CNS 2020 Virtual Meeting | Symposium Sessions

 

#

TITLE

1 STUDYING THE MIND BY MANIPULATING BRAIN NETWORKS
2 FINANCES AND FEELINGS: THE AFFECTIVE NEUROSCIENCE OF SES
3 PRESSING THE PLAY BUTTON: SEQUENTIAL NEURAL REPLAY OF HUMAN MEMORIES
4 FROM WIKIPEDIA SEARCHES TO SINGLE CELL RECORDING: UNCOVERING THE MECHANISMS OF INFORMATION-SEEKING
5 DEVELOPMENT AND PLASTICITY OF HIGH-LEVEL VISION AND COGNITION
6 MOVING FROM A DEFICIT-ORIENTED TO A PREVENTATIVE MODEL IN EDUCATION: EXAMINING NEURAL CORRELATES FOR READING DEVELOPMENT
7 INTEGRATING THEORY AND DATA: USING COMPUTATIONAL MODELS TO UNDERSTAND NEUROIMAGING DATA
8 THE MEETING OF PERCEPTION AND MEMORY IN THE BRAIN
9 CORTICAL GRADIENTS AND THEIR ROLE IN COGNITION
10 SPECIFICS AND GENERALITIES: BEYOND THE SEMANTIC-EPISODIC DISTINCTION
11 DEEP DATA: THE CONTRIBUTION OF CASE STUDIES AND SPECIAL POPULATIONS IN THE ERA OF BIG DATA
12 WHAT DETERMINES CATEGORY SELECTIVITY IN THE CORTEX?

 

SYMPOSIUM 1: STUDYING THE MIND BY MANIPULATING BRAIN NETWORKS

Chair: Joel Voss, Northwestern University

Speakers: Gesa Hartwigsen, Thorsten Kahnt, Michael Fox, Joel Voss

Cognition and emotion rely on large-scale distributed brain networks. However, there is little consensus on how these networks are organized, how their constituent regions interact to achieve function, and other key mechanistic questions. The dominant method for addressing these issues is to measure the natural relationships of activity among network regions while they are engaged by relevant processing demands. A complementary strategy is to utilize more direct functional probes by examining changes in cognition and emotion that result from network perturbations. Speakers in this symposium will discuss recent advances in this area using a network-based framework for noninvasive brain stimulation. They will describe new insights from experiments that use stimulation to probe the network basis of language, decision making, mood, and memory. Across all of these areas, the network-based stimulation framework has yielded strikingly specific, predictable, and reliable influences on targeted networks and their associated functions. Further, there are clear applications of these findings for the treatment of neurologic and psychiatric symptoms that result from network damage and dysfunction. By bringing together researchers using this approach to investigate different functional domains, we hope to foster discovery of general principles governing network control by brain stimulation and of general mechanisms by which brain networks accomplish cognition and emotion.

TALK 1: NEUROSTIMULATION FOR FLEXIBLE LANGUAGE-NETWORK REDISTRIBUTION IN HEALTHY AND LESIONED BRAINS

Gesa Hartwigsen, Max Planck Institute for Human Cognitive and Brain Sciences

Language is sustained by large-scale networks in the human brain. Brain lesions such as stroke often severely affect language function and network dynamics. However, the adaptive potential of the brain to compensate for lesions is poorly understood. In this talk, I will present novel evidence on the potential for short-term reorganization in the healthy and lesioned language network. First, I will show that ?virtual lesions? induced by neurostimulation to key language areas in the healthy brain increase the functional contribution of neighbouring language areas and domain-general control regions. Secondly, I will present a new study emphasizing the adaptive role of homologous right-hemispheric areas in the lesioned language network. As a main finding, perturbation of the left posterior prefrontal cortex in patients with lesions in the left temporo-parietal cortex selectively delayed phonological decisions and decreased task-related activity. The individual response delay was correlated with the upregulation of the lesion homologue, reflecting compensation for the disruption. Moreover, stronger individual tract integrity of the right superior longitudinal fascicle was associated with lesser impairment. These results provide evidence for functional and structural underpinnings of plasticity in the lesioned language network, and a compensatory role of the right hemisphere. I will integrate these data into a model for flexible redistribution in the language network, arguing that compensation for brain lesions may occur both within process-specific language networks and across networks for different cognitive functions. Finally, I will illustrate how this framework can be used to advance stimulation-based treatment of language disorders.

TALK 2: NETWORK STIMULATION TO TEST THE HUMAN ORBITOFRONTAL CORTEX ROLE IN INTERFERENCE-BASED DECISION MAKING

Thorsten Kahnt, Northwestern University Feinberg School of Medicine

Research across species has shown that the orbitofrontal cortex (OFC) is important for decision making. However, it is less clear what specific computations are carried out in this region that make it so important for this function. Recent work from our lab and others has shown that OFC activity is correlated with expectations about specific outcomes. Here we present evidence that these specific expectations are required for decisions that are based on inferred or simulated outcomes, as opposed to behavior that can be based on direct experience alone. Because of its anatomical location, the OFC is not directly accessible to transcranial magnetic stimulation (TMS). However, previous work suggests that TMS affects brain activity not only locally at the stimulation site but also in areas that are functionally connected to the stimulated region. In our experiments, we apply continuous theta burst stimulation (cTBS) to stimulation sites in lateral PFC that are individually selected to be maximally functionally connected to the OFC. We show that such OFC-targeted cTBS selectively disrupts choices that require subjects to infer outcomes, without affecting choices that can be based on direct experiences alone. These behavioral deficits are related to cTBS-induced decreases in the functional connectivity between the OFC and its cortical network. These findings demonstrate the feasibility of indirectly targeting human OFC using TMS and suggest that the OFC contributes to decision making by representing a cognitive map of the task environment that can be used to simulate outcomes when direct experience is missing.

TALK 3:  USING THE HUMAN BRAIN CONNECTOME TO IDENTIFY BRAIN CIRCUIT TARGETS FOR DEPRESSION SYMPTOMS

Michael Fox, Harvard Medical School

Therapies that directly target brain circuits have the potential to treat medication-refractory psychiatric symptoms such as depression. However, antidepressant response to surgical lesions, deep brain stimulation (DBS), and transcranial magnetic stimulation (TMS) has been highly variable across different patients. This variability has resulted in modest overall effect sizes and failed clinical trials. However, variability also provides an opportunity to identify optimal therapeutic targets for specific symptoms, symptom clusters, and disorders. Specifically, incidental variability in the precise location of each patient?s treatment site can be mapped to underlying brain circuits using a wiring diagram of the human brain termed the human connectome. I will describe how this approach can be applied to brain lesions, DBS sites, and TMS sites to identify better brain circuit targets for depression. Recent findings suggest that individualized targeting can be used to tease apart distinct circuits that affect distinct symptom clusters when treated with TMS. In one study, TMS of one target was associated with improvement in dysphoric symptoms, such as sadness and anhedonia, whereas another target was associated with improvement in anxiety and somatic symptoms. These findings indicate that circuit-based approaches for influencing brain function can yield remarkably specific outcomes even for complex mood disorders. This supports the use of such methods for personalized neuromodulatory therapies as well as for investigations into the brain basis of mood and emotion.

TALK 4: STIMULATING THE HIPPOCAMPAL NETWORK TO TEST EPISODIC MEMORY MECHANISMS

Joel Voss, Northwestern University Feinberg School of Medicine

Episodic memory depends on the hippocampus and its coordination with a distributed network of interconnected structures. Recent findings indicate that this hippocampal network can be modulated using network-targeted transcranial magnetic stimulation. This offers the powerful opportunity to directly test hypothesized functional properties of the hippocampal network by measuring the memory changes that occur in response to stimulation. I will describe the progress that has been made in this area to date. Increases in fMRI activity correlation due to stimulation predict corresponding increases in episodic memory ability, indicating that successful performance relies on the interregional coordination of hippocampal network activity. Furthermore, distinct hypothesized posterior-medial and anterior-temporal functional network components are differentially modulated by stimulation, thereby demonstrating their functional independence. The prominent hypothesis that hippocampal network coordination for memory occurs via the synchronization of activity in the theta-frequency band has also been supported by network-targeted stimulation, which more robustly influences hippocampal network activity and memory when delivered using theta patterns versus non-theta patterns. Finally, I will describe our recent work using theta-patterned stimulation during simultaneous fMRI scanning to measure the immediate impact of stimulation on the hippocampus and its role in the network-wide effects of stimulation. Collectively, these findings suggest that it is possible to cause highly specific changes in episodic memory by appropriately targeting portions of the hippocampal network with noninvasive stimulation, yielding new insights regarding brain mechanisms of memory.

SYMPOSIUM 2: FINANCES AND FEELINGS: THE AFFECTIVE NEUROSCIENCE OF SES

Chair: Martha Farah, University of Pennsylvania

Speakers: Moriah Thomason, Joan Luby, Robin Nusslock, Pilyoung Kim

Depression is twice as common at the lowest income levels than at the highest.  Stands to reason, you might say; no need for neuroscience to understand why. But people who are poor during childhood and become more affluent as adults continue to be at elevated risk. It appears that early life socioeconomic status (SES) influences brain development in ways that have lifelong effects on our emotional responses to positive and negative events and to social situations, as well as our ability to regulate our emotions.  This impacts rates of psychopathology, especially affective disorders, and also levels of well-being within the healthy population.  It does so by mechanisms that begin prenatally and operate in postnatal life under the influence of factors such as stress and parenting practices.  The neural differences associated with SES are even associated with parents' feelings and behaviors toward the next generation, their own children.  In this symposium we will hear from four leaders in the affective neuroscience of SES, whose work spans brain activity in prenatal life, early childhood, later childhood to adulthood, and parent-child processes.  While covering different periods of life, the presentations will be unified by a number of common themes: psychosocial and physiological stress, limbic and prefrontal systems and networks, and positive feedback loops operating within individuals and across generations.  A final discussion will solicit thoughts from the speakers and the audience about ways of breaking cycles of disadvantage and despair and promoting well-being for all.

TALK 1: NEURAL CORRELATES OF POVERTY OBSERVED IN THE HUMAN FETAL BRAIN: IMPLICATIONS FOR POSTNATAL WELLBEING

Moriah Thomason, NYU/Langone Medical Center

Prenatal poverty is associated with increased risk for preterm birth, intrauterine growth restriction, neonatal/infant death, and also cognitive and affective regulation in childhood. Here, we address whether prenatal poverty relates to formation of fetal brain circuitry that will support emotion processing in the future. An important target for research is identification of the earliest emergence of socioeconomic status (SES)-related differences in the human brain and their implications for postnatal behavior and wellbeing. We obtained functional MRI data in more than 100 normally-developing human fetuses from primarily low SES families and tested whether amygdala whole brain connectivity relates to familial SES. We observed reduced amygdala connectivity to prefrontal cortex, posterior insula, and cerebellum, and increased local connectivity in fetuses of families with the lowest SES. Some of these differences predict childhood abilities, including self-regulation. Future research confirming that system-level brain organization in utero is altered in fetuses of low SES mothers could motivate new lines of research into physiological processes and chemical and/or epigenetic pathways by which maternal resources program the human central nervous system in the womb.

TALK 2: SES, EARLY EXPERIENCE AND BRAIN DEVELOPMENT: INFORMING A SCIENCE OF NEURODEVELOPMENTAL ENHANCEMENT

Joan Luby, Washington University

There is increasing evidence for the effects of early experiences of poverty, adversity and nurturance on childhood brain development, a problem we have studied at the Early Emotional Development Lab at Washington University.  These effects are known to be enhanced during sensitive periods when neural architecture is maximally informed by the environment for adaptation to future expected experiences.  Evidence for sensitive periods for cognitive enhancement prior to the age of 2 have been inferred in experimental studies in humans and we have shown sensitive periods for maternal support on hippocampal development in longitudinal studies.  Our data and others? find regional specificity of experiences of both adversity and nurturance on brain regions associated with children?s affective functioning and the timing of exposures show that there is both timing and regional specificity to these effects.  These findings along with others from the extant literature, as well as the need for new targeted investigations in developing humans and animal models, will be considered to inform a new science of early childhood neurodevelopmental enhancement.  Such a model could be feasibly used in primary care settings to optimize neurodevelopment.  This could be done by providing clear guidelines for when it is most important to protect developing children from certain forms of adversity and when it is most important for them to experience enhancement through nurturance and stimulation.  The resulting neurodevelopmental enhancement model would be a feasible public health application of findings on adversity, brain development and affective functioning.

TALK 3: EXECUTIVE AND EMOTION REGULATION NETWORKS ASSOCIATED WITH RESILIENCE TO POVERTY AND EARLY ADVERSITY

Robin Nusslock, Northwestern University

Individuals exposed to early-life adversity, including being raised in a family of low socioeconomic status, are vulnerable to emotional and physical problems across the lifespan. However, not everyone exposed to adversity is affected, which raises an important question: what enables some to remain healthy whereas others deteriorate? We first test the hypothesis that heightened activity in the brain?s central executive network (CEN), which regulates emotions and limbic reactivity, might reflect a neurobiological marker of resilience. We enrolled 218 urban youth and characterized their exposure to neighborhood violence. Cardiometabolic health and resting state functional connectivity (rsFC) were assessed. As expected, higher neighborhood violence was associated with greater cardiometabolic problems, but only among individuals who displayed lower rsFC in the CEN. We next examined whether receiving supportive parenting during adolescence helps strengthen connectivity in the CEN and an emotion regulation network (ERN) while growing up in poverty. In a sample of African Americans (N = 119) living in the rural South, poverty status and receipt of supportive parenting were assessed during adolescence and rsFC was assessed using fMRI at age 25. As predicted, more years spent living in poverty presaged less CEN and ERN rsFC among young adults who received low levels of supportive parenting, but not among those who received high levels of such parenting. Collectively this suggests that heightened central executive and emotion regulation tendencies may help protect individuals from the consequences of early-life adversity and that supportive parenting can help foster these tendencies in the face of such adversity.

TALK 4: SOCIOECONOMIC DISADVANTAGE AND THE NEUROSCIENCE OF MOTHER-INFANT ATTACHMENT

Pilyoung Kim, University of Denver

Socioeconomic disadvantage such as poverty can increase distress levels, which may make low-income mothers more vulnerable to difficulties in the transition to parenthood. Cumulative risk ? exposure to multiple stressors ? is one of the main environmental mechanisms by which socioeconomic disadvantage is associated with negative brain and psychological functioning. Cumulative risk has also been linked to negative postpartum outcomes including harsh parenting, which can further influence how socioeconomic disadvantage may be transmitted to the next generation. Thus, the goal of the current study was to investigate whether cumulative risk may disrupt the neural and behavioral development of mother-infant attachment. We examined the association of cumulative risk with the brain response to infant cries and maternal behaviors, in a sociodemographically diverse sample (42% low income) of first-time mothers (N=53). Cumulative risk across socioeconomic (low income, financial stress, food insecurity), physical environment (substandard housing, noise, crowding), and psychosocial (marital dissatisfaction, violence, troubles with social services) domains was associated with reduced brain response to infant cries compared to white noise in several regions including the right insula/inferior frontal gyrus and superior temporal gyrus. Reduced activation in these regions was further associated with lower maternal sensitivity observed during a mother-infant interaction recorded at a home visit. The findings demonstrate that exposure to multiple stressors that are associated with socioeconomic disadvantage may be associated with reduced brain response to an infant?s cry in brain regions that are important for emotional and social information processing, and associated with increased difficulties in developing positive mother-infant relationships.

SYMPOSIUM 3: PRESSING THE PLAY BUTTON: SEQUENTIAL NEURAL REPLAY OF HUMAN MEMORIES

Chair: Eitan Schechtman,

Speakers: Kareem Zaghloul, Marit Petzka, Yunzhe Liu, Leonardo G Cohen

Offline reactivation of memory-related neural patterns is thought to contribute to long-term memory evolution. In rodents, sequential reactivation of neuronal ensembles - conventionally termed 'replay' - has been primarily observed in hippocampal place cells and has been linked to memory consolidation and the planning of future actions. The same replay phenomenon has not yet been observed in humans. Identifying parallel physiological phenomena in humans would be an important advance for understanding neurocognitive mechanisms of memory. Progress towards that goal has recently been achieved using different paradigms and methods, including EEG, MEG, fMRI, and ECoG. This symposium will discuss some of these novel results, all emerging within the past year, that expose several underlying themes, including temporal compression of neural sequences and links to subsequent performance. These demonstrations of replay-like mechanisms in the human brain, taken together, reveal various similarities and differences between human and non-human reactivation. Exploring these avenues could pave the way toward deeper insights into the role of reactivation of sequential neural patterns in memory consolidation, planning, and decision making.

TALK 1: NEURAL MECHANISMS OF HUMAN EPISODIC MEMORY FORMATION ACROSS SPATIAL SCALES

Kareem Zaghloul, National Institute of Neurological Disorders and Stroke, NIH, Bethesda, MD

Episodic memory relies upon our ability to retrieve the memory of individual events that we have experienced at a particular time and place.  The hippocampus and structures in the medial temporal lobe (MTL) play a critical role in this process by representing relations between memories and the spatiotemporal context within which they occur. A parallel line of research, however, has demonstrated that successful episodic memory retrieval involves recovering neural representations that were present in the cortex when memories were first experienced. This has led to the hypothesis that the hippocampus and MTL may promote episodic memory retrieval through a dialogue with the cortex that facilitates the ability to recover these neural representations.  Here we explore this hypothesis by examining neural signals directly captured from the human brain across multiple spatial scales as participants perform a verbal episodic memory task.  We show that patterns of neural activity at both the larger mesoscopic scale of intracranial EEG (iEEG) electrodes and at the smaller microscale of single units in the temporal lobe cortex are reinstated when memories are successfully retrieved.  Moreover, we show that that such reinstatement of cortical activity is locked to the occurrence of coordinated oscillatory activity between the temporal lobe cortex and structures in the MTL.  Together, these data suggest a mechanistic framework through which neural activity in the MTL can promote memory retrieval by initiating the replay of patterns of neural activity in the cortex.

TALK 2: FORWARD REACTIVATION OF SEQUENTIAL MEMORY TRACES DURING SLEEP

Marit Petzka, School of Psychology and Centre for Human Brain Health, University of Birmingham, UK

Our ability to remember past events relies on the re-emergence of learning patterns during sleep. In humans, previous studies focused on simple paired-associate learning. However, episodic memories tend to contain multiple, sequentially experienced elements. Indeed, animal studies have provided evidence for reactivation of learning sequences (?replay?) and suggest that sequential reactivation occurs in a compressed and forward manner. To date, little is still known about the temporal dynamics of sequential memory reactivation during sleep in humans.

Here, we applied targeted memory reactivation (TMR) to cue previously learned sequences of object-face-scene triplets during a post-learning nap using high-density electroencephalography (EEG). Behavioural results confirm that encoding took place sequentially, as the conditional probability to correctly retrieve a face without remembering the following scene was higher than correctly retrieving a scene without remembering the preceding face (p < .001). Importantly, memory performance for sequences that were cued during the nap was higher compared to sequences not cued (p = .014), establishing that TMR for sequences was successful. To capture sequential reactivation, a multiclass LDA classifier was trained on an independent localizer dataset (visual perception of objects, faces and scenes) during wakefulness and validated on sleep data in response to target cues. During sleep, classifier evidence for the emergence of face representations peaked after 800ms cue onset. Most interestingly, classifier evidence for scene representations peaked 600ms after the evidence for face representations, reflecting the order in which the sequences were encoded. Together, our findings reveal forward replay of previously learned memory traces during sleep.

TALK 3: NEURAL REPLAY IN MODEL-BASED LEARNING

Yunzhe Liu,Wellcome Trust Centre for Neuroimaging, University College London, UK

Humans exhibit remarkably flexible behaviour. Such flexibility is thought possible because the brain builds internal models of the world (i.e., cognitive map). How the brain represents, updates and use the world model to support flexible behaviour remains a central question in neuroscience. I will show evidence suggesting neural sequential replay plays a crucial role in representing, updating and generalizing the world model in humans. By building pattern classifiers of MEG sensor activity for each visual stimulus we detected their sequential reactivation during rest. These sequences recapitulated known features of neural replay in rodents and reflected correctly re-assembled orderings, rather than experienced trajectories.  The forward replay of a correctly re-assembled sequence transitioned to that of reverse replay when a sequence was rewarded. We provide further evidence that neural pre-play is a manifestation of abstract structure knowledge. The representation of neural replay is factorized so that a sensory code of object representations was preceded 50 ms by structural code(i.e., sequence position and sequence identity) to allow for fast structural generalization to novel situations. When such a replay mechanism goes wary, it explains key cognitive deficits in psychiatric disorder, like Schizophrenia. I will also show evidence that neural sequential replay supports episodic memory retrieval, model-based planning and decision-making at the trial-by-trial basis. The direction of sequential replay can be flexibly adjusted to suit the current task goal. Together, the evidence suggests a crucial role of sequential replay underlying human cognition.

TALK 4: REPLAY OF HUMAN PRACTICE PREDICTS EARLY SKILL LEARNING

Leonardo G Cohen, National Institute of Neurological Disorders and Stroke, NIH, Bethesda, MD

Neural replay, spatiotemporal brain activity associated with task performance during rest, has been reported during sleep and linked to overnight memory consolidation.  Wakeful replay contributes to memory formation in rodents (1) but its role in relation to skill formation or even presence in the context of human motor practice is not known. Here, we analyzed data collected in 31 subjects  (2) who learned a sequence of keypresses with the non-dominant left hand. Training consisted of 36 alternating practice and rest periods (10 seconds each) lasting a total of 12 minutes. MEG recordings were obtained to assess resting-state and task-induced brain activity dynamics. Support vector machine (SVM) classifiers were constructed for individual key-press events during practice and then used to identify replay of sequence-related MEG dynamics during wakeful rest periods (3). Replay was assessed over sixteen different timescales (25-2500ms) pertaining to biologically relevant replay durations (4). Replay events were observed as early as the first rest period, remained present over the 36 rest periods and for at least 5-minutes after the end of practice. Optimal replay duration was 50-100ms, with a majority of subjects showing peak replay rates at 75ms durations. Replay of the trained sequence during rest periods prior to performance asymptote predicted rapid offline consolidation of the new skill. Source analysis identified a distributed medial temporal and sensorimotor network underlying wakeful neural replay. We conclude that motor practice elicits sustained neural replay during wakeful rest intervals that predict early skill learning.

SYMPOSIUM 4: FROM WIKIPEDIA SEARCHES TO SINGLE CELL RECORDING: UNCOVERING THE MECHANISMS OF INFORMATION-SEEKING

Chair: Tali Sharot, University College London

Speakers: Eric Schulz, Danielle Basset, Ethan Bromberg-Martin, Irene Cogliati Dezza

People spend a substantial amount of time seeking out information (e.g., asking questions, reading, internet browsing). The human pursuit of knowledge drives intellectual development, is integral to social interactions, crucial for learning and decision-making. An important research challenge is understanding how people decide what they want to know. As massive amounts of information are becoming available to people this challenge is more pertinent today than ever. Despite the central role of information-seeking to human behavior research on information-seeking has been surprisingly limited in comparison to other domains of human cognition and behavior, but has been experiencing revitalization in recent years. This symposium will showcase some of the most recent discoveries in this domain. The presented studies aim to uncover the computational rules and neural mechanisms that support information-seeking as well as individual differences in information-seeking strategies and the relationship between these strategies and mental health. The symposium brings together speakers from different disciplines including psychology, neuroscience and computer science, to provide new insight into information-seeking and its neural basis. Eric Schulz will characterize human strategies for information-seeking in complex environments; Danielle Bassett will present a study looking at people's Wikipedia searches that reveals how people create knowledge networks; Ethan Bromberg-Martin will present evidence for a neural network mechanism of information-seeking; Irene Cogliati Dezza will present evidence for information-seeking alterations in psychopathology; Tali Sharot will conclude by presenting a theory of the motives that drive information-seeking and describe potential applications of this work for assessing mental health.

TALK 1: USING STRUCTURE TO EXPLORE EFFICIENTLY

Eric Schulz, Harvard University

Many types of intelligent behavior can be framed as a search problem, where an individual must explore a vast set of possible actions, while carefully balancing the exploration-exploitation dilemma. Under finite search horizons, optimal solutions are normally unobtainable, yet humans and other animals regularly manage to solve these problems gracefully. How do they accomplish this? We propose that two simple principles can explain this: generalization over features and uncertainty-guided exploration. Together these form a model that learns from past observations to generalize to similar options and seeks out uncertainty eagerly in order to gain more information about the search space. This model can be used to predict participant's search behavior in a complex multi-armed bandit task. Its parameter estimates can also be used to gain meaningful insights into developmental differences in generalization and directed exploration. Furthermore, we can use this model to describe customers' purchasing decisions in large-scale data set (1.6 million orders) online food delivery website. Finally, I will end by describing ongoing work that puts this model to a test in a multi-armed bandit task with rats, in which we find similar principles influencing animals' motor variability.

TALK 2: HUNTERS, BUSYBODIES, AND THE KNOWLEDGE NETWORK BUILDING ASSOCIATED WITH CURIOSITY

Danielle Basset, University of Pennsylvania

The information gained when practicing curiosity promotes well-being over extended timescales. The open-ended and internally driven nature of curiosity, however, makes characterizing the diverse styles of information seeking that accompany it a daunting endeavor. A recently developed historicophilosophical taxonomy of curious practice distinguishes between the collection of disparate, loosely connected pieces of information and the seeking of related, tightly connected pieces of information. With this taxonomy, we use a novel knowledge network building framework of curiosity to capture styles of curious information seeking in 149 participants as they explore Wikipedia for over 5 hours spanning 21 days. We create knowledge networks in which nodes consist of distinct concepts (unique Wikipedia pages) and edges represent the similarity between the content of Wikipedia pages. We quantify the tightness of each participants' knowledge networks using graph theoretical indices and use a generative model of network growth to explore mechanisms underlying the observed information seeking. We find that participants create knowledge networks with small-world and modular structure. Deprivation sensitivity, the tendency to seek information that eliminates knowledge gaps, is associated with the creation of relatively tight networks and a relatively greater tendency to return to previously-visited concepts. We further show that there is substantial within-person variability in knowledge network building over time and that building looser networks than usual is linked with higher than usual sensation seeking. With this framework in hand, future research can quantify the information collected during curious practice and examine its association with well-being.

TALK 3: A NEURAL NETWORK FOR INFORMATION SEEKING

Ethan Bromberg-Martin, Washington University

Do you want to know what your future holds? Humans and animals often express a strong desire to seek information about the properties of uncertain future rewards, even when there is no way for them to use this information to influence the outcome. However, little is known about the neuronal mechanisms that sustain information seeking. In particular, how does the brain anticipate opportunities to gain information and generate the motivation to pursue them? I will present evidence that these cognitive and motivational processes are served by a novel population of information-anticipatory neurons in an anatomically connected network including the anterior cingulate cortex, dorsal striatum and ventral pallidum. We trained monkeys to perform tasks which yield probabilistic juice rewards and which offer opportunities to gaze at visual cues: either informative cues that perfectly predict future reward outcomes or non-informative cues that do not predict future outcomes. We found that a substantial proportion of neurons in the network have strong and selective information-anticipatory activity: ramping activity that anticipates the moment the animal expects to gain information to resolve uncertainty about future rewards. Moment-to-moment fluctuations in their activity predict the animal?s future information-anticipatory gaze shifts, and pharmacological perturbation of the basal ganglia nuclei that contain these neurons causally interferes with information seeking. Our results demonstrate a cortico-basal ganglia pathway for seeking information about future events, in parallel with the well-known pathways for seeking primary rewards like food and water. I will discuss the implications for theories of motivation, learning, and decision making.

TALK 4: INFORMATION-SEEKING IMPAIRMENTS IN BEHAVIORAL ADDICTION AS A NOVELTY FAILURE

Irene Cogliati Dezza, University College London

Information-seeking is an important aspect of human cognition that supports healthy functioning of decision-making and goal-directed processing. Despite its adaptive and ubiquitous role in human daily activities, we have rather limited understanding on the mechanisms subtending information-seeking in both healthy individuals and in psychopathologies. Here, we sought to formalize the computational basis of healthy information-seeking, as well as how those components could be compromised in behavioral addiction. We investigate and model human behavior using a novel variant of a classical decision-making task and a novel computational model. This approach allows us to dissociate the relative contributions of information and reward on decision-making, as well as the influence of novelty and general uncertainty. Overall, we found that healthy subjects were motivated by both information gain and reward gain in their choices. In contrast, problem gamblers showed a decreased reliance on information gain as a consequence of a failure in representing novelty. This finding both sheds light on the computational mechanisms underlying healthy human choice behavior, and how they go awry in an addictive population without the confound of illicit substance consumption. Methodologically, this work offers promising novel experimental and computational approaches to study the mechanisms underlying reward-based learning and decision-making in both healthy and pathological populations.

SYMPOSIUM 5: DEVELOPMENT AND PLASTICITY OF HIGH-LEVEL VISION AND COGNITION

Chair: Zeynep Saygin, The Ohio State University

Speakers: Daniel Dilks, Rhodri Cusack, Zeynep Saygin, Marina Bedny

What determines the development and plasticity of cortical specialization? Recent evidence points to connectivity as the general mechanism that underlies this specialization. Daniel Dilks introduces evidence of adult-like functional connectivity of face and place networks in infants. Face networks show biases in connectivity with foveal primary visual cortex (V1) while place networks show connectivity with peripheral V1. Rhodri Cusack also finds evidence of adult-like structural connectivity of face and place networks in infants but further shows that tool networks are not adultlike and undergo prolonged maturation until 9 months of age. Zeynep Saygin shows that another highly experience-dependent visual region, the visual word form area (VWFA), already shows privileged functional connectivity with language areas at birth. These three studies suggest a connectivity-based mechanism to earmark functional specialization as well as a role for experience in further shaping connectivity and specialization. Marina Bedny directly explores the plasticity that occurs with experience and finds that congenitally blind individuals have distinct regions within 'visual' cortex that are selective to higher-cognitive domains. Remarkably, these regions show preferential functional connectivity with prefrontal areas that have analogous task-based responses, suggesting that connectivity constrains functional specialization even in cases of large-scale reorganization due to atypical experience. Together, these presentations suggest that early-developing or innate connectivity provides a scaffold for functional specialization of cortex, and constrains how experience may shape this functional specialization.

TALK 1: CONNECTIVITY AT THE ORIGINS OF DOMAIN SPECIFICITY IN THE CORTICAL FACE AND PLACE NETWORKS

Daniel D. Dilks, Emory University

It is well established that the adult brain contains a mosaic of domain-specific networks. But how do these domain-specific networks develop? Here we tested the hypothesis that the brain comes prewired with connections that guide the development of particular domain-specific networks. Using resting-state fMRI in the youngest sample of newborn humans tested to date, we found that cortical networks that will later develop strong face selectivity (including the ?proto? occipital face area and fusiform face area) and scene selectivity (including the ?proto? parahippocampal place area and retrosplenial complex) by adulthood, already show adult-like patterns of functional connectivity in as little as 27 days of age. We further asked why these networks always develop selectivity for faces and scenes, respectively, and not for other domains (e.g., scene selectivity in regions that are typically face selective, and vice versa), and found that the proto face and scene networks show differential functional connectivity to primary visual cortex (V1), with face regions biased toward foveal V1, and scene regions biased toward peripheral V1. Given that faces are almost always experienced at the fovea, while scenes always extend across the entire periphery, these distinct inputs may place powerful constraints on the function that each system will ultimately take on. Taken together, these results strongly support the hypothesis that innate connectivity shapes the development of the cortical face and scene processing networks, providing novel evidence for what may be a general mechanism of the origins of domain-specific networks.

TALK 2: CATEGORY-SELECTIVE VISUAL REGIONS HAVE DISTINCTIVE SIGNATURES OF STRUCTURAL CONNECTIVITY IN INFANTS

Rhodri Cusack, Trinity College Dublin

By four months, infants can form categories of similar-looking objects, but it is unclear when they begin to make the rich cross-modal, motoric and affective associations that are characteristic of adult visual categories. These associations are thought to be encoded by long-range brain connectivity and are reflected in the distinctive signature of connectivity of each category-selective region in the ventral visual stream. Category-selective ventral visual regions are already functioning in young infants, but their long-range connectivity has not been investigated. Therefore, we used MRI diffusion tractography to characterize the connectivity of face, place and tool regions in 1-9 month infants. Using a linear discriminant classifier, we found that the face and place regions had adult-like connectivity throughout infancy, but the tool-network underwent significant maturation until 9 months. This suggests that the face and place regions have long-range connectivity that is either innately specified or learned in the first months of infancy, while the more protracted development of the tool network is consistent with it developing as motor function develops, and infants learn to reach. This emerging long-range connectivity could reflect young infants developing category-specific rich associations.

TALK 3: SELECTIVITY DRIVEN BY CONNECTIVITY: INNATE CONNECTIVITY PATTERNS OF THE VISUAL WORD FORM AREA

Zeynep M. Saygin, The Ohio State University

The human brain is a patchwork of different functionally specialized areas. What determines this functional organization of cortex? One hypothesis is that innate connectivity patterns shape functional organization by setting up a scaffold upon which functional specialization can later take place. We tested this hypothesis here by asking whether the visual word form area (VWFA), an experience-driven region that only becomes selective to visual words after gaining literacy, was already connected to protolanguage networks in neonates scanned within one week of birth. We found that neonates showed adult-like functional connectivity, and observed that i) the VWFA connected more strongly with frontal and temporal language regions than regions adjacent to these language regions (e.g., frontal attentional demand, temporal auditory regions), and ii) language regions connected more strongly with the putative VWFA than other adjacent ventral visual regions that also show foveal bias (e.g. fusiform face area, FFA). Object regions showed similar connectivity with language areas as the VWFA but not with face areas in neonates, arguing against prior hypotheses that the region that becomes the VWFA starts out with a selectivity for faces. These data suggest that the location of the VWFA is earmarked at birth due to its connectivity with the language network, providing novel evidence that innate connectivity instructs the later refinement of cortex.

TALK 4: CONGENITAL BLINDNESS REPURPOSES VISUAL CORTICES FOR HIGHER-COGNITION AND CHANGES THEIR CONNECTIVITY

Marina Bedny, Johns Hopkins University

A growing body of evidence suggests that intrinsic connectivity patterns constrain the functional specialization of cortex. Are these constraints compatible with large-scale functional change as a result of experience? We tested the hypothesis that in blindness different parts of ?visual? cortex are incorporated into distinct higher-cognitive networks using task-based and resting-state data. Congenitally blind (N=23), adult-onset blind (N=10) and blindfolded sighted controls (N=18) took part in three higher-cognitive tasks that activate different fronto-parietal networks. Each task had multiple difficulty levels: 1) auditory sentence processing (grammatically complex vs. simpler sentences) 2) solving math equations of varying difficulty and 3) non-verbal executive go/no-go task (frequent go, infrequent go, no-go). In congenitally blind individuals, different networks within ?visual? cortex preferentially responded to linguistic, numerical and non-verbal go/no-go tasks and showed task-specific sensitivity to cognitive load. Responses were larger in congenitally blind than in sighted and adult-onset blind participants. Congenital blindness was also associated with functional connectivity changes: all occipital networks tested showed reduced resting-state correlations with sensorimotor and auditory areas and enhanced correlations with prefrontal cortices. Furthermore, each occipital network showed preferential enhancements with prefrontal areas that have analogous task-based responses (i.e. language-responsive ?visual? areas showed preferential correlations with language-responsive prefrontal areas). Blindness enables ?visual? cortices to develop selective higher-cognitive responses and changes resting-state connectivity. These findings suggest that intrinsic connectivity constraints are compatible with dramatic functional change as a result of experience.

SYMPOSIUM 6: MOVING FROM A DEFICIT-ORIENTED TO A PREVENTIVE MODEL IN EDUCATION: EXAMINING NEURAL CORRELATES FOR READING DEVELOPMENT

Chair: Tzipi Horowitz-Kraus, Cincinnati Children's Hospital

Speakers: Tzipi Kraus, Nadine Gaab, Heikki Lyytinen, Michael Skeide and Jolijn Vanderauwera

Reading is a cultural invention and needs to be explicitly taught. Learning to read leads to high-level plasticity in a number of neural circuits, including vision, language and executive functions which makes it a great model to study experience-dependent plasticity in the developing brain. However, 3-10% of children struggle with reading acquisition, which continues into adulthood and poses future academic, socio-economic and mental health challenges in life. The etiology of reading difficulty is thought to lie within the dynamic interplay of genetic risk factors and environmental as well as cutural influences.  In this symposium we will discuss the developmental trajectories and corresponding structural and functional neural circuits of learning to read starting at the pre-reading stage. Functional and structural MRI, Diffusion tensor imaging and EEG data obtained from English, German, Finnish, Hebrew and Dutch speaking children will be presented to provide a wide overview of the various factors influencing typical and atypical reading development in children worldwide. Furthermore, we will provide an overview about genetic and environmental factors that can influence experience-dependent plasticity during the process of learning to read. This includes a discussion of familial risk and its role in a multi-risk model as well as the role of home literacy environment and screen exposure time over the time course of learning to read. Similarities and differences across languages and orthographies as well as between a variety of different neuroimaging modalities will be discussed.

TALK 1: NEUROBIOLOGICAL CORRELATES FOR ENVIRONMENTAL FACTORS CONTRIBUTING TO FUTURE READING ABILITIES

Tzipi Horowitz-Kraus, The Educational Neuroimaging Center, Faculty of Education in Sciences and Technology and Jolijn Vanderauwera, Université catholique de Louvain, Belgium, Harvard Medical School, USA; KU Leuven, Leuven, Belgium

Environment has a major contribution to children’s reading abilities. Home literacy environment and joint storytelling may be helpful for future reading abilities. On the other hand, screen exposure time may minimize the time children spend reading. Neuroimaging studies have demonstrated the involvement of executive functions, visual processing and language networks, all support future reading abilities, in young children listening to stories. However the same networks may be engaged during screen exposure as well. In a series of studies, we examined the neurobiological correlates for home literacy environment and of screen exposure, focusing on executive functions, language and visual processing in young children. The relationship between home literacy vs screen exposures with the activation and connectivity of neural circuits supporting these networks in preschoolers and school-age children was examined using functional MRI and EEG.  Results demonstrate the recruitment of visual processing and executive functions networks, as well as white matter tracts related to these abilities both crucial for reading, in preschoolers and school-age children during a resting-state and task conditions with increased screen time. Similar regions were positively correlated with increased home reading environment.  We conclude that screen exposure competes with neural circuits originally used for reading and narrative comprehension and therefore, exposure to screens should be monitored carefully. We also suggest that children exposed less to home literacy environment and to increased screen time may eventually have a reduced reading ability.

TALK 2: THE TYPICAL AND ATYPICAL READING BRAIN: HOW A NEUROBIOLOGICAL FRAMEWORK OF READING DEVELOPMENT CAN  INFORM EDUCATIONAL PRACTICE AND POLICY

Nadine Gaab, Harvard Medical School, Boston USA

Various developmental disorders are diagnosed in early childhood, but divergent trajectories of brain development may already be present in preschool, at birth or prenatally. Here we will present results from our longitudinal studies which investigate whether observed functional and structural brain differences associated with reading impairments and developmental dyslexia are already present in infants and preschoolers, how they develop over time, and which aspects of these functional and structural differences are prospectively associated with subsequent language and reading outcome. We will further introduce a multiple deficit model that illustrates reading impairment as an outcome of multiple risks and protective factors interacting within and across genetic, neural, cognitive, and environmental levels from infancy to adolescence. Additionally, we will place a special emphasis on new findings from our longitudinal studies that characterize neural protective and compensatory mechanisms in young children at a heightened risk but who subsequently develop typical language and reading skills. Understanding the early developmental trajectories of language and reading skills, behaviorally and in the brain, will allow for better understanding of the etiological basis of reading impairments and will help inform early screening, identification and remediation practices. Finally, current and potential implications of these findings for contemporary challenges in the field of developmental cognitive neuroscience as well as for education and clinical practice in general, are discussed

TALK 3: FUNCTIONAL AND STRUCTURAL SIGNATURES OF DYSLEXIA BEFORE AND AFTER LITERACY INSTRUCTION

Michael Skeide, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany

Developmental dyslexia, a severe deficit in literacy learning, is one of the most common neurodevelopmental learning disorders. Yet, it is not well understood whether existing accounts of dyslexia capture potential causes of the deficit or consequences of reduced reading experience. Here, we followed a cohort of 32 children from preliterate to school age using functional and structural magnetic resonance imaging techniques. Based on reading and spelling tests administered at the end of second grade, these children were classified as dyslexics (N=16, age range preliterate age: 5.1-6.4 years, age range school age: 8.0-9.1 years, 5 female, 11 male) and controls (N=16, age range preliterate age: 5.0-6.0 years, age range school age: 7.11-8.11 years, 7 female, 9 male). This longitudinal design allowed us to disentangle potential neural predispositions for developing dyslexia from differences in literacy experience. In our sample, dyslexia reveals itself already at a preliterate age from differences in auditory cortex gyrification (F(1,24)=9.64, p=0.0048; FWE-corrected; η2 =0.19) and downstream connectivity (resting state functional connectivity: F(1,24) = 14.73, P = 0.0009, η2 = 0.32; streamline density: F(1,24) = 15.16, P = 0.003, η2 = 0.39) within the speech processing system. Our results support the notion that dyslexia may be related to subtle early cortical formation defects altering auditory cortex folding and hampering speech processing.

TALK 4: PRECURSORS OF DIFFICULTIES ASSOCIATED WITH THE DEVELOPMENTAL STEPS TOWARDS FULL LITERACY

Heikki Lyytinen, University of Jyväskylä , Finland

The Jyväskylä Longitudinal study of Dyslexia has documented the developmental steps learners have to take to reach full literacy and the bottlenecks which may make it difficult. The first step prepares learner to acquire the basic reading skill. At least in a fully transparent writing environment consistent at grapheme-phoneme level it is easy to understand that a potential bottleneck is the difficulty one may face in differentiating the phonemes from each. To learn the basic reading skill, one had to learn to connect the letter/grapheme representing each of the phonemes. Thus e.g. n, m and l are acoustically so close that differentiation is difficult. Thus, it is not any surprise that it is auditory sensitivity which is needed. Children at familial risk for dyslexia resulting from a parent with dyslexia were observed to have about fifty-fifty likelihood of ending to face dyslexia. Mismatch negativity observed at age 3-5 days of life fails to show up among that half of infants who ended facing dyslexia 10 years later. The next step, learning to comprehend written material, observed on the basis of PISA measures at age 15 was shown up being related to the development of spoken language. Children at familial risk whose expressive and receptive language was late at the age of 2.5 years ended up facing severe problems in reading comprehension 12 years later.

 

SYMPOSIUM 7: INTEGRATING THEORY AND DATA: USING COMPUTATIONAL MODELS TO UNDERSTAND NEUROIMAGING DATA

Chair: Brandon Turner, The Ohio State University

Speakers: Birte Forstmann, Michael Frank, Alison Preston, Brandon Turner

Our understanding of cognition has been advanced by two traditionally non-overlapping and non-interacting groups. Mathematical psychologists rely on behavioral data to evaluate formal models of cognition, whereas cognitive neuroscientists rely on statistical models to understand patterns of neural activity, often without any attempt to make a connection to the mechanism supporting the computation. Both approaches suffer from critical limitations as a direct result of their focus on data at one level of analysis (cf. Marr, 1982), and these limitations have inspired researchers to attempt to combine both neural and behavioral measures in a cross-level integrative fashion. The importance of solving this problem has spawned several entirely new theoretical and statistical frameworks developed by both mathematical psychologists and cognitive neuroscientists. In this symposium, we will highlight a few of these efforts both at a methodological and application level.

TALK 1: CORTICOSTRIATAL COMPUTATIONS IN LEARNING AND DECISION MAKING

Michael Frank; Brown University

The basal ganglia and dopaminergic systems are well studied for their roles in reinforcement learning and reward-based decision making.   Much work focuses on 'reward prediction error' (RPE) signals conveyed by dopamine and used for learning. Computational considerations suggest that such signals may be enriched beyond the classical global and scalar RPE computation, to support more structured learning in distinct sub-circuits ('vector RPEs').   Such signals allow an agent to assign credit to the level of action selection most likely responsible for the outcomes, and hence to enhance learning depending on the generative task statistics. I will first describe the computational models spanning levels of analysis from implementation to function. I will then present evidence across species and methods -- from fMRI and EEG in humans to calcium imaging of striatal dopamine terminals in rodents -- that RPE signals are modulated by instrumental task demands, in accordance with vector RPEs.

TALK 2: MUTUAL BENEFITS: COMBINING REINFORCEMENT LEARNING WITH SEQUENTIAL SAMPLING MODELS

Birte U. Forstmann; University of Amsterdam

Reinforcement learning models of error-driven learning and sequential-sampling models of decision making have provided significant insight into the neural basis of a variety of cognitive processes. Until recently, model-based neuroscience research using both frameworks has evolved separately and independently. Recent efforts have illustrated the complementary nature of both modelling traditions and showed how they can be integrated into a unified theoretical framework, explaining trial-by-trial dependencies in choice behavior as well as response time distributions. Here, we expand on this framework. I will outline a theoretical background of such an integration and review current efforts towards this goal. Next, I will show data from ultra-high field functional magnetic resonance imaging testing concrete predictions from this novel framework in the brain. Finally, I will argue that the integration of both modelling traditions provides mutual benefits for both fields, and highlight promises of this approach for cognitive modelling and model-based cognitive neuroscience.

TALK 3: NEUROCOMPUTATIONAL MECHANISMS OF KNOWLEDGE ACQUISITION AND GENERALIZATION

Alison R. Preston; University of Texas at Austin

We acquire knowledge by connecting events that are experienced at different times and places. Prior work has shown that the brain represents the structure of an environment within cognitive maps that support flexible behavior within that environment, such as taking a shortcut to get to a goal location. Real-world adaptive behavior, however, is also supported by recognizing higher-order connections among distinct contexts that have similar internal structures; for example, navigating a new city may be facilitated by inferring that cities generally have downtown, midtown, and uptown areas with distinct properties. We used neurocomputational approaches to examine how the brain forms and uses generalized cognitive maps to support novel decisions. Participants learned object pairs (AB, BC) that were drawn from distinct triads (ABC) which shared the same internal structure. We found that hippocampal and frontoparietal regions formed higher-order cognitive maps wherein hierarchically-aligned representations coded cross-triad relationships with a common geometric structure. Critically, hierarchical cognitive maps were formed despite the lack of explicit reinforcement to do so. Using a combination of neural network simulations and computational modeling of neural data, we further showed that such hierarchically aligned maps provide an efficient representation of both individual events and the common structure across those events. Using these computational approaches, we show that frontoparietal representations promoted efficient inference by coding inferred event relationships with a director vector within the geometric structure that was consistent across triads, speeding inference decisions. These data show how neural representations extend knowledge beyond direct experience to support generalization through inference.

TALK 4: PROBABILISTIC LINKING FUNCTIONS FOR MIND, BRAIN, AND BEHAVIOR

Brandon M. Turner; The Ohio State University

The link between mind, brain, and behavior has mystified philosophers and scientists for millennia. Scientists who study cognition infer underlying processes either by observing behavior (e.g., response times, percentage correct) or by observing neural activity. These two types of observations have traditionally supported two separate lines of study. The first is led by cognitive modelers, who rely on behavior alone to support their computational theories. The second is led by cognitive neuroimagers, who rely on statistical models to link patterns of neural activity to experimental manipulations, often without any attempt to make a direct connection to an explicit computational theory. Recent progress has been made by forming statistical associations between manifest variables of the brain (e.g., EEG, fMRI) and manifest variables of behavior (e.g., response times, accuracy) through hierarchical latent variable models (Turner et al., 2018).  Within this framework, one can make inferences about the mind in a statistically principled way, such that complex patterns of brain-behavior associations drive the inference procedure. In this talk, I will discuss a recent approach called joint modeling that mutually constrains what we learn about the cognitive process from both the computational model and the neurophysiology. The central idea of this approach is to use the information in the neurophysiology to enhance or guide what the cognitive model says about the cognitive process of interest. I will highlight the utility of this approach from a methodological perspective as well as summarize a few key applications.

SYMPOSIUM 8: THE MEETING OF PERCEPTION AND MEMORY IN THE BRAIN

Chair: Marc Coutanche, University of Pittsburgh

Speakers: Marc Coutanche, Chris Baker, Jennifer Ryan, Morgan Barense

Perception and memory are intrinsically linked. Perceptual processes influence which information enters memory, and existing memories influence how we process perceptual input. This symposium will examine questions that speak to how and where perception and memory meet in the brain. The speakers will each tackle this topic in a unique way, giving an opportunity to identify shared cognitive and neural principles. In the first talk, Marc Coutanche will ask how information at distinct perceptual and conceptual levels can predict encoding success, and how levels of ventral stream reactivation affect memory outcomes, using behavior, fMRI, and convolutional neural networks. Next, Chris Baker will present fMRI findings that address the different levels of granularity that are elicited in high-level visual cortex during memory encoding and recall, revealing a spatial organization associated with recall. We will then move onto Jennifer Ryan, who will discuss how visual exploration and hippocampal binding processes are inherently linked, drawing on findings from behavioral, neuropsychological, neuroimaging, and computational modeling methods. Finally, Morgan Barense will present behavioral and neuroimaging studies of predictive coding, a continuous bridge between memory and current perceptual input, to ask how prediction errors at naturalistic event boundaries affect episodic memory updating. With overlapping goals but different perspectives, we hope to identify commonalities that can shed light on these issues, while raising new questions.

TALK 1: ROLES OF PERCEPTUAL AND CONCEPTUAL HIERARCHIES IN THE FORMATION OF MEMORIES

Marc Coutanche, University of Pittsburgh

A visual stimulus is represented at multiple levels across the human visual system: from low-level visual properties to high-level meaning. What roles do these levels play in memory formation? I will present results from several recent investigations that address this question. First, I will discuss a study that used convolutional neural networks to identify how early and later stages of the visual hierarchy help predict whether an image will be encoded into memory. Findings from two behavioral and one fMRI experiment suggest images are more likely to be successfully remembered when they are discriminable at early visual levels, but more similar at higher visual levels, where the relevant stage depends on the use of single or multiple semantic categories. Second, I will present findings from an fMRI study of how pattern reactivation of novel concepts relates to memory outcomes. Here, participants were introduced to image?word associations for novel rare animals, and were then asked to retrieve the associations one month later, each during an fMRI scan. The results suggest that the degree of reactivation of item and taxonomic-category (e.g., 'mammal') information within different regions of the ventral stream, relates to distinct memory outcomes. Together, these findings suggest that different perceptual and conceptual levels play important and distinct roles in achieving successful memory performance.

TALK 2: DISTINCT PROFILES OF PERCEPTION AND MEMORY IN HIGH-LEVEL VISUAL CORTEX

Chris Baker, National Institute of Mental Health

High-level visual cortex has been characterized by the presence of category-selective regions that respond preferentially to certain classes of stimuli (e.g. scenes, faces, objects). During memory recall, these regions are commonly thought to exhibit similar responses to those observed during perception, although these studies often focus on limited stimulus properties. In an item-based fMRI recall task, we investigated the nature of representations elicited during encoding and memory recall in category-selective regions by decoding multiple levels of information. Stimuli were trial unique and ranged in granularity from broad stimulus class (scenes, objects) to types of objects or scenes (e.g. natural, manmade) to individual sub-categories (e.g. living room, cupcake). While we find that the patterns of response in object and scene-selective cortex do contain information about recalled items that resembles that during encoding, this information tends to be quite coarse allowing decoding of stimulus class (objects versus scenes) but not other stimulus dimensions that are detectable during encoding. Further, we observed segregation within category-selective cortex between those voxels showing strongest effects during encoding and those during recall. Finally, in a whole-brain analysis, we observed the strongest similarity between encoding and recall in regions anterior to the category-selective cortex. These results highlight key differences in representational structure and spatial distribution between encoding and recall. More generally, these results are consistent with our prior work suggesting systematic relationships between regions engaged during perception and those engaged during recall throughout high-level visual cortex.

TALK 3: THE RECIPROCAL LINK BETWEEN MEMORY AND VISUAL EXPLORATION

Jennifer Ryan, Rotman Research Institute, Baycrest

The oculomotor and hippocampal memory systems interact in a reciprocal manner, on a moment-to-moment basis. Memory influences ongoing viewing behavior by increasing the efficiency of active vision. Conversely, eye movements serve to accumulate information from the visual world, contributing to the formation or updating of coherent memory representations. Eye movements may also contribute functionally to memory retrieval by reconstructing the rich, vivid, spatiotemporal details from memory. These interactions are mediated by the vast structural and functional links between the two systems. Findings from human and non-human animals, using behavioral, neuropsychological, neuroimaging, and computational modeling methods, will be highlighted to show that visual exploration and hippocampal binding processes are inherently linked, and that such an exploration-binding link is altered with hippocampal dysfunction.

TALK 4: PAST MEETS PRESENT: PREDICTION ERROR DRIVES EPISODIC MEMORY UPDATING

Morgan Barense, University of Toronto

How does the brain link past, present, and future? The concept of predictive coding provides a framework that bridges memory and perception. We draw on past experience to make predictions, and then compare those predictions to present perceptual input. This comparison process allows the brain to segment continuous experience, learn from error, and adaptively integrate new information into memory. Converging evidence from animals and humans has implicated prediction error, or surprise, as a key mechanism that renders established memories malleable. We developed a naturalistic paradigm to elicit prediction error during memory reactivation. To create surprising event boundaries, we interrupted narrative videos immediately before the expected conclusion. Through a series of behavioral and neuroimaging studies, we demonstrated that prediction errors at event boundaries allow episodic memories to be destabilized and updated with new, semantically relevant information. The effect of prediction error on memory was critically time-dependent, consistent with reconsolidation theory. Using fMRI, we showed that trial-wise neural activity in the hippocampus, ventral tegmental area, and angular gyrus was related to prediction error and memory updating. Our findings support the idea that the brain switches between internal and external modes of information processing. After an event boundary, internally-oriented processing (e.g., pattern completion, replay) strengthens episodic memory. However, surprising or salient perceptual input triggers a switch to externally-oriented processing. After a surprising event boundary, the hippocampus is primed to integrate new details into memory. Broadly, our findings showcase the dynamic interplay between perception and memory, through the overarching framework of predictive coding.

SYMPOSIUM 9: CORTICAL GRADIENTS AND THEIR ROLE IN COGNITION

Chair: Daniel Margulies, CNRS

Speakers: Boris Bernhardt, Noam Saadon-Grosman, Micah Murray, Jonathan Smallwood

While cortical areas have long been considered the building blocks of cortical processing, an emerging perspective suggests cortical functions are mediated along spatial gradients of organization. Cortical gradients provide a general framework for describing a global hierarchy that spans different processing streams, and establishes a mechanism for the large-scale structure of the cerebral cortex to enable sensory integration and diverse forms of cognition. We will address this topic through talks describing different implications and challenges for a gradient-based model of cortical organization and its role in cognition: How are different features of cortical connectivity and microstructure organized along gradients, and how does their deviation in disease account for atypical function? How do sensory topographies impact on the cortical layout within association cortex? How can we reconcile a global hierarchy of 'primary sensory-motor to higher-order cognitive functions' with the observation of that primary cortical areas also demonstrate multisensory properties? And what is the role of large-scale cortical gradients in enabling the dynamics underlying distinct mental states? The four 25-minute presentations will be followed by a 20-minute moderated discussion and Q&A from the audience.

TALK 1: THE INFLUENCE OF BRAIN STRUCTURE ON TYPICAL AND ATYPICAL BRAIN FUNCTION

Boris Bernhardt, Montreal Neurological Institute, McGill University

Neuroscience has the potential to explain how brain function arises from its underlying structure, and how brain dysfunction emerges from diseases associated with structural abnormalities. My talk will overview new work from the lab that analyzed cortex-wide microstructural coordination in humans, and that derived novel measures of structural wiring and hierarchies via advanced modeling of multimodal MRI data.  Leveraging post-mortem histological and transcriptomic techniques, we could furthermore show that these microstructural hierarchies derived from in vivo imaging reflect underlying cytoarchitecture and gene expression patterns. Studying large cohorts of healthy individuals as well as patients with structural brain anomalies, our work furthermore shows that new models of structural wiring can make robust predictions of typical as well as atypical functional connectivity and dynamics. Our results advance our understanding of how microstructural properties produce a hierarchical cortical wiring scheme that governs large-scale functional gradients and signal flow in cortical areas.

TALK 2: CORTICAL SOMATOSENSORY HIERARCHICAL GRADIENTS

Noam Saadon-Grosman, Shahar Arzy, Yonatan Loewenstein, Hebrew University

Multiple body maps in different cortical areas characterize information processing in the cortex. Additionally, electrophysiological studies in non-human primates have demonstrate hierarchical relationship between several somatosensory-responsive regions. However, a large-scale understanding of cortical somatosensory processing directions, analogous to the dorsal and ventral streams in the visual cortex, has been lacking. Therefore, we set out to characterize somatosensory hierarchies in the entire cortical representation. We applied phase-encoded bilateral full-body light touch stimulation under functional MRI. We quantified selectivity, a measure of the specificity of the response to a preferred body-part (the fMRI equivalent of neuronal receptive field), as well as the response?s laterality, a measure of the dominance to contralateral response. Incorporating multi-modal cortical parcellation, we defined gross anatomical regions and computed selectivity and laterality along four spatial axes originating from the central sulcus. Our results suggest somatosensory hierarchical gradients that follow three anatomically distinct directions: parietal (from the central sulcus posteriorly in the lateral-parietal lobe), frontal (from the central sulcus anteriorly in the frontal) and medial (inferiorly and anteriorly in the medial wall). We propose that as in the visual domain, these directions are streams of somatosensory information processing.

TALK 3: A MULTISENSORY PERSPECTIVE ON PRIMARY CORTICES

Micah M. Murray, University Hospital Center and University of Lausanne

The turn of the 21st century introduced evidence dramatically changing our conception of functional brain organization and cortical gradients. Anatomic evidence in non-human primates showed that primary cortices were directly (i.e. monosynaptically) interconnected. Some even proposed that the whole neocortex is essentially multisensory in nature. In this talk, I will overview our efforts to provide evidence in humans that primary cortices are indeed fundamentally multisensory and play an active role in multisensory processes and perception. This evidence is provided from a full pallet of human brain imaging, mapping, and stimulation methods. First, there is both convergence and integration occurring within primary visual and auditory cortices at early post-stimulus stages. Second, these processes are behaviourally relevant, can be linked with excitability changes, and impact perceptual outcome. Third and more generally, early-latency multisensory processes extending from primary to lateral-occipital regions play a direct role in recognition memory. Finally, we extend such findings to show how multisensory processes across the lifespan are tethered to global cognition and its breakdown, providing a potential access point for screening and treatment. Together, these data underscore how multisensory research and its applications in basic, clinical, and applied research is changing long-held models of graduated functional brain organization.

TALK 4: NEUROCOGNITIVE HIERARCHIES AS A STATE SPACE FOR ON-GOING THOUGHT

Jonathan Smallwood, University of York

Our experience is not always focused on events in the outside world, we often focus internally on self-generated mental content. Understanding the neural basis of these different patterns of ongoing thought requires understanding how the cortex leverages the constraints imposed by its organisation to produce different neurocognitive states. This talk considers evidence that uses machine learning, experience sampling and neural activity to establish that neurocognitive hierarchies can provide a coordinate space for understanding different modes of neurocognitive operation. In particular, these studies suggest that different types of states can fall at either extreme of a dimension resembling how the brain responds to task demands. These results establish that neural hierarchies provide a flexible coordinate space within which to understand the dynamics of unconstrained thought, and demonstrate neural patterns resembling the brain?s response to external task demands, capture important aspects of self-generated experience.

SYMPOSIUM 10: SPECIFICS AND GENERALITIES: BEYOND THE SEMANTIC-EPISODIC DISTINCTION

Chair: Chi Ngo, Max Planck Institute for Human Development

Speakers: Chi Ngo, Dagmar Zeithamova, Alexa Tompary, Sean Polyn

An adaptive memory has to serve both the need to construct generalized knowledge across experiences to optimally guide behaviors in novel situation, and the need to keep individual episodes distinctive to minimize interference. These functions are thought to rely on distinct memory systems. The former underscores an appreciation of the commonalities across overlapping experiences, whereas the latter retains the specificity and distinctiveness of individual episodes. These two memory systems are inextricably intertwined and exert joint influences on behavior. However, generalized knowledge and episodic memory have often been investigated in separate lines of research. The work presented in this symposium will characterize the bidirectional influences between generalized knowledge and memory of specific instances from the neural and behavioral levels of analyses. The presentations will cover a set of central questions: (1) when do generalized knowledge and episodic memory emerge in early development; (2) how neural representations of specific instances and generalized knowledge representations may emerge across learning, and how categorization decisions are supplemented by memory for specific experiences; (3) how the structure of prior knowledge explains the extent of distortions in episodic retrieval; and (4) how categorical structure influences both behavioral performance and neural signals during free recall. Collectively, our symposium will include some of the newest and most exciting work in this line of research by integrating theories drawn from developmental science, cognitive neuroscience, and neurocomputational science, and with methods ranging from behavioral, computational, functional magnetic resonance, and transcranial magnetic stimulation approaches.

TALK 1: GENERALIZED KNOWLEDGE AND EPISODIC MEMORY IN DEVELOPMENT

Chi Ngo, Max Planck Institute for Human Development, Berlin

Young children display prodigious capacities to extract generalized knowledge about the environment and to build strong semantic memory and yet they have difficulty remembering specific events. This observation predicts a developmental lead-lag relation between constructing schematic knowledge and episodic memory. However, past research has primarily studied these processes in isolation and focused on different age windows, creating critical blind spots in our understanding of the relative emergence of generalization and episodic memory. Here, we directly tested the prediction of developmental precedence of generalization over episodic memory. We administered a novel task that allowed for assessing generalization and episodic memory in children aged 4-8 and young adults with common stimuli and task demands. Participants learned a series of events, some of which shared commonalities with one another (e.g., Tabaluga was seen in different contexts, paired with different musical instruments each time). Generalization was operationalized as the ability to make a novel inference based on the series (e.g., Tabaluga would choose a novel musical instrument over objects from other semantic categories). Episodic memory was operationalized as detailed memories of the individual episodes, probed at different levels?context binding, item conceptual and perceptual precision. Although generalization and episodic specificity both improved with age, generalization performance exceeded some aspects of episodic specificity early in life. Crucially, generalization did not depend on memories of individual episodes, suggesting that generalization does not arise from abstraction over episodic memories in early development.

TALK 2: MEMORY SPECIFICITY AND CONCEPT GENERALIZATION

Dagmar Zeithamova, University of Oregon

Concept learning and episodic memory have been typically studied in distinct lines of research, assumed to rely on different memory representations and competing memory systems. In contrast, a single system view of concept learning assumes that concept generalization relies on specific representations formed by the episodic memory system and the existence of "generalized" representations is not necessary. To resolve between these views, we conducted a series of studies using two experimental paradigms that allow for simultaneous tracking of specific and generalized representations in the brain and behavior. One set of studies used binary-dimension stimuli well suited to fitting formal categorization models that assume reliance on specific vs. generalized concept representations respectively. Model predictions derived from behavior were related to brain activation. The results showed both specific and generalized concept representations emerging across learning in distinct loci, both contributing to categorization performance. The second set used face blend stimuli well suited for behaviorally testing both face-specific memory and category generalization. Neural pattern similarity analysis revealed neural representations of specific faces as well as category-level (generalized) representations that emerged across learning. Together, these findings demonstrate the existence of generalized concept representations in brain and behavior, but also reveal how categorization decisions are supplemented by memory for specific exemplars, reconciling competing theories of concept representation.

TALK 3: SEMANTIC KNOWLEDGE DISTORTS EPISODIC MEMORY: BEHAVIORAL AND NEURAL INVESTIGATIONS

Alexa Tompary, University of Pennsylvania

Retrieval is not a veridical recapitulation of past events, but instead an imperfect recombination of event-specific details and other general knowledge.  Integrating these sources of information may improve the ?signal? of a memory, but introduce systematic errors if there are discrepancies between them. However, it remains unclear how the structure of semantic knowledge, like category typicality, biases new episodic memories. We predicted that typical (compared to atypical) category members would be more prone to bias by prior knowledge. In a series of behavioral experiments, participants encoded and retrieved image-location associations. Most members of a category (e.g. birds) were located near each other, but some typical and atypical category members were in random locations. Critically, we used a continuous retrieval measure of location memory to develop two measures: error, a measure of episodic specificity, and bias towards other category members, a measure of the influence of semantic knowledge. First, location memory was more precise for images that were spatially consistent with their category membership. Second, retrieval of typical category members was more biased towards category neighbors, relative to atypical members. Both effects replicated across multiple experiments and were disrupted when images were not arranged by category. This suggests that episodic retrieval is supported both by event-specific details and prior knowledge, and the structure of this knowledge explains the extent of distortions in memory. An ongoing experiment applying transcranial magnetic stimulation to the left anterior temporal lobe will clarify whether these sources of information are underpinned by distinct neural mechanisms.

TALK 4: NEURAL SIGNATURES OF TIME AND MEANING IN CATEGORIZED FREE RECALL

Sean Polyn, Vanderbilt University

During memory search, generalized semantic knowledge interacts with episodic memories of recent experience.  We examine these interactions in the categorized free-recall task.  In this task, participants study a series of items drawn from taxonomic categories associated with distinct neural signatures (celebrities, landmarks, and objects), and then recall these items in whatever order they come to mind. These interactions reveal themselves in the behavioral dynamics of the task: Participants produce recall sequences that simultaneously display strong temporal organization (whereby successively produced responses tend to come from nearby list positions) and semantic organization (whereby successively produced responses tend to be meaningfully related).  These interactions are also revealed in the task?s neural dynamics: Category-specific neural patterns can be tracked during both study and recall periods, and these patterns show integrative effects, whereby information about the category identity of items from the recent past persists in the neural signal. We have developed a neurocognitive modeling framework to explain these neural and behavioral dynamics. The model describes how semantic information can be integrated into a representation of temporal context, and predicts the representational structure of neural activity and the organizational effects observed during recall. We use this framework to infer the representational structure of memories for the study experience, and the nature of the executive processes guiding search through these memories. Finally, I?ll describe recent behavioral and neuroimaging experiments in which we use a distraction task to disrupt temporal and category integration, which has corresponding effects on both recall behavior and task-related neural signals.

SYMPOSIUM 11: DEEP DATA: THE CONTRIBUTION OF CASE STUDIES AND SPECIAL POPULATIONS IN THE ERA OF BIG DATA

Chair: Erez Freud, York University

Speakers: Erez Freud, Ella Striem-Amit, Shayna Rosenbaum, Bradford Z. Mahon 

Classic neuropsychological case studies helped found the field of cognitive neuroscience, showing which cognitive faculties can be dissociated from one another, and contributing to the discovery differentiated processing streams. However, given the benefits of big datasets and large sample sizes for reliability, do case studies and the research of small unique populations still have a role in the future of the field? This symposium highlights patient and population research across action, perception, memory, and language, illustrating the benefits of well-characterized deep individual data to cognitive neuroscience. Talks will challenge the dissociation of action and perception in vision, by inspecting the role of the dorsal stream in object recognition; explore the role of hand motor experience for tool and action representations in individuals born without hands; explore the language pathways using direct electrical stimulation mapping in awake brain surgery; and inspect the role of the hippocampus in pattern separation across modalities and content domains. Across the four content domains that are discussed, causal evidence resulting from case studies and special populations places firm constraints on plausible theoretical distinctions. These allow for the generation of new hypotheses about the brain and mind, which, in turn, can be further examined in large datasets.

TALK 1: THE ROLE OF THE DORSAL PATHWAY IN OBJECT PERCEPTION

Erez Freud, York University

According to the two visual pathways hypothesis, the ventral visual pathway promotes vision-for-perception, while the dorsal pathway promotes vision-for-action. Seminal single-cases neuropsychological investigations supported this functional dissociation. However, accumulating evidence challenges this binary distinction and suggests that regions in the dorsal pathway derive object representations that might play a functional role in object perception. In my talk, I will discuss evidence from psychophysical, kinematic and neuroimaging studies with visual agnosia patients that were aimed to explore the nature of object representations in the dorsal pathway. The results from these studies highlight (a) the plausible role of the dorsal pathway in object perception, (b) the interplay between shape representations derived by the ventral and the dorsal pathway and (c) the association between the developmental trajectories of vision-for-perception and vision-for-action. Together, these findings are consistent with the view that object perception is not the sole product of ventral pathway computations, but instead relies on a distributed network of regions.

TALK 2: PERCEPTION AND ACTION WITHOUT HANDS

Ella Striem-Amit, Georgetown University

Our hands are at the core of our action system, affecting our representation of both actions and manipulable objects, such as tools. But what role do the hand motor features themselves, or our experience using them, play in these representations? I will present a series of fMRI experiments addressing this question by investigating individuals born without hands, who use their feet to perform everyday actions. These works revealed representations abstracted from the hand-specific features, as well as representations affected by motor knowledge and experience. First, I will present findings showing typical processing for visual hand images and actions, regardless of motor experience. Second, I will present findings showing the effect of the absence of motor use knowledge for some common objects which the dysplasics cannot use, affecting a distributed system integrating different attributes of object knowledge. Third, I will present findings related to action execution, which reveal a gradient between body-part selective regions and effector-invariant regions, allowing exploration of the different levels of abstractness in representing actions. Finally, I will discuss the benefits of thespecial populations, congenitally deprived of experience in specific manners to address the causal role of experience in shaping our brain and mind, drawing parallels across studies of people born blind, deaf or without hands.

TALK 3: PATTERN SEPARATION FOLLOWING DENTATE GYRUS LESIONS

Shayna Rosenbaum, York University

Healthy older adults and individuals with hippocampal compromise experience a notable decline in episodic memory. These memory problems may be due to difficulties discriminating highly similar inputs belonging to separate, yet overlapping, events into discrete episodes at encoding, a process known as ?pattern separation.? Pattern separation in humans is often estimated behaviourally with visual recognition tests in which participants must select previously studied everyday objects from visually and conceptually similar lures and from dissimilar foils. However, the types of stimuli, domains, processes, and representations it impacts remain unclear. To what extent do presumed deficits in pattern separation extend to modalities other than vision, such as audition? Within vision, is it more evident for one class of stimuli, such as scenes, than another, such as faces? Is the extent of the deficit affected by prior knowledge? To address these issues, I will present a series of studies involving a unique individual with bilateral lesions to the dentate gyrus, a region of the hippocampus strongly associated with pattern separation. A first set of studies involve tasks requiring discrimination of novel auditory stimuli within memory and perception. I will then describe how categorical perception, which refers to greater differentiation of stimuli at a perceived boundary compared to within boundaries, might relate to pattern separation. Findings from this work illustrate how the study of single cases continue to contribute to hypotheses and theories that may steer the field in new directions.

TALK 4: DIRECT ELECTRICAL STIMULATION MAPPING OF LANGUAGE PATHWAYS DURING AWAKE BRAIN SURGERY

Bradford Z. Mahon, Carnegie Mellon University

An emerging approach for understanding the neural substrates of speech processing emphasizes integrated functional analysis of cortical regions, major white matter pathways, and behavioral consequences of lesions to those structures.  Language mapping with direct electrical stimulation in awake neurosurgery patients undergoing removal of brain tumors offers a powerful approach for testing hypotheses about the cortical and subcortical systems critical for language processing. I describe a case series in which the first patient was tested with detailed neuropsychological testing pre- and post-operatively, and the second patient was studied using cortical and subcortical electrical stimulation mapping during awake brain surgery. Both patients had gliomas in the dominant frontal lobe. The first patient experienced a reduction in verbal fluency subsequent to partial resection of the Frontal Aslant Tract, which connects the pre-supplementary motor cortex with the inferior frontal gyrus. Motivated by those findings, we designed a novel test of speech fluency that was administered during the awake portion of the second patient?s surgery. We found that electrical stimulation of the Frontal Aslant Tract specifically disrupted speech fluency, leaving lexical access and articulatory processes intact. The findings are interpreted in the context of the hypothesis that the Frontal Aslant Tract mediates the integration of syntagmatic relations among words with positional level planning.

SYMPOSIUM 12: WHAT DETERMINES CATEGORY SELECTIVITY IN THE CORTEX?

Chair: Talia Konkle, Harvard University

Speakers: Michael Arcaro, Marius Peelen, Rebecca Saxe, Hans Op de Beeck

What determines category selectivity in the cortex? Methodological advances are providing new insights into the nature of category representation in the brain and how it emerges in development. For example, controlled rearing studies in non-human primates are revealing the pivotal role of intensive early experience in category selective responses, and the first human infant neuroimaging experiments are revealing what kind of category selectivity is already present in the first few months of life. At the same time, deep neural network models are providing ways to operationalize hypotheses about the feature tuning along the visual hierarchy.  In this symposium, we take on the question of what kind of category-related information is represented along the ventral visual stream, in the context of why it is there and how it comes to be that way.  The goals of the symposium are to present a compact view of the range of perspectives on these questions, to discover the points of agreement and tension, and ideally to articulate the directions we can take with our new tools in order to arrive at an accepted explanation of category-selectivity in the cortex.

TALK 1: CORTEX IS CORTEX: UBIQUITOUS PRINCIPLES DRIVE FACE-DOMAIN DEVELOPMENT

Mike Arcaro, University of Pennsylvania

How does the brain develop category-selective tuning? One possibility is that the brain is innately organized into anatomically distinct regions, each processing different biologically important image categories. However, genetic specification of something as particular as an image category seems in tension with known activity-dependent, self-organizing wiring properties of the visual system. I will present an argument that seemingly complex phenomena like category selectivity can be accounted for by such ubiquitous principles of brain organization. We performed functional neuroimaging in non-human primates across development, finding that the newborn visual system already comprises an extensive retinotopic proto-architecture that carries with it a topographic organization for scale and shape. We propose that this architecture emerges early in development as a consequence of molecular cues and self-organizing principles and provides the scaffolding for subsequent experience-dependent specializations throughout the visual system. Throughout development, monkeys preferentially look at faces. This systematic foveation enforces a retinotopic regularity of face experience that, in conjunction with an intrinsic retinotopic architecture, can account for the observation that face-selective domains develop in parts of the proto-architecture representing the central visual field. Visual experience was critical for the development of these domains, as monkeys reared without seeing faces did not develop domains selective for images of faces, but did develop domains for other experienced image categories. Thus, the development of category-selective domains involves the same kind of activity-dependent, self-organizing rules that are widely viewed as sufficient to account for the exquisite organization and complex receptive field properties of primary visual cortex.

TALK 2: CATEGORY-SELECTIVE REGIONS IN VISUAL CORTEX: WHAT ARE THEY FOR?

Marius Peelen, Donders Institute for Brain, Cognition and Behaviour

Visual input supports diverse tasks such as object recognition, action guidance, tool use, spatial navigation, inferring the emotions of other people, and interpreting written text. These tasks each require representations of the outside world that emphasize some high-level dimension(s) of their critical stimuli, while abstracting over other irrelevant dimensions. We argue that the demand for these diverse types of representations is a driving force behind the evolution and development of category selectivity in visual cortex. Focusing on tools, we will present evidence: 1) that category-selective responses are closely aligned with knowledge of what a thing means to the observer; that is, the cognitions and behaviors an object is associated with (e.g., knowing that an object can be used as an effector); 2) that specific visual features, visual input, and even visual experience are not necessary to elicit category-selective responses; and 3) that category-selective regions are structurally and functionally connected to regions in other parts of the brain supporting related tasks more broadly (e.g., the fronto-parietal tool network). We conclude that category-selective regions in visual cortex exist because they provide specialised perceptual input to wider networks that serve diverse behavioral goals, including object recognition, but also tool use, reading, and navigation. What exactly category-selective regions represent about their preferred category can only be fully understood by considering the goals that are achieved by the whole systems to which they contribute.

TALK 3: SOCIAL ORIGINS OF CORTICAL FACE AREAS

Rebecca Saxe, MIT

To what degree is category-selectivity of the ventral visual stream specified at birth, and how much arises from experience? Using functional magnetic resonance imaging in awake human infants (age 2-9 months), we have found that cortical regions responding preferentially to faces, scenes and objects are identifiable, with similar large-scale spatial organization to that of adults. In infants, category preferences appear simultaneously in occipital, temporal, parietal and frontal areas, with no hint of a posterior-to-anterior sequence. Yet the functional responses are less category-selective than in adults. What explains the earliest cortical responses to high-level visual categories? I propose that these responses reflect biased connectivity, both bottom-up and top-down. Connectivity from earlier visual areas could confer sensitivity to the characteristic visual statistics of faces, scenes, and objects; whereas connectivity from parietal and frontal areas enhances responses for category-relevant functions in cognition and action. For example, faces are not just a pattern of foveated, curvy, smooth objects: infants choose to look at faces to engage in positively valenced, contingent social interactions. Using functional Near Infrared Spectroscopy in infants (age 6-12 months), we have found that medial prefrontal cortex is active in response to these cues of social value, and predicts the infant?s subsequent looking to faces. The strongest version of my hypothesis is therefore that activity related to social value during social interactions, plausibly in medial prefrontal cortex and/or subcortical regions, directly guides the development of face-selective responses in occipital and temporal cortex.

TALK 4: FACTORS DETERMINING WHERE CATEGORY-SELECTIVE AREAS EMERGE IN VISUAL CORTEX

Hans Op De Beeck, KU Leuven

A hallmark of functional localization in the human brain is the presence of areas in visual cortex specialized for representing particular categories such as faces and words. Why do these areas appear where they do during development? I will survey and present recent neuroimaging and computational evidence that seems contradictory in the light of simple hypotheses aiming to explain the characteristics and location of emerging category selectivity. Instead, we need to integrate at least three factors to explain the data: (i) pre- existing selectivity for properties of the stimulus, (ii) the computational hierarchy of the visual system, and (iii) domain- specific patterns of connectivity to nonvisual regions.  The resulting framework posits that the cortical location of category selectivity is constrained by which category will be represented, how it will be represented, and why the representation will be used.