative experiences and their preferences in social interactions. Our findings revealed age-related differences in how individuals respond to positive and negative social experiences. Older adolescents were more efficient in devaluing trust toward individuals who did not reciprocate trust, even when the outcomes did not directly affect their gains. Additionally, we will explore how these behavioral patterns correlate with neural responses, aiming to understand how reward-based learning and social cognition networks interact across adolescent development. This integrated approach provides insights into the dynamics of trust and social learning and illuminates the neural mechanisms that shape adaptive social behavior during adolescence.

The human brain’s potential to represent mental states—such as needs, emotions, wonders, thoughts, goals and plans—has long held philosophers, psychologists, and neuroscientists spellbound. This ability, commonly known as mentalizing, anchors behaviours to the mind and enables human beings to imagine and infer the intentional stance behind observable actions. Mentalizing lays the foundations for seeing oneself from the outside and others from the inside and is thus of essence in everyday social networks. Concurrent with the development of social bonds and attachment relationships, mentalizing begins in the first year of life. Long before the emergence of words, infants search for proximity and are curious about and rely on non-verbal modalities including body postures, gestures, movement dimensions and qualities, kinaesthetic and facial expressions, touch, gaze, and paralinguistics in order to learn about agency of others and workings of the external world. Hence, mentalizing is first and foremost an embodied phenomenon, inextricably embedded within and shaped by the interactive, body-to-body dialogues and the surrounding space. Notwithstanding the embodied and spatial roots of mentalizing, the mechanisms explaining the translation of spatial cognition to social cognition remain poorly understood. Converging evidence from a growing body of research indicates that the navigation of the sphere of mind could be a legacy of evolutionary tinkering with neural algorithms, transmitters, and circuits which have primarily emerged in response to seeking and exploring the physical world. Physical navigation integrates position, orientation, and locomotion with neural computations related to distances and durations, contributing fundamentally to the direction of goal-directed behaviours. It has been argued, however, that the brain regions and neural, geometric representations involved in the organization of cognitive maps of the spatial layout could have expanded to encode the real-world social relationships. Implicit in recent findings is the idea that the neural systems of spatial cognition may have been repurposed over time to pave the path for the evolution of a social brain. In line with this possibility, the aim of the current presentation is to review the nexus between embodiment and spatial-social cognition and discuss the potential of an integrating framework in exploring the neural architecture of mentalization.

Social interaction takes up a large portion of our social life. Social neuroscience research has extensively studied human social interaction but predominantly uses third-person paradigms where participants passively observe ongoing events – like watching TV. However, real-life social interaction is anything but passive observation. There is an urgent need to switch from this third-person perspective to a second-person perspective, where participants can actively interact with other individuals as they would in everyday life. In this three-year PhD project, I propose three work packages (WPs) to explore the neural correlates underlying real-time social interaction, focusing on the role of empathy, mentalizing, and mental health disorders. In WP1, I will systematically review the literature on real-time social interaction in patients with mental disorders and conduct a meta-analysis of the underlying neural correlates. Then, I will develop a novel “interactive EmpaToM (iEmpaToM)” paradigm for functional magnetic resonance imaging to collect brain activation data during real-time social interactions. Using iEmpaToM, WP2 will study the neural mechanisms of social interactions and how empathy and mentalizing demand influence the process. WP3 will explore the relationship between autistic traits and neural activity during social interactions. Together, this PhD project will help gain novel insights into real-time social interaction and individual differences related to mental health.

Social neuroscience has set out to uncover the neural mechanisms and dynamics of everyday social interactions. Hyperscanning, a promising methodology in which the online neural activity of two or more persons is simultaneously recorded, consistently demonstrates interpersonal neural synchrony as a neural marker of social communication, coordination, and cooperation. Hyperscanning analyses remain standardised despite recent efforts. Namely, the connectivity measures used for analysing the statistical covariance of neural signals are almost-always model-dependent and insensitive to more complex, nonlinear dynamics, despite both social and neural dynamical systems demonstrating both stochasticity and nonlinearity. Information theory, the mathematical theory of communication, and its probability-based measures are particularly well-suited to address this shortfall given their relaxation of statistical and modelling assumptions and their more recent advancements in addressing higher-order informational dynamics within complex systems. After introducing and motivating this mathematical framework, I demonstrate the use of various information-theoretic measures, evaluate the performance of various estimators, and benchmark the measures against standard connectivity metrics. In decreasing levels of abstraction, these evaluations are undertaken on proof-of-concept small-world networks, large biophysical dual-brain simulations of interpersonal neural connectivity, and finally on an empirical EEG-hyperscanning dataset. This thesis presents a necessary and important step forward for the both the standardisation of hyperscanning analyses as well as continuing efforts towards uncovering the complex dynamics of social systems.

There is accumulating evidence that the human cerebellum is heavily implicated in adult social cognition. Yet, its involvement in the development of Theory of Mind (ToM), a hallmark of social cognition, remains elusive. In a functional MRI study involving children with emerging ToM abilities (N=41, age range: 3-12 years) and adults (N=78), we showed that children with ToM abilities activated cerebellar Crus I-II in response to ToM events during a movie-watching task, similar to adults. This activation was absent in children lacking ToM abilities. Functional connectivity profiles between cerebellar and cerebral ToM regions differed as a function of children's ToM abilities. Notably, task-driven connectivity shifted from upstream to downstream connections between cerebellar and cerebral ToM regions from childhood to adulthood. Greater dependence on connections emerging from the cerebellum early in life suggests an important role of the cerebellum in establishing the cognitive processes underlying ToM in childhood and thus for the undisrupted development of social cognition.

The context of social interaction is a rich ground for multimodal behaviours and brain activity that dynamically coordinate between those engaged to support successful communication. This interpersonal exchange represents the ecological niche where learning concepts typically occurs over the lifespan, starting with children-caregiver interactions. This context may be particularly important for abstract concepts which, given their immaterial nature, may especially rely on interactive dynamics of face-to-face communication, both in terms of behavioural coordination and neural synchronisation. This study aims to characterise the behavioural and neural coordinative dynamics that underlie primary-school-aged children’s abstract concept learning in interaction with caregivers. We present an ongoing multimodal fNIRS hyperscanning study, where caregivers and their 8 to 9-year-old children take part in a novel, interactive concept learning task. The primary aim is to build a dataset of brain and behavioural measures to identify significant predictors of successful learning. Additionally, we present an unprecedented pipeline for verbal analysis integrating AI-based transcription and automatic annotation of coordinative verbal behaviours, namely question-answer, turn-taking and backchanneling. With these data, we identify the specific behaviours contributing to abstract concept learning. To measure brain activity, we use the Hitachi ETG4000 NIRS device with 22 channels for each member of the dyad. We focus on left dlPFC and TPJ, key regions involved in social cognition, language, and learning. Using Wavelet Transform Coherence analysis, our study will uncover whether and when brain-to-brain synchronisation between caregivers and children predicts successful learning in the above-mentioned regions of interest. Behaviourally, we hypothesise that key coordinative behaviours (i.e., question-answers, backchanneling, turn-taking), which have been found to be relevant in face-to-face communication, predict abstract concept learning. Neurally, we hypothesise that inter-brain synchrony will contribute to successful learning, although the pattern may be modulated by the behaviours of the dyad. As we take an embodied approach to social neuroscience by simultaneously considering the dyads’ behavioural and brain-to-brain coordination as predictors of learning, this study will enhance our understanding of the interplay between brain and behaviour within social learning contexts.

Being mimicked (BeMim), the state of having your actions copied by another person, is believed to foster liking and affiliation. To uncover the neural and cognitive mechanisms underlying this effect, we employed functional Near-Infrared Spectroscopy (fNIRS) to monitor brain activity in two participant groups subjected to different types of mimicry. In the Choice BeMim condition, participants pointed to a painting and observed a confederate expressing a preference for the same or a different painting. In the Motor BeMim condition, participants pointed to a painting and then saw a confederate make the same or different arm movement to another painting. Brain activity in the temporal and parietal cortex was recorded using fNIRS. Behavioral results indicated a preference for confederates who mimicked participants' choices, which was reflected in activation patterns in the right Intraparietal Sulcus (IPS) and left Inferior Parietal Lobule (IPL). Conversely, for Motor BeMim, a subtle behavioral liking effect was associated with clear brain activity in the right Angular Gyrus, supporting previous findings linking this region to motor coordination and the sense of self (Farrer et al., 2008). This provides the first neuroimaging study of the mechanisms by which mimicry causes people to like each other in the context of real face-to-face social interactions.

Encoding new memories takes place against the backdrop of a rich library of information acquired through one’s life. Several studies show that prior knowledge, such as schemas, strengthens encoding and accelerates recall of new memories that are in agreement with it (congruent), while others show the opposite pattern where prediction violation (related to information being incongruent with a schema) facilitates learning. To reconcile the contradictory findings in these two lines of research, a recent framework, the schema-linked interaction between the medial temporal and medial prefrontal regions (SLIMM model), postulates that memory shows a non-linear, U-shaped function with degrees of congruency to prior information. In other words, highly congruent and highly incongruent information with a schema benefit the process of consolidation during learning. However, the SLIMM model remains under scrutiny since empirical evidence is scarce and not sufficient to support its hypotheses yet. Furthermore, the neural underpinnings of such learning processes remain unknown. While some models suggest a trade-off between the medial prefrontal cortex (mPFC) and the medial temporal lobe (MTL) for congruent and incongruent effects respectively (e.g. SLIMM), other models predict an essential role of MTL structures in encoding information congruent to existing knowledge structures. In this study, we use behavioural methods as well as neuroimaging (fMRI) to understand whether and how the representation of prior knowledge enhance encoding and retrieval of new events. We developed a novel spatial schema paradigm, which compares three conditions with varying degrees of congruency to previous knowledge and test the seemingly contradictory behavioural findings in the literature. Our results demonstrate a mnemonic advantage for congruent events, while incongruent events and those lacking a strong prior schema exhibit a disadvantage, suggesting that reaffirming expectations facilitates learning. In the concurrent fMRI study, we directly compare learning systems in the brain that support learning under certain (congruent) and uncertain (incongruent) conditions and investigate the formation and update of schema representations with newly acquired information. This study could lead to a better understanding towards a refined neuroscientific model of how brain networks interact to successfully integrate new information with previous knowledge schema.

Too often still, people lose their lives or livelihoods in tragic accidents that have their root causes in lapses of attention. With the technology available to us today, the development of assistance systems which can detect attentional lapses and prevent such accidents should be viable. One way to achieve this is to use objective, physiological signals in which sudden changes may indicate an imminent lapse. Our study aims at identifying physiological markers that are maximally diagnostic and reliable, first in the lab and then in the 'real-world'. In the lab, we simultaneously record electroencephalographic (EEG) activity and changes in pupil size as we monitor performance in modified versions of two sustained attention paradigms: the Mackworth Clock Task (MCT) and the Sustained Attention to Response Task (SART). In both monotonous tasks, participants are required to stay focused for a long period of time and respond appropriately to infrequent target events. Preliminary results thus far mostly support our preregistered predictions: in the seconds preceding an incorrect response in the SART – when we assume that participants were not paying attention – we observe an increase in activity in the alpha frequency band (8-14Hz) in occipital electrodes, a marker previously associated with attentional focus. Interestingly, this effect is not observed in the MCT data. We also observe changes in pupil diameter, a marker previously associated with neuromodulatory arousal. Further analyses will target time-on-task effects and other measures such as the spectral slope of the EEG power spectrum that indexes excitability fluctuations and how different measures can be combined to increase reliability. Finally, further comparing the data between the two tasks will provide insights into how much any given diagnostic marker generalises across situations with different response demands. This lab-based experiment will critically inform the design of a second, more ecologically valid experiment.

In everyday life, spoken speech streams are often masked by noise or competing speech streams in the surrounding. Thus, listeners reconstruct the continuous speech signal from a comprised sensory input using predictive processing. In an EEG study, we investigate how acoustic and higher-level linguistic factors contribute to individual comprehension performance in challenging competing speech situations. Participants (N = 42) heard two sentences presented simultaneously and were instructed to follow one speaker while ignoring the other one. They subsequently repeated the target sentence. The individual task difficulty was adjusted using an adaptive staircase procedure to account for peripheral differences. We used a generalized logistic mixed effects model to predict word comprehension from linguistic predictability and acoustic information quantified by the amount of glimpsed target signal. Additionally, we quantify the neural tracking of acoustic and linguistic features using forward temporal response functions and predict behavioral performance within and across participants from the cross-validated trial-by-trial model fits. As hypothesized, behavioral results (N = 42) revealed that word surprisal (z = -15.39, p < .001), word audibility (z = 19.17, p < .001) as well as their interaction (z = 8.43, p < .001) contribute to comprehension performance. Neural analysis is ongoing. First results indicate that acoustic (z = 4.70, p < .001) and linguistic word (z = 3.35, p < .001) and phoneme level (z = 3.11, p = .002) neural representation of the target sentence positively predict comprehension performance.

Features of varying complexity are extracted from music along the auditory pathway. Traditionally, sound perception has been described in a sequential model, in which low-level features such as pitch, loudness or sound location precede grouping and structure-building mechanisms (Koelsch, 2011). It has been shown than higher-level structures modify the perception of individual tones, e.g. when the identification of mistuned tones is facilitated in the tonal context (Brattico et al., 2001). In this study, we aim to go further and ask how the grouping of music into phrases affects the perception of a seemingly unrelated feature - the spatial location of sound. The musical material consisted of 12 multi-phrase melodies based on compositions by J.S. Bach. As the music was presented to the participants, the location of the sound changed, either at a phrase boundary or within a phrase. Participants attended two sessions of the experiment, during which neural and behavioural data were collected. The electrophysiological (EEG) data were recorded while participants listened passively to music. In the behavioural part, participants were asked to press a button to indicate the detection of a change in the sound source location. The results of the experiment indicate that the musical structure has an effect on the ability to detect changes in the location of the sound source. EEG data analysis reveals enhanced mismatch negativity (MMN) response at the boundary condition, compared to within-phrase condition (t = 4.1179, p & .001). Although behavioural data did not show increased sensitivity to detect location changes at the boundary, the bias in responses was modulated by phrase structures in music. The results of this study allow us to conclude that the perception of sound’s spatial location can be modulated by arbitrary musical formations - phrases. The processing of abstract-level structures in music alters the performance in lower-level acoustic tasks.

Advances in receive coils, parallel imaging, and pulse sequence design have increased fMRI temporal resolution, with repetition times (TRs) ≤1000 ms now standard for ~whole-brain imaging at 3T. 'Fast fMRI' studies in vision show departures from canonical hemodynamic response functions (HRFs) for short stimuli, with HRF timing differences across and within regions. Some auditory fMRI studies suggest faster peak latencies to sound presentation than predicted by canonical HRF models. HRFs to simple stimuli can show different shapes across brain regions, suggesting that components of the HRF relate to distinct neural processes, like inhibition, a particularly important mechanism in auditory function. However, mechanistic inferences based on HRFs are premature when reliability, regional variation, and inter-individual differences in auditory BOLD responses are not well characterized. Here we characterize the voxelwise HRF to short, naturalistic auditory stimuli to assess reproducibility across sessions within and across participants in auditory relevant areas. We additionally investigated whether more detailed modelling of auditory HRFs would reveal additional information about neural encoding of auditory salience. Five healthy adults participated in two identical fMRI sessions at 3T (6 runs each, TR = 1s, 2x2x2 resolution) in which they passively listened to environmental sounds that varied in perceived salience. HRFs to these stimuli were computed across the entire brain. Estimated responses using deconvolution were classified using data reduction techniques; HRFs were also approximated using several parameterized models. The voxelwise correlation of an individual participant's auditory HRFs was generally high across sessions, with notable HRF shape and timing differences between multiple auditory and frontal areas. Other regions showed robust HRFs, but ones that changed across runs and sessions with increasing exposure to sounds. We also found distinct HRF classes across cortical areas and relate these to quantitative MR maps along with tonotopic and motor somatotopic maps acquired in the same participants.

Curvature has been suggested to play a crucial role in supporting visual object processing and functional selectivity in high-level visual regions. While for artificial stimuli there may be a clear definition of curvature, this is more challenging for natural images, and the definition of curvature can vary from the curvature of the global shape of objects cropped from background to local elements of textures, both of which may deviate from our subjective percept of the curviness of individual natural images. How can we quantify perceived curvature of natural images, and how does this perceived curvature relate to patterns of brain activity? To improve our understanding of perceived curvature, we gathered extensive curvature ratings for 1,854 objects across 27,961 natural images of the THINGS database, compared their alignment with fMRI responses to computed curvature measures (Li & Bonner, 2020; Walther & Shen, 2014), and developed a neural network model that predicted perceived curvature for new images. Perceived curvature exhibited high reliability (r = 0.93). Computed curvature only weakly correlated with perceived curvature (r = 0.26 and r = 0.22) but also weakly correlated with each other (r = 0.14). In the human visual system, perceived curvature generally accounted for more variance across higher-level visual cortex than other measures and corresponded best to known category selectivities (e.g., Li & Bonner, 2020; Long et al., 2018). Given the validity of this curvature measure, we aimed at providing an automated quantification of perceived curvature for novel images. To this end, we finetuned a convolutional neural network to predict the perceived curvature of images. The prediction performance of our model closely approached noise-ceiling (cross-validated R²= 81%), and generalized to an external dataset of more controlled object images (Long et al., 2018; R²= 72%). Together, our results highlight the importance of perceived curvature as a mid-level summary statistic and provide an approach for the automated quantification of perceived curvature in natural object images.

Introduction: Over a century ago, William James posited that our physiological responses shape our affective states. Indeed, afferent sensory signals, like heartbeats, can induce feelings of anxiety and shape our perception. Crucially, during such states, self-touch is commonly observed in humans and primates, raising the question of whether we may actively use self-touch as an anxiolytic that modulates the effects of afferent signals. Soothing effects are typically attributed to a soft touch by others, activating specific C tactile fibers. But despite anecdotal and correlational evidence, its function in self-touch has not been empirically tested. Aim: Leveraging the known effect that states of anxiety facilitate aversive processing, our study investigates whether active self-touch (ST) can modulate the perception of fearful facial expressions in anxiety-inducing contexts. Method: Using a within-subjects design (2x2x2), we examine the effects of touch type (active self-touch vs. object touch), anxiety induction (threat vs. safe), and emotion (fearful vs. happy faces) on facial emotion discrimination. Participants are asked to discriminate varying intensities of happy or fearful faces compared to their neutral counterparts. Applying a ‘threat of scream’ paradigm, blocks containing unpredictable, distressing screams (threat) are alternated with blocks that are free from screams (safe). During active ST, participants stroke the dorsal part of their left forearm with their right hand at a C-tactile targeted velocity while stimuli are displayed. The control involves the same motion stroking an object (object touch). ECG captures heart rate and variability as markers of affective state in each block. Trait and state anxiety are assessed using the State-Trait Anxiety Inventory. Expected Results: We expect enhanced sensitivity to fearful faces during threat blocks, indicated by a lower discrimination threshold. Importantly, we expect this effect to be mitigated by active ST, but not by the control condition. No significant changes are anticipated for anxiogenic contexts and active ST on happy faces. Outlook: If confirmed, this study could provide initial evidence that engaging in active self-touch has an anxiolytic function, reducing the effects of potentially threatening stimuli, and thus potentially reducing anxiety. Follow-up studies will then assess whether this effect differs between active self-touch and passive touch.

The processing of tactile stimuli relies on complex dynamics within cortical microcircuits across primary and secondary sensory cortices. Tactile evoked responses have been extensively studied using EEG and LFP measurements, but are poorly understood on the micro scale. In this project we use a neural mass model to explore the dynamics of somatosensory evoked responses. The model is capable of simulating EEG- and LFP-like signals, and therefore serves as a fundamental tool for studying the circuit dynamics of the somatosensory cortex. Insights from previous modelling approaches lack layer-specific and cell-type specific insights into the generation of ERPs. Here, we present a layer-specific somatosensory cortex model that aims to provide insights into the yet unknown dynamics underlying tactile evoked responses. The model consists of two cortical columns representing the primary and secondary somatosensory cortex, each with a granular, supra-, and infra-granular layer. Three interneuron types and one pyramidal population form each layer. The mean firing rate and the membrane potential of the cell populations are defined based on the established Jansen-Rit model. Connectivity, cell counts and synaptic properties are obtained from animal studies and prior models. An observation model transforms the firing rates and membrane potential into EEG- and LFP-like signals. This allows us to fit the model to real recordings, improving its interpretability and validity. Our approach reveals the role of different neuronal populations in generating tactile evoked responses and sheds light on the origins of early and late responses to tactile stimuli. In addition to providing a biologically plausible and efficient way to understand somatosensory processing, our model bridges the gap between macroscopic measurements and microscopic neural dynamics.

Early-life adversity (ELA) consists of experiences of harmful input (abuse) or lack of necessary stimuli (neglect) during youth development, which can affect brain structure and function and potentially lead to psychopathological outcomes. Changes in hippocampal volume due to ELA have been previously reported. However, no study has yet assessed the longitudinal structural and functional effects of adversities on different hippocampus subfields through a three-dimensional perspective of ELA (abuse, neglect, and household challenges). Thus, the neurobiological mechanisms of this relationship are not completely understood. In this project, we aim to achieve the aforementioned goal with the Adolescent Brain Cognitive Development Study, a large-scale longitudinal dataset comprising maltreatment questionnaire results and structural and functional magnetic resonance (MRI) imaging of 10,000 individuals aged 8-14. For that, T1w and T2w images of the hippocampus subfields will be segmented through the novel 'HippUnfold' pipeline, and an Exploratory Factor Analysis will be conducted to categorize questionnaire results into abuse, neglect, and household challenges. With that, we can evaluate ELA effects on the developmental course of the hippocampus microstructure (T1w/T2w ratio), in which accelerated maturation is expected to be linked with experiences of abuse (Study 1). In Study 2, an array of behavioral functional MRI tasks will build latent variables through Principal Component Analysis to represent memory and emotion processing and to detect intra- and interindividual differences. Structural and functional hippocampal maps will be used to predict these behavioral markers, followed by their application in combination with ELA dimensions. We hypothesize ELA effects on behavioral impairments to be more prominent in the anterior region of the hippocampus. Lastly, Study 3 focuses on understanding the role of resilience in the relationship between ELA and hippocampus maturation. Stressor resilience scores obtained through machine learning approaches will be combined with ELA factors and structural and functional hippocampal development into biophysical and latent change score models, in which we expect to find multiple latent dimensions supporting adaptation to ELA. With this project, we aim to contribute to comprehending the impact of ELA on the maturation of hippocampal structure and function in addition to probing brain markers of resilience and risk for ELA.

The perception of actions performed by others (henceforth action perception) is fundamental for human behaviour, enabling us to interact effectively with others in different contexts. During aging, brain mechanisms underlying sensory and cognitive functions undergo substantial declines, which may impact older adults’ ability for action perception. While previous studies have identified a distributed network of brain areas and temporal hierarchies of visual action perception in young adults, the effect of aging on the cognitive and neural representations of daily actions remains unclear. Here, we curated a set of photos of context-rich daily actions and quantified the visual, action-related, and social-affective features of each photo stimulus. We will first examine whether the intuitive action representation of older adults follows a similar organization as that of the young. In the main part of the study, we will conduct a magnetoencephalography (MEG) experiment, a technique with high temporal and spatial resolution, to reveal the underlying neural mechanisms and dynamics. Results of this study may shed light on how the aging brain processes and represents observed daily actions performed by others and help us better understand age-related changes in action perception in specific and social cognition in general.

Theory-of-Mind (ToM) is the ability to reason about other people’s thoughts and beliefs. A milestone in the development of ToM is achieved around 4 years, when children pass the critical test of ToM—the false belief task. A similar developmental breakthrough around 4 years is observed in reasoning about abstract concepts, such as possibility and relations, and it has been proposed that ToM and reasoning in these domains may share a general structure of reasoning. However, cultural differences in the development of both ToM and relational reasoning have been observed between the US and China, raising the question whether the development of reasoning across these domains follows a universal common trajectory or depends on culture. To address these questions, in this study, we examined the developmental trajectories of abstract reasoning (i.e., possibility reasoning and relational reasoning). We hypothesized that there is a positive correlation between children’s performance on ToM task and abstract reasoning tasks, and that this relation may depend on culture. We recruited typically developing 3- to 5-year-old children from Germany and China. Children participated in a battery of false belief tasks, a possibility reasoning task, a relational reasoning task, and an inhibitory control task in 2 sessions via online meeting. Preliminary data showed that German (N = 43) and Chinese children (N = 46) demonstrated similar age-related development in false belief task and possibility reasoning task. However, German and Chinese preschoolers showed different developmental trajectories of relational reasoning. Specifically, German children showed a consistent tendency to reason individually in the preschool years, while Chinese children demonstrated a transition from individual reasoners to relational reasoners between 3 and 5 years. Our findings evidenced that the development of ToM is positively correlated with possibility reasoning in both cultures but not with relational reasoning, suggesting that the development of ToM and possibility reasoning may depend on a general reasoning process that develops in the preschool years, independently of culture. In contrast, relational reasoning seems to be shaped by culture and accordingly is not associated with a more general development of reasoning ability at this age.

The large-scale organisation of the human brain and its relation to cognition is a topic that has been extensively studied over time. Cortex is known for its six-layered architecture with spatial variations across the cortical mantle but the functional consequences of these variations remain incompletely understood. This project aims to explore how agranular allocortical regions relate to isocortical regions, inspired by the dual origin theory of cortical organisation, which posits that there are separate waves of lamination arising from two 'proto' cortical origins.. In this project, we seek to examine this topic from an inside-out perspective, starting from the allocortical areas. We investigated the structural and functional dynamics between the allocortex and the isocortex in the human brain, the impact of these dynamics on human cognition, and the distinctions and connections between two specific allocortical subregions that are thought to be the dual origins, namely the hippocampus and amygdala. To do so, we examined the brain using an integrative and multi-scale approach. We first mapped the microstructural axes differentiating hippocampal from amygdalar patterning across the whole cortex using high-resolution structural MRI. Then we assessed functional relevance using resting-state functional MRI data. Our findings indicate a divergence of ventral and dorsal trends in the cortical projections of both functional connectivity and microstructure profile gradients of hippocampus and amygdala. Furthermore, our preliminary results point to within amygdala and hippocampus variations of aligning with either the ventral or dorsal trends of cortex. In future work, we will analyse openly available task-fmri data acquired during various cognitive tasks to further explore the dynamics of how hippocampus and amygdala subregions interact and diverge in different conditions.

The structure and function of the human brain are shaped by ontogenetic processes that interact with environmental input to allow for complex cognitive and behavioural skills to form. In particular, the social environment may be relevant for humans. Yet, how the social environment, and specifically variations in socioeconomic status (SES), may impact brain structure and function across the lifespan is not fully known. Understanding this relationship is important to create accurate models spanning biological and social factors influencing adaptation and mental health across the lifespan. In this project, we aim to elucidate how neurobiological processes of maturation and degeneration are impacted by individuals’ social environments, with a focus on aspects of SES. For this, we will model the lifespan trajectory of structural markers of cortical plasticity (cortical myelin) and graph properties of structural covariance networks in the HCP lifespan data using generalized additive models. This flexible modelling framework will allow us to compare the effects of SES on patterns of variation in (micro-)structural features and align those to a general sensorimotor-association axis. Charting growth curves for different SES groups will help identify cortical regions most impacted by SES and the critical periods where these are most pronounced across the lifespan. This will ultimately inform future studies on the mechanisms driving these changes in neurodevelopment.

The development of large-scale, dynamic models that accurately reflect the complexity of spatial cognition remains a significant challenge. We present an open-source, extendable Python toolkit to model spatial cognition in the framework established by the so-called BB model (Bicanski & Burgess 2018) and its predecessors – called the Spatial Cognition Box (SCB). The SCB is designed to generate egocentric and allocentric spatial cell responses and their systems-level interaction, in interactive environments of arbitrary shape and object composition. Recall of spatial snapshots and replay of trajectories, as well as mental navigation can be triggered at will. The SCB Toolkit thus offers a flexible and user-friendly way to generate neural responses to parallel behavioral experiments or for integration with other computational models. Finally, the SCB can easily be integrated with other toolkits such as RatInABox (T. George et. al., 2024). The toolkit features tutorials and comprehensive documentation, and we endeavor to make it an adaptable tool for both experimentalists and modelers to advance the understanding of spatial cognition.

This thesis investigates the hierarchical network architecture of auditory-speech processing in the autistic brain. Autism spectrum condition (ASC) is characterized by atypical sensory-perceptual processing and challenges in cognitive and social functions. Despite extensive research, a comprehensive neurobiological framework integrating both low and high-level abnormalities in ASC remains elusive. Neuroanatomy and brain imaging suggest that this architecture supports the integration of abstract concepts, cognition, and behavior. We will use resting-state functional magnetic resonance imaging (rs-fMRI) data from the Autism Brain Imaging Data Exchange (ABIDE) and Longitudinal European Autism Project (LEAP) to analyze interactions of sensory inputs from primary regions into motor and higher-order cognitive areas. Functional connectivity analysis will compare neural patterns in ASC individuals with typically developing controls. The study will assess differences across sensory modalities and examine the developmental trajectories of these network architectures from adolescence to adulthood, exploring potential compensatory mechanisms in ASC. We hypothesize that ASC individuals will exhibit distinct network architectures guiding sensory inputs compared to typically developing controls. These differences will manifest across auditory, visual, and somatosensory modalities, influencing sensory-motor integration, cognition, and behavior. Since the hierarchical model of auditory-motor speech perception, motor feedback is suggested to serve as an error prediction signal influencing the selection of auditory categories for decision making and action, we expect functional connectivity disruptions in the motor network to impact language comprehension. We also hypothesize that developmental trajectories of these network architectures will reveal compensatory mechanisms in ASC individuals with varying cognitive abilities. Results will identify inter-individual variations within the autism spectrum and differences in sensory modalities, which will allow to propose refined autistic phenotypes and a better understanding of the neurobiological framework for sensory-motor integration in ASC. Ultimately, the study is expected to provide insights into developmental changes and compensatory effects in the neural networks of ASC individuals.

Language is an integral feature of human behaviour and cognition. Linguistic processes have been shown to recruit several distinct regions of the brain with one of the most crucial regions being Broca’s area situated in the inferior frontal gyrus subserving language processing and speech production. Despite extensive research, the question of whether a structural and functional equivalence of Broca’s area exists in non-human primates remains open for a long time. Numerous approaches have been implemented to explore the evolution of Broca’s region to advance our understanding of the neuro-evolutionary underpinnings of human language. While homologues of Broca’s area have been depicted in macaques and chimpanzees, studies have documented a critical misalignment between peripheral sulcal boundaries at the macroscopic level and the cytoarchitectonic borders of Brodmann area 44 and 45 at the microstructural level. Furthermore, while most studies have drawn parallels based on non-human primates like macaques which are evolutionarily distant from humans, there exists a lack of detailed comparison on our closest evolutionary relatives – chimpanzees. Hence, a comprehensive evaluation between chimpanzees and humans to understand the microstructural details of the cytoarchitectonic organization of area 44 and 45 could help provide insight into the neural underpinnings of language. Additionally, beyond the cytoarchitectural delineation of Broca’s area, the Arcuate Fasciculus, a white-matter tract connecting Broca’s area to Wernicke’s also plays a crucial role in language. Interestingly an anatomical homologue of the arcuate fasciculus also exists in non-human primates, in whom language is absent. Understanding whether the arcuate fasciculus enables similar projections to the temporal and inferior frontal regions in non-human primates still remains a decisive question. Therefore, within the scope of the present project, we aim to delineate the cytoarchitectural characteristics of Broca’s area (BA 44 and 45) and map the endpoints of the Arcuate Fasciculus leveraging both gold standard histological techniques and cutting-edge MRI methods in 6 chimpanzee brains with particular attention devoted to the anatomical characteristics of the left and right hemispheres. This two-fold approach will allow us to examine the cellular architecture of Broca’s area (Brodmann area 44 and 45) as well as the neural connective pathways of arcuate fasciculus in chimpanzees.

The transition between interlocuters’ utterances in conversation (commonly called turns) is remarkably fast, with average silent intervals (gaps) in the order of about 200 ms. In the behavioral literature, it has been found that speech rate prior to the turn boundary affects gap duration. For instance, questions asked after an utterance are answered faster when preceded by faster speech rates. Speech entrainment has been proposed as a viable mechanism to anticipate the turn exchange and speed up turn-taking. Specifically, prior studies suggest that neural oscillations in the theta range (4-8 Hz) can entrain to the amplitude envelope of speech. The envelope shows maxima that accompany syllables. While these are not isochronous, their rhythmicity may be strong enough to entrain neural oscillations. In order to test the beneficial role of entrainment for reducing gaps between speakers we manipulated speech such that three conditions emerged: One with the original quasi-isochronous speech rate, one with isochronous syllable rate where (strong) entrainment is excepted, and one with anisochronous syllable rate, where reduced or no entrainment is expected. We established the effect of the three conditions on timing in turn-taking by auditorily presenting Yes/No-questions that twenty participants had to respond to. These questions were embedded in a Guess-Who game that participants played together with a prerecorded digital interlocutor. We found slower response times to anisochronous syllable lengths than original syllable lengths, in support of a role of syllable rate for identifying possible turn completion points.

This project will investigate the role of predictive processing in speech perception, focusing on its potential to explain atypicalities in autistic communication. Predictive processing is a theoretical framework that suggests that the brain constantly generates predictions about the world and then updates these predictions based on incoming sensory information. This process is thought to be crucial for both perception and production, as it allows us to make sense of the world and to interact with it effectively. In speech perception, it helps us to understand spoken language by allowing us to anticipate upcoming sounds based on our knowledge of language and the context of the conversation. Previous research has shown that individuals with Autism Spectrum Conditions (ASC) often exhibit differences in predictive processing compared to neurotypical individuals, and exposure to distracting sensory stimuli may interfere with their ability to direct attention to socially relevant information. These differences may contribute to their communication difficulties, such as problems with understanding spoken language and producing fluent, accurate speech. This project will use a variety of methods to investigate predictive processing and auditory attention in ASC, including behavioural experiments, neuroimaging studies, and computational modelling. Behavioural experiments will be conducted to assess speech perception under both intrinsic and extrinsic auditory degradation conditions. Participants will be presented with spoken words in varying levels of background noise and auditory distractions and asked to identify the words. In both conditions, participants will be asked to identify the words they hear. Measures of speech perception accuracy and response times will be collected compare the effects of both types of degradations and to assess auditory attention abilities. Electrophysiological recordings (MEG or EEG) and eye-tracking will be used to measure brain activity and cognitive load during speech perception tasks. Computational models of predictive processing will be employed to simulate the neural underpinnings of speech perception and these simulations will be compared to the MEG data. The findings of this research will help us to better understand the cognitive underpinnings of ASC communication difficulties and to develop new interventions. This poster is planned work currently in development and will be presented to gain feedback

Large language models have been experiencing a remarkable increase in performance and popularity, especially since the invention of the Transformer architecture first, and later the launch of OpenAI's ChatGPT 3.5. The language and reasoning capabilities exhibited by these models are increasingly nearing human levels of proficiency. This raises the question of whether the mechanistic innovations behind these models are related to workings of the human brain. Studies on representational alignment have shown that layer activations of transformer-based large language models can predict neuroimaging data acquired during language processing, yielding modest but significant correlations. However, these analyses rely mainly on static representations and a careful comparison of dynamics is not performed. We consider that analyzing relationships in dynamics should be deemed particularly relevant under the working hypothesis that the brain can be understood as a dynamical system. This project presents the idea of a new approach to understanding the relationship between large language models (LLMs) and the human brain in terms of their dynamics, by leveraging representational similarity, dynamical system theory and Koopman theory.

Successful maintenance of items in working memory has been linked to oscillatory mechanisms in low-frequency ranges. Previous research indicates that working memory content can be encoded as specific phases of these oscillatory signals. However, generalizing this mechanism to higher cognitive functions remains challenging. This study investigates whether higher cognitive content, such as the semantic representation of words, can be encoded in a phase-dependent manner. Specifically, we examine whether the binding of multiple semantic representations influences the phase coding of individual items. We conduct an EEG study during a verbal working memory retention task where participants memorized semantically bindable or unbindable adjective-noun pairs. To enhance the read-out of distributed neural activity related to working memory content, a high-intensity auditory impulse stimulus is presented. We train a decoder to differentiate activity patterns between adjectives and nouns. We hypothesize that for unbindable pairs, the semantic representation of each item would be maintained phase-dependently in working memory, allowing the decoder to distinguish between the two word categories. In contrast, for bindable pairs, the individual semantic representations would be integrated and not maintained separately in different phases, impairing the decoder’s performance.

Functional Magnetic Resonance Imaging (fMRI) is a neuroimaging technique that indirectly measures neural activities of the human brain over time by detecting changes in blood oxygenation levels. The analysis of fMRI data can provide information about functional networks, brain states, and brain dynamics between diseased and healthy individuals. However, it is hard to correlate this information with specific cognitive activities among individuals. To overcome this limitation, we will take multi-dimensional fMRI data which can be studied at different levels from 1D temporal data (individual voxel activity over time) to static 3D (parallel activation of voxels at a point in time) to 4D (spatiotemporal dynamics). We will unfold different perspectives of brain dynamics by exploring the hidden (latent) structure of these increasingly higher-dimensional data using machine learning methods. We want to leverage deep generative models (e.g. variational autoencoders and their variants) or deep encoder networks with contrastive learning to extract latent structures. Firstly, latent models on 1D temporal data could provide reliable estimates of functional connectivity or functional coupling matrix (e.g. for computing functional gradients) and can identify clusters of typical temporal signatures. Similarly, static 3D allows us to identify clusters corresponding to “snapshots” of brain states and common trajectory patterns. Finally, 4D (3D + time) data can be used to train sequence models (e.g. recurrent encoders or transformers) that learn to map the spatiotemporal patterns into the latent representation by studying short snippets of state space dynamics which might correspond to dynamical brain states. The outcomes of different levels help in correlating cognitive activities with neural activities. Overall, we hope deep latent learning tools will provide a methodological research direction that will yield better insight into the brain as a complex system.

Artificial neural networks (ANNs) have gained significant popularity and efficiency in recent years, particularly for data analysis and complex computations across various interdisciplinary fields. Despite their widespread use, the internal workings of ANNs often resemble a "black box," prompting researchers to explore and understand their functionality through diverse methodologies. Representational learning is a key area of study that aims to uncover the structure and operational mechanisms of these networks. One prominent approach within representational learning focuses on the geometric properties of both the network and the data it processes. In this study, we aim to analyze network representations by leveraging geometric methods, probabilistic machine learning techniques, and generative models such as variational autoencoders.

Dopaminergic neurons (DN) in the substantia nigra (SN) suffer from iron overload in age, increasing the risk for Parkinson’s disease (PD). Neuromelanin (NM) chelates iron, protecting DN against oxidative stress, but becomes toxic when oversaturated in age[1]. Therefore, monitoring iron in DN is crucial for early PD diagnosis and understanding its pathophysiology. Herein, we studied iron accumulation in chimpanzee brains across the lifespan[2]. We measured cellular iron in DN and extended the recently proposed non-invasive MRI-based method for DN-iron quantification [3] to the entire primate lifespan. 22 postmortem brains of chimpanzees (0.1-56 y, 8 f) were ethically collected within the Evolution of Brain Connectivity (EBC) project [2]. Quantitative maps of iron-sensitive effective transverse relaxation rate (R2*) were acquired on a 7T MRI scanner using multiparametric mapping [4]. For five brains (0.1,1.7,16,30 and 44 y), cellular iron quantification was performed using synchrotron X-ray fluorescence microscopy (XRF). The biophysical model of iron-induced MRI contrast based on the static dephasing theory [3] was informed by the cellular iron concentrations and used to estimate the contribution of DN to the effective transverse MRI relaxation rate (R2*) in the SN at different ages. The R2* in the SN and cellular iron load of DN both increased with age. The similarity in NM accumulation and total iron levels in SN between chimpanzees and humans qualifies the chimpanzee as a suitable animal to study the lifespan trajectory of human DN. The cellular NM-iron accumulation determined with XRF in a large group of neurons was well described by an exponential saturation with a time constant of 47 years. We linked the cellular iron concentration to the R2* using our biophysical model and tested its validity. The model was applicable for animals older than 16 years. We quantified the age-related iron accumulation of NM-clusters in DN in the SN of our closest relatives, chimpanzees, using XRF. We demonstrated that the biophysical model of iron-induced R2* in the SN is applicable after puberty opening the way for a non-invasive biomarker of DN density and iron load, which may be used to diagnose and monitor PD in humans. [1] Zecca et al., Nat Rev Neurosci 2004 [2] Gräßle et al., Methods Evol Ecol 2023 [3] Brammerloh et al., NeuroImage 2021 [4] Lipp et al., Magn Res Med 2023.