Session II: Monday, 30 June 2026, 18.00-19.15 (CEST)

Modal cognition, comprising the ability to represent and reason about different possible realities, is a defining feature of human cognition. However, its developmental trajectory remains debated. Eye-tracking studies suggest that infants can represent multiple possible object identities (Cesana-Arlotti et al., 2018, 2022), whereas tasks assessing reasoning about possible locations indicate a developmental breakthrough around age 4 (e.g., Mody & Carey, 2016; Redshaw & Suddendorf, 2016). It has been proposed that reasoning about possible identities and possible locations may rely on partly distinct cognitive processes (Brody et al., 2025; Jasbi et al., 2019; Mazalik & Halberda, 2026), which may help reconcile these conflicting findings. To address this issue, we adapted two established paradigms to create parallel tasks: a possible identities task and a possible locations task. In both tasks, adult participants are presented with either ambiguous stimuli (two possible identities/locations) or unambiguous stimuli (one possible identity/location). On each trial, subjects make a speeded judgment about whether a presented stimulus depicts a possible or impossible identity/location. We predict longer reaction times when participants must represent two possibilities compared to one. In addition, we test whether previously reported increases in pupil dilatation for ambiguous versus unambiguous object identities (Cesana-Arlotti et al., 2018, 2022) can be replicated, and whether a similar pattern emerges for reasoning about possible locations. Gaze patterns will be investigated exploratorily. Comparing RT and eye-tracking data between the two tasks will allow us to gain further insight into the cognitive mechanisms underlying modal reasoning about what and where.

The ability to understand and represent others' mental states, also known as Theory of Mind, is essential to social cognition (Premack and Woodruff, 1978). A developmental breakthrough occurs around the age of 4, when children are able to understand that others can hold false beliefs about the world (Wellman et al., 2001). At this time, children pass both location- and content-based false-belief tasks, with performance correlated across tasks (Grosse Wiesmann et al., 2017). To extend this line of research, we will examine performance in adults across content- and location-based false-belief tasks. Here, we aim to investigate whether an agent’s belief modulates reaction times (RTs) and gaze patterns during two false-belief paradigms: a false-content task and a false-location task. In each task, participants will view videos in which an agent either hides one of three different objects in a specific location (content condition) or hides an object in one of three different locations (location condition). A second agent will then either change the content or the location, while the first agent was either present (true belief) or not (false belief). At the end of each trial, participants will be presented with an image of one of the objects or locations and asked to respond as quickly as possible to one of two questions probing the real or believed contents/locations. We hypothesize that RTs will be shorter when the believed location/object matches reality than when it does not. Furthermore, in false-belief trials, RTs are expected to be faster when presented with the real compared to the believed contents/locations. Finally, we predict that RTs will be similar across the two tasks. Gaze patterns will be explored as an additional measure. Comparing RTs across conditions and between tasks will advance our understanding of the cognitive mechanisms underlying ToM, particularly in representing beliefs about “where” versus “what.”

TBA

Multisensory perception is shaped not only by sensory inputs but also by expectations based on prior experience or context. This dissertation project investigates the neural processing of multisensory contextual plausibility in modulating vibrotactile perception across the lifespan. Previous research (Kang et al., 2022, 2024) has shown that the subjective plausibility of perceptual events experienced in virtual reality (VR) can be affected by the degree of multisensory congruence between audiovisual contextual information and vibrotactile stimulation in immersive vehicle-riding scenarios. This preliminary study focuses on younger adults and parametrically manipulates the degree of plausibility of the vibrotactile augmentation. Based on a perceptual distance model (Rosenkranz & Altinsoy, 2023), vibrotactile stimuli are generated by systematically varying levels of perceived plausibility. Specifically, stimulus intensity levels are selected based on their perceptual distance from plausible tactile simulation experiences in VR, as derived from ratings provided by an independent sample in prior experiments. By replacing binary contrasts (plausible vs. implausible) with graded manipulations, the paradigm enables a fine-grained investigation of how cortical activity may scale along a continuum of multisensory contextual plausibility. Cortical responses are measured using functional near-infrared spectroscopy (fNIRS). In the dissertation project, continuous age analyses across the adult lifespan will further assess age-related differences in neural sensitivity to multisensory contextual plausibility. The findings will advance understanding of how the brain represents graded multisensory plausibility across adulthood and inform the design of more realistic virtual environments. Results of the current poster focus on findings from younger adults.

Children and adolescents typically display more inconsistent and less reward-maximising behaviour than adults, a pattern often described as non-greedy decision-making. While such variability could reflect adaptive exploration, accumulating evidence suggests that it may instead arise from learning noise, reflecting inconsistent belief updating in uncertain environments. Adopting a neurocomputational approach, the present study aims to disentangle learning noise and exploration in order to understand what drives non-greedy decision-making across development. 174 participants aged 8 to 30 years will complete a reinforcement learning task during functional magnetic resonance imaging (fMRI), with concurrent electrocardiogram (ECG) recording. Computational modelling will be used to estimate latent parameters underlying non-greedy decision-making. We hypothesise that younger participants will show reduced task performance, driven by higher levels of learning noise. At the neural level, this is expected to be reflected in weaker encoding of relative value in the ventromedial prefrontal cortex (vmPFC) and stronger activation of the dorsal anterior cingulate cortex (dACC). Physiologically, we expect reduced heart-rate deceleration in response to prediction errors in younger individuals, with the extent of deceleration inversely related to learning noise. To further probe learning noise, we examine its relation to exploration, cognitive effort, and real-life behaviour. We predict that learning noise will be unrelated to strategic exploration, but associated with reduced willingness to exert cognitive effort, as well as increased impulsivity, frustration, failures of control, and elevated (subclinical) ADHD symptoms. Together, this study aims to determine whether developmental differences in non-greedy decision-making reflect changes in choice strategies or variability in belief-updating.

Background: Glaucoma is an age-related visual disorder characterized by progressive visual field defects, leading to blindness in advanced cases. fMRI studies in diseases like retinitis pigmentosa and macular degeneration have reported task-dependent activity in the deafferented visual cortex despite the absence of bottom-up visual input. We recently reported similar task-related responses during visual stimulation in glaucoma as well. Do these activations reflect bottom-up plasticity or top-down modulation from higher cortical regions? Building on findings of our intervention-based study [Misra et al., MedRxiv, 2025], we motivate a 7T laminar fMRI study to address this question by resolving layer-specific activity in the visual cortex. Work leading up to the project: To probe plasticity in glaucoma, we used 3T fMRI to study the effects of motor training (unimodal intervention; UMI) and motor-cognitive training (multimodal intervention; MMI) on resting-state functional connectivity (FC) in 12 glaucoma patients (GL) and 20 healthy controls (HC), randomly assigned to UMI (5GL/11HC) or MMI (7GL/9HC). Resting-state fMRI was acquired before and after 3-months intervention. Whole-brain FC was estimated using bilateral seeds in Brodmann areas 1, 4, 9, 17, 18, 41, and 42, and longitudinal changes were assessed using paired t-tests (FWE corrected). In HC undergoing MMI, increased FC was observed between primary and secondary visual cortices and between auditory and visual regions (pFWE < p; 0.05), alongside increased motor-insular FC. UMI effects were limited to FC between cerebellum, midcingulate, and frontal eye fields. In GL undergoing MMI, increased FC was observed in audiovisual and visuomotor networks, including connections with the primary visual, inferior temporal, auditory, inferior parietal, and prefrontal cortices. No significant longitudinal changes were observed in GL undergoing UMI. Current project: These findings indicate the capacity for cortical plasticity in glaucoma, particularly within audiovisual and visuomotor networks. Therefore, we are currently conducting a 7T laminar fMRI study to investigate activity in the deafferented visual cortex during unimodal and cross-modal tasks. Laminar imaging will allow us to test whether task-related activations preferentially arise in middle layers (feedforward input) or superficial and deep layers (feedback), providing mechanistic insight into cortical plasticity and informing future rehabilitative strategies.

Introduction. Humans continuously process information from multiple sensory modalities. Multisensory integration (MI) enhances perceptual reliability by combining inputs across senses. Evidence suggests MI is altered in some pain conditions, but it is unclear whether this reflects general feature of chronic pain or is specific to central pain disorders. This study aimed to extend the previous investigation of audio-visual MI mechanisms to include the effects of nociceptive stimuli, examining it in healthy individuals and patients with chronic low back pain (cLBP). Methods. Thirty healthy participants and 30 patients with cLBP completed two newly developed MI pain tasks: a pain-induced multisensory illusion paradigm and a reaction time task with unimodal and bimodal visual, auditory, and nociceptive stimuli. Stimuli were applied to the arm or lower back in counterbalanced order. Reaction time data were analyzed using mathematical and computational modeling, including race model analysis. Results. Both groups showed faster responses to bimodal than unimodal stimuli (p < .001), with no group differences. Race model analysis revealed significant violations across all stimulus combinations and sites in both groups (p < .001), indicating bimodal stimuli were processed more efficiently than predicted by independent race models. In the multisensory illusion task, both groups reported fewer visual flashes when paired with painful stimulation (p < .001). Discussion. In our two newly developed MI pain tasks, we show that pain integrates with visual and auditory signals in a manner comparable to other sensory modalities. Notably, MI is preserved in cLBP, indicating that alterations are condition-specific rather than a universal feature of chronic pain.

Background The optic nerve is the white matter tract responsible for communicating visual information from the eye to the brain. From anatomical studies we know that the optic nerve has an organized fiber structure that maps each point in the visual field to a corresponding fiber bundle in the optic nerve. If the optic nerve is damaged, we suspect that the visual field would eventually be affected in an organized manner as well. The optic nerve can be damaged by certain eye diseases, such as glaucoma. Glaucoma is a progressive eye disease characterized by the loss of retinal ganglion cells. This loss is associated with structural damage in deeper parts of the visual system, including the optic nerve. If we could localize and quantify degeneration in the parts of the optic nerve corresponding to the visual field loss in people with glaucoma, we would be able to better explain the propagation of glaucomatous degeneration in the optic nerve. This, in turn, may help towards providing a more accurate prognosis and fine-grained evaluation of treatment responses. Our work leading up to the project In earlier work, we demonstrated that we can use 3T diffusion MRI to quantify clinically meaningful white matter changes in individuals with glaucoma by mapping the cross-section of the intraorbital optic nerve and assessing its fiber density and cross-section using fixel-based analysis. Our method can provide information beyond what current clinical examinations can provide to capture neurodegeneration. Current 7T project The location of glaucomatous damage in the optic nerve can vary a lot between patients. When describing neurodegeneration in glaucomatous brains, we usually describe averages between a healthy population and a glaucomatous population. To advance towards personalized healthcare, we aim to locally describe subject-specific neurodegeneration patterns in the optic nerve using our cross-sectional mapping approach to directly relate observed visual field defects to structural damage. We have applied our technique to cutting-edge diffusion MRI scans on the 7T MAGNETOM Terra.X Impulse Edition system in the Otto-von-Guericke-University Magdeburg, Germany. As of now, we are in the process of collecting scans of healthy controls and people with glaucoma. Using these high-resolution scans, we aim to quantify and visualize neurodegeneration in the cross-section of the optic nerve that directly relates to the observed visual field loss.

Introduction: Brainstem degeneration is an early feature of spinocerebellar ataxia type 3 (SCA3) (Faber et al., 2021) affecting structures responsible for binaural synchronization such as the medial superior olive. Consequently, binaural processing may be affected. Previously, we demonstrated that the upper limit of Huggins Pitch (HP) — a psychoacoustic phenomenon dependent on binaural synchronization, differed significantly between SCA3 mutation carriers and age-matched controls (Jacobi et al., 2024). This study aimed to investigate HP processing longitudinally in SCA3 mutation carriers to assess its potential as a marker for disease progression and onset prediction. Methods: Ataxia severity was assessed using the Scale for the Assessment and Rating of Ataxia (SARA). Upper limit of Huggins pitch perception was psychoacoustically determined via a 3 forced choice computer based behavioral task. Cortical HP N100 responses were recorded with magnetoencephalography (MEG). MEG recordings were analyzed using dipole source modeling. We collected longitudinal psychoacoustic and neuromagnetic data from 18 pre-ataxic and ataxic SCA3 mutation carriers. Results: HP thresholds significantly declined between baseline (t1) and follow-up (t2) (1.82 vs. 1.44 kHz; p=0.049) over an average interval of 2.83 (±1.08) years. Ataxic mutation carriers showed a significant HP threshold decline (1.65 vs. 1.30 kHz; p=0.039; n=9), while a trend was observed in pre-ataxic mutation carriers (2.20 vs. 1.75 kHz; p=0.09; n=4). HP thresholds significantly correlated with SARA sum score at follow-up (r=-0.82; p<0.001). We replicated the previous finding of a significant difference in the N100 amplitude between SCA3 mutation carriers and an age matched control group. Discussion: HP thresholds decline with disease progression in SCA3 mutation carriers, at a rate exceeding that expected from healthy ageing alone, suggesting potential as a marker for both tracking disease progression. While the behavioural decline has a neurophysiological correlate in the N100 amplitude, the psychoacoustic measure appears more sensitive, possibly reflecting early brainstem alterations not yet fully captured by cortical MEG responses. Future work will investigate oscillatory neural activity and connectivity signatures of cerebellar network degeneration, to better understand the network-level mechanisms underlying psychoacoustic deterioration in SCA3.

Contemporary models of the neurobiology of language predominantly rely on correlational neuroimaging methods and broad constructs inherited from linguistic theory (e.g., semantics, syntax), which often fail to map cleanly onto neural architecture. In contrast, the brain’s cytoarchitecture and the capacity to causally perturb cognition via neurostimulation suggest that far greater functional precision is achievable. Here, we propose elementalism: a neuromodulation-driven framework for modelling language in the brain grounded in high process specificity and whole-brain inference. We conducted a preregistered systematic review of neurostimulation studies using TMS, tES, and DES. Searches across four databases returned 12,763 records, of which 176 papers (1999–2025) met inclusion criteria following screening and extraction. These studies collectively reported 547 significant neuromodulation outcomes across 22 languages and a conservatively estimated 3,489 participants. Each outcome was qualitatively analysed to infer the most specific linguistic subprocess causally modulated, termed an element. Elements were organised within a seven-level hierarchy of process specificity, ranging from discipline-level descriptors (e.g., morphosyntax) to fine-grained activities and objects (e.g., non-adjacent dependency processing, articulatory control of specific speech units). Across the literature, traditional linguistic terms were frequently recoverable (e.g., semantics: n = 176; phonology: n = 74), but rarely delineated discrete functional tissue, instead exhibiting substantial degeneracy. Leveraging coordinate-level TMS data, we demonstrate how an elementalist approach enables bottom-up functional characterisation of language-eloquent regions. As a proof of concept, qualitative clustering of IFG stimulation sites reveals three plausible elements: executive control over language, temporal sequencing of speech units, and non-local feature combination. We argue that elementalism reconciles older modelling traditions by combining causal neuromodulation evidence, task heterogeneity, and whole-brain generalisation. The resulting open, continuously extensible database provides a principled resource for language neuroscience and relevance for awake craniotomy language mapping.

Obesity prevalence continues to rise globally, with more than half of the world’s population projected to be overweight or obese by 2030. Unhealthy eating behaviors, including heightened preference for high-calorie foods and hedonic eating in the absence of hunger, contribute substantially to the development and maintenance of obesity. Neuroimaging studies suggest that these behaviors are linked to dysregulation within brain networks involved in reward processing and decision-making. Recently, the gut microbiome has emerged as a potential modulator of eating behavior and metabolic outcomes through the microbiome–diet-gut–brain axis. In particular,Prebiotic supplementation has been proposed as a promising strategy to influence gut microbial composition and, consequently, neural responses related to food reward. The present preregistered study investigates whether a microbiome-targeted prebiotic intervention influences food-related decision-making and gut–brain interactions in adults with overweight and obesity. In the randomized controlled MIFOOD trial, participants receive either a daily 30 g inulin supplementation, a neurocognitive behavioral intervention, or placebo across six months. Food decision-making is assessed using task-based functional MRI, while gut microbiome composition is measured using shotgun metagenomics. Additional outcomes include eating behavior measures, blood-based markers such as short-chain fatty acids and gut hormones, metabolic markers, and body composition. Confirmatory analyses will test whether the prebiotic intervention alters neural correlates of wanting-related food decision-making compared to placebo. Exploratory analyses will investigate potential microbiome and metabolic pathways underlying these effects, as well as potential sex differences in intervention responses.

The hippocampus is traditionally associated with memory, yet functional neuroimaging and neuropsychological evidence suggests a broader role in processing scene information. Hippocampal activation occurs during perception of contextual, spatial, and navigation-related stimuli, as well as during scene imagination – processes thought to rely on internally maintained scene representations. It remains unclear whether this activity reflects perceptual scene processing or successful encoding into long-term memory. To investigate whether hippocampal activation during encoding relates to later recognition, young adults (N = 29; 18 - 35 years) completed an implicit encoding task during 3T fMRI. Participants viewed 120 colour scenes and 120 greyscale objects for 3000 ms each, making property-based judgements (e.g., “Is the stimulus jagged or smooth?”). Outside the scanner, participants completed an old/new recognition test with confidence ratings. Explanatory variables for each category and memory outcome were entered into a GLM in FSL, and percentage signal change was extracted from a probabilistic hippocampal ROI. Category × memory interactions were assessed with a 2 × 2 repeated-measures ANOVA, separately for each hemisphere. In the left hippocampus, there was a significant main effect of stimulus category (F = 8.58, p = 0.007), with greater activation for scenes than objects, but no effect of memory outcome (F = 0.65, p = 0.429) and no interaction (F = 2.10, p = 0.158). In the right hippocampus, no significant effects of stimulus category (F = 3.56, p = 0.070), memory outcome (F = 0.08, p = 0.783), or the interaction (F = 2.32, p = 0.139) were observed. Behaviourally, memory was better for objects than scenes (p = 0.001), indicating the left hippocampal effect was not driven by memory performance. These findings suggest that hippocampal engagement during encoding is driven by stimulus category, with preferential engagement for scenes over objects, rather than subsequent memory success. The effect was left-lateralised, contrasting prior reports of a right bias. Overall, the results highlight a role for the hippocampus in processing spatial scene information rather than supporting memory encoding in a domain-general manner. Future work could explore relationships with scene memory confidence or item-level memorability to clarify links between scene perception and memory.

The extent to which reward history interferes with inhibitory control in adolescent obesity remains unclear. Using an adolescent subsample of the HCP-Development study, we will test whether previously rewarded or loss-associated cues disrupt No-Go performance in adolescents with obesity to a greater extent compared to their healthy-weight peers. In a Go/No-Go paradigm, stimuli were either novel or previously associated with either a reward or a loss during a previous task. Firstly, we hypothesise that adolescents with obesity will show a greater rate of reward history driven commission errors compared to their healthy-weight peers. Secondly, to further characterise the cognitive processes underlying task performance differences between the two groups, we will examine drift-diffusion parameters. We expect a stronger starting point bias towards responding on No-Go trials with previously rewarded stimuli, along with noisier evidence accumulation in adolescents with obesity. Thirdly, we will investigate whether adolescents with obesity exhibit altered task-related neural responses in brain areas involved in habitual control of behaviour and inhibition. Identifying behavioural and neural correlates underlying compromised interplay of habitual and goal-directed behaviour may refine mechanistic accounts of executive function impairment in adolescent obesity and inform future intervention strategies.

Cognitive control is a multi-scale, dynamic, context-dependent process. However, efforts to understand how it emerges from the brain using task fMRI have largely relied on single-scale, static, context-agnostic approaches. This epistemological mismatch has limited our understanding of cognitive control. Here, we introduce a computational method that aligns with the ontological characteristics of cognitive control and the design of cognitive neuroimaging tasks. Our approach integrates methods from statistical physics (Iterative Coarse Graining, ICG) and routine neuroimaging (dynamic functional connectivity, General Linear Model) into a unified pipeline. The method iteratively clusters pairs of brain regions based on the strength of their task-related dynamic functional connectivity, yielding progressively coarser scales of organisation. We first validated the method in silico using simulated timeseries with prespecified modular structure and varying task designs. We compared our approach with commonly used task fMRI methods, including Pearson correlation and background connectivity, and found that it successfully isolated dynamic, context-dependent functional connectivity, outperforming traditional approaches. Next, we applied the method to openly available 3T task fMRI data from 48 healthy adults performing three tasks tapping into the core components of cognitive control (Go/No-Go, Task Switch, N-back). After standard preprocessing and denoising, data were parcellated using a 1024-region cortico-subcortical atlas. For each task, ICG generated 11 spatial scales (1024 regions to a global signal). We focused on (1) how the multi-scale scaffold emerged (which regions paired at each scale) and (2) how regional beta coefficients evolved across scales (how regions were recruited). We also examined similarities and differences across tasks in light of theories of shared and distinct cognitive control components (Miyake et al., 2000). Our method revealed a conserved cross-task multi-scale connectivity backbone reproducing the three cortical gradients of functional organisation. Despite this shared structure, each task recruited the backbone differently across scales, suggesting task-specific functional reconfiguration. Together, our results support the multi-scale, dynamic and context-dependent nature of the neural mechanisms underlying cognitive control.

The ongoing continuous experience of life is automatically segmented into smaller and meaningful events. Such segmentation is a core component of the cognitive system that affects memory, prediction of the future and enables effective planning. However, past research has mainly relied on unimodal perceptual experiences. Thus, there is a significant gap in the literature regarding how the brain represents the complexities of naturalistic multisensory events. Our experiences are inherently multimodal. If asked to think about a recent workday, we could divide this event into smaller units such as arriving at work, a morning meeting, and a lunch break outside to enjoy the first day of spring. Each event is made up of multiple sensory modalities, the chirping of birds in the near distance, the sight of bluebells and crocuses peeking up from the earth, and the sound of your friend’s voice on the phone as they tell you about their weekend. Previous research has shown that event boundaries can influence episodic memory by disrupting temporal order memory for items crossing a boundary, increasing the perceived temporal distance between events, and enhancing memory for contextual information. However, it remains unclear how boundaries at different hierarchical contextual levels influence these memory processes. Little is known about how the magnitude of a boundary (degree of contextual change) produces graded effects on memory organization. This study addresses this gap by examining how different types of event boundaries influence episodic memory. Participants study events where contextual shifts occur either within a single sensory modality (e.g. a change in visual or auditory context), or simultaneously across multiple modalities (concurrent changes in both visual and auditory contexts). Effects are measured on temporal order, perceived temporal distance, source memory, and recognition memory. Importantly, this design allows us to test if larger contextual shifts (changes across multiple sensory modalities) produce stronger boundary-driven memory effects. This study is part of a project aiming to understand how multimodal and hierarchical events are represented in episodic memory. Future neuroimaging studies will focus on identifying neural signatures of event encoding and recall, evidence for cross-modal integration of boundary cues, and the effects of hierarchical levels and predictability on encoding and subsequent memory performance.

Bodily rhythms, such as the heartbeat or respiration, generate continuous interoceptive input to the brain, prominently via vagal afferents. Even at rest, these signals engage the central autonomic network (CAN), a distributed set of brain regions that integrates autonomic information and coordinates autonomic output. Our prior work linked resting heart rate variability (HRV), an index of vagal cardioregulation, to functional connectivity (FC) in cortical midline regions, often considered CAN nodes. However, scan-averaged connectivity may miss interoceptive dynamics reflected in HRV-linked fluctuations in functional connectivity. Here, we test whether intra- and inter-individual differences in HRV are associated with the variability in FC of the CAN using an expanded, literature-informed CAN parcellation. We will analyze 7T resting-state fMRI with concurrent finger photoplethysmography in Amsterdam–Trondheim (N=56; 2×15 min) and test replication in the Welsh Advanced Neuroimaging Database (N=178; 155 with 7T; 10 min). CAN ROI time series will be analyzed in sliding windows (100s; 1-TR step). Within each window, ROI-to-ROI FC will be estimated using window-weighted least squares and Fisher z-transformed. HRV is defined as the average RMSSD across 2 minutes. Dynamic FC variability will be quantified per CAN edge as the across-window SD of Fisher z-transformed ROI-to-ROI connectivity. For each edge, we will test whether mean RMSSD predicts edge variability using a flexible factorial model with subject and run as well as covariates (age, sex, mean FD); edge-wise results will be FDR-corrected. We expect higher mean HRV to be associated with higher CAN FC variability and a reproducible subset of HRV-associated edges. Within participants and one session, we expect higher dynamic HRV to be associated with greater dynamic CAN FC. If supported, these findings connect trait vagal cardioregulation to the flexibility of intrinsic autonomic circuitry during rest, beyond what scan-averaged FC captures. A replicated, network-defined association between HRV and CAN would clarify how individual differences in vagal function shape the moment-to-moment coordination of central autonomic processing, advancing mechanistic models of interoceptive control.

Background: The Montreal Cognitive Assessment (MoCA) has emerged as a routine neuropsychological instrument for telemedicine, including a 5-word memory task that yields three retrieval conditions: free recall, category-cued recall, and multiple-choice recognition. Hippocampal subfield volumetry has demonstrated that specific subfields are differentially affected across the Alzheimer’s disease continuum and are selectively associated with distinct memory processes. However, whether the MoCA’s memory retrieval conditions map onto specific hippocampal subfield volumes remains unexplored, particularly when administered via telemedicine. Objective: To investigate the associations between hippocampal subfield volumes and remotely administered MoCA memory retrieval profiles (free recall, category-cued recall, and multiple-choice recognition) in a telemedicine memory clinic cohort spanning the cognitive spectrum from Subjective Cognitive Complaints (SCC) through Mild Cognitive Impairment (MCI) to mild dementia. Methods: Approximately 60 participants will be recruited from the Outpatient Clinic for Memory, Dementia and Parkinson’s Disease through the National Telemedicine Network. The sample will include patients on the Aegean islands across the cognitive disorder spectrum from SCC through MCI to mild dementia. MoCA memory data will be extracted from clinical assessments at diagnosis, recording performance across all three retrieval conditions. Structural T1-weighted MRI, undergone no further than 6 months from MoCA administration, will be processed using the FreeSurfer hippocampal subfield segmentation module to extract bilateral volumes of all hippocampal subfields. All segmentations will undergo visual quality control. Partial correlations between hippocampal subfield volumes and recall condition scores will be computed, controlling for age, sex, education, and estimated total intracranial volume. Linear regression will test whether subfield volumes differentially predict encoding versus retrieval failure profiles. Conclusion: This study is designed to examine whether the MoCA’s brief memory assessment architecture, including cued and recognition conditions, captures hippocampal subfield integrity relevant to clinical interpretation when administered via telemedicine. Secondly, it demonstrates the feasibility of conducting cognitive neuroimaging research in geographically dispersed island populations through telemedicine.

Odours we experience daily are composed of multiple components, and the perception of an unfamiliar note within a familiar smell depends on situational context. Detecting unusual notes in food odours is essential to avoid ingesting contaminants, yet this caution must be balanced against the risk of rejecting safe food. Our previous research showed that the perceptual dominance of an attended food component in an odour mixture declines sharply as the proportion of a contaminant increases. In contrast, the perceived dominance of an attended non-food odour remains comparatively robust. In the current study, we investigate whether this asymmetry depends on knowledge of an odour’s identity. Healthy participants complete a learning task establishing an association between an odour and its source. Successful manipulation demonstrates that different participants can learn to identify the same ambivalent odour as originating from either an edible or an inedible source. Participants rate familiarity, intensity, pleasantness, and edibility before and after learning to track learning-induced changes in odour evaluation. Analysis of perceptual dominance decisions across a range of odour mixtures tests whether learned odour identity alters the perceptual organisation of ambiguous odours, and whether identity-based learning alone is sufficient to reproduce previously observed differences between food and non-food odour processing. Using functional MRI, we further examine whether learned odour identity is reflected in stimulus-specific neural activation patterns. Multivariate pattern analyses assess whether neural responses to ambivalent odours resemble patterns evoked by canonical food or non-food odours, allowing us to test whether learned semantic identity reshapes the neural representation of olfactory stimuli.

Human perception is not a merely a passive reflection of sensory input but is strongly shaped by current goals, expectations, and task demands. Prior work has shown the anticipatory and task-driven nature of perception (Uithol et al., 2021). Building on this idea, the present fMRI study investigates how task instructions modulate the neural processing of faces and expressions. A key innovation of our paradigm is that participants first completed a calibration step, where they select faces that, in their view, best represent gender categories (male, female, or androgynous). This procedure yields an individualized stimulus set that accounts for variability in how gender cues are perceived. With this calibrated stimulus set, we then tested whether neural responses differ when participants perceive and categorize either emotion or gender of the same stimuli. Multivariate pattern analysis (MVPA) was applied to decode task-dependent activation patterns across the brain. Results suggest that task instructions bias activity in early visual cortex (V1, V2), face-sensitive areas (FFA) and key limbic areas. We successfully decoded the cued task from distributed activation patterns. These findings highlight how top-down, task-driven influences shape perception, providing insights into the nature of goal-driven cognition.

The dorsolateral prefrontal cortex (DLPFC) is a principal target for transcranial magnetic stimulation (TMS) in treating major depressive disorder, with therapeutic effects thought to be mediated by its connectivity with the subgenual anterior cingulate cortex. As both regions are involved in autonomic regulation, short-term heart rate changes following DLPFC stimulation may serve as physiological markers to identify stimulation targets. We employed neuro-cardiac guided TMS in a cohort of healthy participants to examine the effects of stimulation intensity and DLPFC target specificity on heart–brain coupling (HBC). We used generalized additive models to assess nonlinear effects of stimulation intensity and target location on HBC, while accounting for pain ratings and other side effects. Intra-subject repeatability across three sessions was evaluated using intraclass correlation coefficients. We observed a non-linear modulation of HBC depending on stimulation intensity and target location, with greater effects at the F3 lateral and F3 posterior targets compared to sham. By evaluating these effects across sessions within participants, we demonstrate the robustness of our results beyond the influence of pain and other side effects on HBC modulation. Exploratory analyses of the directionality show a consistent decrease in HR only at the F3 lateral target with suprathreshold stimulation. These results demonstrate that HBC is modulated in a target- and intensity-specific manner, with particularly consistent effects at F3 lateral sites within the DLPFC. The findings enhance the understanding of TMS-modulated heart-brain interactions, offering a potential framework for optimizing individualized rTMS treatment protocols for depression.

The goal of the proposed project is to understand the neural and computational substrate of perceptual multistability. When humans and animals are exposed to particular forms of ambiguous stimuli (e.g., the Necker cube and binocular rivalry), they experience multiple percepts that switch over time. The existence and dynamics of perceptual switching have been investigated with a diverse range of tools and approaches, and have provided important insights into a wide range of questions in cognitive neuroscience, from core principles of information processing in the brain to aspects of higher cognitive processes such as visual awareness. Although the classical accounts have provided important insights into this phenomenon, they struggle to explain multiple aspects of recent data. The main goal of this project is to develop a novel framework, based on reinforcement learning, in which perceptual multistability is treated as an active decision process, with perceptual alternation being an action that results from an internal decision-making process. The project will also aim to identify how these computations are realized at the neural level, through analysis of the dynamics of relevant regions such as the prefrontal cortex during perceptual multistability.

Research on the neural correlates of conscious somatosensory perception dates back to the 1960s, when Benjamin Libet demonstrated that cortical activity could be recorded even in the absence of reported stimulus awareness. Subsequent EEG studies using electrical stimulation identified a mid-latency event-related potential (ERP) component occurring approximately 140 ms post-stimulus as a likely neural correlate of conscious somatosensory perception, whereas earlier components around 60 ms were thought to primarily reflect physical stimulus properties rather than perception itself. More recently, studies employing mechanical stimulation have reported perception-related neural activity as early as ~60 ms post-stimulus, suggesting that the temporal dynamics of conscious perception may depend on stimulus modality. In addition to stimulus modality, pre-stimulus brain states appear to play a critical role in sensory awareness. In particular, alpha oscillations have been strongly linked to perceptual performance, with both pre-stimulus alpha amplitude and phase-locking strength correlating with detection rates. However, it remains unclear how the brain uses anticipatory cue information to prepare for incoming somatosensory stimuli, and whether preparatory mechanisms differ between artificial electrical stimulation and more naturalistic mechanical stimulation. In this project, we investigate mental preparation for near-threshold mechanical stimuli using an EEG-based detection paradigm with graded stimulus intensities around the perceptual threshold. The study aims to extend previous findings by examining how attentional state and preparatory neural activity during the pre-stimulus window modulate conscious mechanical versus electrical somatosensory perception.

Theories of learning suggest that prediction errors drive how we update our beliefs about the world. Computational models propose that the rate at which we learn from these errors should adapt to environmental volatility: updating rapidly when outcomes are unstable and slowly when they are stable. However, most research has focused on how objectively experienced volatility shapes learning. Less is known about whether subjective beliefs about volatility can independently modulate reward learning behaviour and its neural substrates. We address this question using a two-armed bandit reversal learning task where participants track which of two options is more likely to yield a reward. Crucially, we independently manipulate objective environmental volatility and participants' subjective beliefs about volatility through explicit instruction. This creates conditions where participants experience identical reward environments but hold different expectations about their stability (believe-stable vs. believe-volatile), allowing us to dissociate the effects of experienced versus expected uncertainty on learning. We have taken this paradigm into an fMRI study using a novel multi-echo acquisition at 3T. Behavioural pilot data from the fMRI-adapted task (N=10) confirm that objective volatility significantly increases switch rates (p=.030, dz=0.815), and that believed volatility shifts behaviour in the same direction (dz=0.442). Computational modelling using the Rescorla-Wagner framework successfully captures individual learning trajectories, generating trial-by-trial prediction error estimates for each participant. Our pilot fMRI data (N=4) demonstrate that reward prediction error signals estimated from individually fitted Rescorla-Wagner models correlate with striatal activation, and condition-wise comparisons suggest that both objective and believed volatility modulate the strength of striatal prediction error encoding. In the full study (target N=30), we will extend these analyses with multivariate decoding to investigate whether patterns of brain activity can distinguish experienced from expected volatility, and whether these are encoded in shared or distinct brain regions. Updated results from the ongoing study will be presented.

Predictive coding suggests that the brain generates predictions about sensory input, compares them with incoming signals, and updates an internal world model based on prediction errors. Potential neural implementations of this influential framework remain debated. In discrete settings, a world model can generate predictions for a finite set of possible observations. This becomes challenging in continuous environments, where observations cannot be enumerated. Stochasticity exacerbates the problem because the same observation may lead to multiple plausible futures, not captured by a single “best prediction.” We introduce a probabilistic world model that generalizes predictive coding to continuous and stochastic settings and identify potential anatomical correlates in the brain. Rather than minimizing prediction error, the model maximizes the likelihood of observed data, explicitly modeling the environment’s probability distribution. We tentatively map the computation of probabilities onto the medial prefrontal cortex (mPFC), implicated in schema formation, abstract task representations, and the integration of hippocampal information across episodes. The hippocampal system (modeled as an RNN) encodes temporal context and interacts bidirectionally with the OFC/mPFC component (a normalizing flow), thereby implementing the probabilistic world model. At the algorithmic level, we employ normalizing flows, neural networks that transform unknown distributions into a tractable form via learned invertible mappings. We hypothesize that the same circuitry used to quantify the likelihood of an episode may also support the generation of novel imagined episodes, making the mPFC/OFC a plausible substrate, given its bidirectional interactions with the hippocampus and its proposed role in cognitive maps of latent or partially unobservable states. The model learns sequence probabilities in a self-supervised way without biologically implausible reconstruction loss. Our formulation resolves the core issue in continuous stochastic environments and offers a common account of replay, planning, and imagination.

Childhood environment plays a critical role in shaping brain development, impacting an individual’s lifelong cognition and mental health. Previous studies have suggested that environmental exposures may either accelerate or delay brain development, with downstream consequences for psychopathology and cognitive function. The hippocampus is a uniquely plastic structure and a key node in distributed networks underlying, for example, emotion and the processing of information in its environmental context. Research has demonstrated associations between isolated aspects of the childhood environment and hippocampal macrostructure, which although informative, may fail to account for the additive effects of exposures. Moreover, longitudinal studies have mainly relied on the estimation of age-related mean differences, offering limited guidance to track individual change over time. In this context, our main objective is to examine how individual rates of change in hippocampal macrostructure, across its anterior-posterior and proximal-distal axes, are reflected by the environment in its totality – i.e., an “exposome”. For this, we will use T1-weighted imaging data from 828 participants across three time-points from the ABCD Study (age range 9 – 14 years), quantified using a novel unfolding approach implemented in the HippUnfold toolbox. To track individual changes in hippocampal macrostructure, a normative model using Hierarchical Bayesian Regression with SHASH likelihood will be fitted with the Predictive Clinical Neuroscience Toolkit. Resulting normative Z-scores will then be used to estimate conditional SDS-gain metrics (also referred to as velocity scores), which measure an individual's rate of change relative to their previous measurements. Finally, a composite exposome score, based on a previously validated exposome factor spanning socioeconomic, family, and school-related domains, will be computed and used as the predictor for the velocity scores across hippocampal axes. With this study, we hope to clarify how the environment contributes to the pace of hippocampal macrostructural development, and to provide evidence on region-specific plasticity during a period of great environmental sensitivity.