How the human brain specialized for language and cognition is a fundamental issue for the cognitive and brain sciences, motivating the deep question: what makes us unique? I will present recent scientific evidence, including from collaborative work with the Max Planck Institute for Human Cognitive and Brain Sciences, on structural and effective connectivity studies in human and nonhuman primates. The results provide insights into how the human arcuate fasciculus crucial for language and the fronto-temporal cognitive brain system evolved from a common primate ancestor to humans, monkeys and apes. I also overview behavioral, neuroimaging and neurophysiological studies using Artificial Grammar Learning paradigms with both nonhuman primates and human patients. At every juncture, the results from our and the broader field of research regularly identify both cross-species correspondences and human-unique specialization, informing theoretical positions on primate evolutionary prototypes and how the human language and cognitive system evolved.

Young children show large individual differences in the number of words they produce and/or understand. Previous studies showed that these differences are partially attributable to genetic differences between children. In this talk, I will discuss genetic influences contributing to vocabulary size during early development. Although this period only spans the first few years of life, I will show that the genetic factors associated with vocabulary size may differ for infancy and toddlerhood. I will then discuss how these differences in genetic architecture shape associations with later-life traits. For example, especially genetic influences identified for word understanding in toddlerhood are related to various language and literacy-related traits assessed in middle childhood and early adolescence. Finally, I will discuss the potential relevance of genetic influences underlying early language development for behavioral traits, such as Attention-Deficit/Hyperactivity Disorder.

Abstract tba

The fundamental question in cognitive neuroscience—what are the key coding principles of the brain enabling human thinking—still remains largely unanswered. Evidence from neurophysiology suggests that place and grid cells in the hippocampal-entorhinal system provide an internal spatial map, the brain’s SatNav—the most intriguing neuronal coding scheme outside the sensory system. Our framework is concerned with the key idea that this navigation system in the brain—potentially resulting from evolution—provides the blueprint for a neural metric underlying human cognition. Specifically, we propose that the brain maps experience in so-called ‘cognitive spaces’. We test the overarching model that—akin to representing places and paths in a spatial map—similar coding principles are involved in the formation of such cognitive spaces. Importantly, we investigate if these domain-general principles support a broad range of our fundamental cognitive functions, ranging from spatial navigation, memory formation, learning, imagination, and perception to time processing, decision making, and knowledge acquisition. In this talk, I will give an overview of our theoretical framework and experimental approach and will present show-case examples from our fMRI, MEG, and virtual reality experiments identifying cognitive coding mechanisms in the hippocampal-entorhinal system and beyond.

Social connections are proposed to be a fundamental basic need of humans. Yet, loneliness and isolation are increasing in societies all around the world, particularly in young people (Hammond 2019, Twenge 2019). How does a lack of social contact affect the brain and cognition? While it is not possible to assess whether an animal subjectively feels lonely, animal models using objective isolation (i.e., assigning social animals to isolated living conditions) can give insight into the causal effects of unmet social needs. Animal research has consistently shown that a lack of social interaction leads to increased reward sensitivity, higher anxiety and inflexibility during learning – particularly during adolescence (Tomova et al. 2019, Orben et al. 2020). Yet, it is unclear how well results from animal models of isolation can be translated to humans. Do social isolation and loneliness in human adolescents cause similar modulations in brain function and cognition? Previous research in adult humans has shown that acute loneliness induced via short-term isolation affects brain functioning in a similar level as food craving after fasting (Tomova et al. 2020). In adolescents aged 16-19 years, short term isolation increases reward sensitivity (indicated by increased reward seeking and reward learning) and threat learning (Tomova et al. 2022). Access to virtual social interactions remediates some, but not all effects of isolation. The implications of this research in the light of loneliness and mental health problems will be discussed.

Transformer-based large-scale language models are able to generate highly realistic text. They are duly able to express, and at least implicitly represent, a wide range of sentiments and color, from the obvious, such as valence and arousal to the subtle, such as determination and admiration. Here, we explore these representations and their use for understanding the inner sentimental workings of single sentences. To do this, we take advantage of recent ideas in distributional reinforcement learning by training predictors of the quantiles of the distributions of final sentiments of sentences from the hidden representations of an LLM applied to prefixes of increasing lengths. After showing that predictors of distributions of valence, determination, admiration, anxiety and annoyance are well calibrated, we provide examples of using these predictors for analyzing sentences, illustrating, for instance, how even ordinary conjunctions such as “but” can dramatically alter the emotional trajectory of an utterance. We then show how to exploit the distributional predictions to generate sentences with sentiments in the tails of distributions. This is joint work with Chris Gagne.

Recovery from any spinal cord injury – and its attendant neurodegenerative processes – can follow a complicated trajectory spanning several years after injury. The ability to track injury-induced structural changes across the neuroaxis provides the opportunity to quantify pathological processes driving disability and recovery-related plasticity. During my talk I will present evidence from quantitative MRI to highlight the role of preserved tissue bridges, and the extent of progressive volume, myelin and iron changes along the projections of the corticospinal tract. Moreover, I will show how serial myelin and iron sensitive multiparametric mapping during a period of intensive motor skill acquisition revealed temporally and spatially distributed, performance-related microstructural changes in the grey and white matter across the motor system in SCI patients. Finally, I will show latest developments of high-resolution MRI sequences at 7T and optimized post-processing methods to assess the interaction of degenerative changes and recovery-related plasticity at the level of the spinal cord and brain, simultaneously.

The development of real-time functional magnetic resonance imaging (rt-fMRI) has expanded the reach of neurofeedback for self-regulation training of brain activity. Over the last 15 years there have been increasing efforts to utilise fMRI-neurofeedback in neurorehabilitation and psychiatry. I will first explain the setup and basic signal properties utilised for fMRI-based neurofeedback. I will then describe the rationale for its use as a therapeutic tool in psychiatry and neurology, focusing on our development of protocols for depression and Parkinson’s disease. I will review the current evidence for clinical fMRI-neurofeedback protocols and provide an overview of developments in the field. I will go on to discuss how clinical fMRI-neurofeedback can be taken to the next level, focussing on questions of protocol and trial design. In particular I will outline our current work incorporating information obtained from other neuromodulation approaches (such as deep brain stimulation) in the design of neurofeedback protocols (https://neuripides.eu). I will also argue that neurofeedback treatment needs to incorporate elements of psychological therapies and transfer technologies in order to become a clinically viable therapy.

The rules and mechanisms allowing for the development of sensorimotor maps in the human brain are still poorly understood. Congenital and acquired hand loss are key models for studying the impact of developmental stages on sensorimotor organisation and reorganisation. But the driving mechanisms, limits and behavioural consequences of these changes are still debated. In particular, it is not clear how topographical relationship and compensatory behaviour interplay in shaping brain (re)organisation. For example, increased BOLD activity was reported in the missing hand cortex during arm, lips and feet movements of congenital one-handers (Makin et al., 2013; Hahamy et al., 2017). However, it is not clear whether this dramatic brain remapping resulting from atypical development bears any functional relevance (Muret and Makin, 2021). In this talk, I will present a series of experiments combining functional MRI (3T) and behavioural assessments to investigate i) the role of cortical proximity and body usage in such remapping, and ii) the information content underlying it (see Muret et al., 2022). Using conventional univariate fMRI analysis, I will show that sensorimotor representations in humans are highly plastic, but that this plasticity is restricted by the developmental stage of input deprivation rather than by cortical proximity (Root*, Muret* et al., 2022). Using multivariate analysis, I will also show that increased usage of compensatory body-parts per se can result in different patterns of altered information content. In addition to replicating previous findings, these results further tone down the role of cortical proximity in both remapping and atypical development. Finally, these results suggest that some of this remapped activity may be functionally relevant, but that different mechanisms may be at play depending on the developmental stage of input deprivation.

There is an increasing consensus in the neuroscience community that the next major advances in neuroscience require us to span scales, species, and tools. In this talk, I’ll present examples of how our group is taking on this challenge by directly comparing MRI with microscopy. I’ll overview the technical challenges for performing these kind of studies, particularly in whole-brain samples. I’ll describe a range of kinds of investigation that this can enable. Time permitting, I’ll discuss how we’re aiming to integrate this into a broader range of imaging studies, including virtuous cycles of discovery-hypothesis-discovery linking in-vivo, ex-vivo, animal, and population neuroimaging.

Optically Pumped Magnetometers (OPMs) have recently become small enough and sensitive enough to measure magnetic field changes from the human brain. These devices offer many advantages over traditional cryogenic systems. They can be worn and made tolerant to levels of subject movement far exceeding traditional neuroimaging constraints. This opens up new possibilities in clinical and developmental neuroscience. As these sensors are small, they also allow us to build custom arrays designed to measure optimally from different structures. For example, we have begun to look for interactions between the spinal cord and cortex. The sensors do have several fundamental constraints however- they only operate within a narrow range around zero magnetic field and are very sensitive to external magnetic noise. I will outline some recent work and offer some perspectives on the exciting challenges ahead.

Just as cultural knowledge is wrapped in language shaped by historical practice, so biological facts are wrapped in physiology shaped by evolutionary forces. And just as unwrapping language to release the knowledge it contains requires deep generative models of text, so unwrapping physiology to reveal biological principles requires deep generative models of the body. The biological task is harder, for physiology---the grammar of biology---is hidden from view, glimpsed from the narrow, distorting apertures of imperfect medical instruments, and distributed across multiple, interacting scales of organisation. Though the nervous system is undoubtedly the most complex, the structuring pressure on its organisation is arguably highest, and its accessibility to generative models of the right expressivity therefore greatest. Here I examine the challenges and potential rewards of developing multi-modal, >3D deep generative models of the brain, drawing on analyses involving >10^6 individual brain volume images across >10^5 distinct patients powered by >5 petaFLOPS of compute, in the context of an array of representational, predictive, and prescriptive tasks.simulations of Bayesian belief updating in the brain and relate them to predictive processing and sentient behaviour.