Seminars

In realistic choice tasks, especially sequential ones like mazes, actions are separated from their consequences by many steps of space and time. A central computational problem in decision making -- which recurs in various guises such as credit assignment and planning -- is spanning these gaps to work out the long-term consequences of candidate actions. I review recent experimental and theoretical work aimed at understanding the mechanisms by which the brain solves this problem. First, I review recent studies in rodents that monitor neural signatures of reward expectancy in rodents to monitor how the brain propagates information about individual experiences with outcomes to distal choice points. Second, I report ongoing theoretical work that aims to clarify how the brain can judiciously build and query cognitive maps so as to achieve effective decisions while minimizing computational costs. This offers a formal, resource-rational perspective on a range of issues such as habits and slips of action in the healthy brain, but also suggests candidate mechanisms that may underlie dysfunctions such as compulsion, rumination, and avoidance.

How do babies start to make sense of their language? We used to think that infants convert the speech they hear into a sequence of discrete categories, and that once they got good at that, they could store words much the way a dictionary does, as a sequence of category labels. This would be efficient, and useful: according to the phonological principle, to a first approximation, two portions of speech with the same labels are the same word; and with different labels, different words. Nowadays, psycholinguists are not so sure this is the right way to think about young children. Why not? (a) Toddlers are resistant to interpreting minimally-deviant phonological strings as novel words--"tog" is a bad "dog," not a new non-canine thing. (b) Transcriptions of speech to infants suggest that the phonological principle is not actually sound, in infants' experience, because parental realizations of words are so variable. (c) Computational modeling has failed to show a path from phonetics to categories in the first place, undermining the phonological foundation. Does this mean children don't have phonological categories? I'll have a go at defending infant phonology, and suggest how I think we can make progress in building better, more realistic models.

For humans and nonhuman primates alike, deeper and more numerous social connections promote health, well-being, and survival. Precisely how primates navigate the multidimensional social relationships that structure daily life and shape survival and reproductive success remains a mystery. Here, we combine ethological analyses with new wireless recording technologies and computer vision to uncover neural signatures of natural behavior in unrestrained, socially interacting rhesus macaques. Neuronal activity in prefrontal and temporal cortex robustly encoded 24 species-typical behaviors, and also signaled the presence and identity of neighboring monkeys. Male-female partners demonstrated near-perfect reciprocity in grooming, a key behavioral mechanism supporting friendships and alliances, and neural activity maintained a running account of these social investments. When confronted with an aggressive intruder, behavioral and neural population responses reflected empathy and were buffered by the presence of a partner. Surprisingly, neural signatures in prefrontal and temporal cortex were largely indistinguishable and irreducible to visual and motor contingencies. Our work reveals a highly distributed neurophysiological ledger of social dynamics, a potential computational foundation supporting communal life in primate societies, including our own.

Learning to behave adaptively in a complex world relies on our ability to learn and deploy useful representations of the environment that are appropriate for the task at hand. In this talk, I will show that representation learning can be understood as sequential sampling over a multidimensional observation space. And I will present results from ongoing work investigating (1) how this process is implemented in the human brain; (2) what priors might support sampling in naturalistic environments; (3) a role for affect in biasing which representations are sampled.

Music is ubiquitous in human life and has been developed by all human societies. Music modulates brain activity in multiple brain regions with very different functions, including auditory, memory, motor, and dopaminergic and reward-related areas, among many others. In this vein, researchers have leveraged music to probe the human brain, with the cognitive neuroscience of music emerging as a new field of scientific interest in the last 20 years. In this talk, I will discuss recent findings from our lab in which we leverage music to assess the neural mechanisms of reward and its intersection with memory formation, motor processing, and social interactions. First, I will show how music can be used to assess the effects of both reward and novelty on long-term memory formation at the behavioral and neural levels. This first study includes the use of computational modelling to measure novelty in music. I will then turn to the use of music to understand motor processing, introducing the results of a study assessing music improvisation. In this work, we measured behavioral, physiological, and neural responses while people improvised with a music therapist and results will highlight the important connection between motor and reward-related processes while music-playing. Finally, I will also talk about a new wearable device developed at the lab that allows for the tracking of reward-markers of music‒the goosebumps we experience while listening to music‒in naturalistic environments. I will end by showing how this device can be potentially used in experiments assessing emotion in real-world scenarios.

Humans live a cognitively constrained existence. At any one time, we can only choose from a small set of cognitive operations and process a fraction of our experiences. Despite this, our cognition demonstrates remarkable adaptability in navigating a complex and partially known world. In this talk, I will highlight two types of cognitive constraints that are critical in shaping adaptive behavior in human long-term memory: architectural constraints and resource constraints. I will demonstrate how various behaviors in free recall and cued recall tasks can be explained through the lens of a memory system that optimizes its performance within these constraints. When searching for memories, people initiate and continue their recall in a way that maximizes the effectiveness of their retrieval cues, stop the search when success seems unlikely, and prioritize stronger memories. Finally, I will discuss our efforts to link cognitive constraints with existing neuroscience literature on successful memory formation and recall. Aligning cognitive models with neural patterns helps us better understand the cognitive mechanisms behind these patterns while also offering a useful tool for further evaluating the cognitive theory.

Naturalistic experimental paradigms in cognitive neuroscience arose from a pressure to test, in real-world contexts, the validity of models we derive from highly controlled laboratory experiments. In many cases, however, such efforts led to the realization that models (i.e., explanatory principles) developed under particular experimental manipulations fail to capture many aspects of reality (variance) in the real world. Recent advances in artificial neural networks provide an alternative computational framework for modeling cognition in natural contexts. In this talk, I will ask whether the human brain's underlying computations are similar or different from the underlying computations in deep neural networks, focusing on the underlying neural process that supports natural language processing in adults and language development in children. I will provide evidence for some shared computational principles between deep language models and the neural code for natural language processing in the human brain. This indicates that, to some extent, the brain relies on overparameterized optimization methods to comprehend and produce language. At the same time, I will present evidence that the brain differs from deep language models as speakers try to convey new ideas and thoughts. Finally, I will discuss our ongoing attempt to use deep acoustic-to-speech-to-language models to model language acquisition in children.

Music is ubiquitous in human life and has been developed by all human societies. Music modulates brain activity in multiple brain regions with very different functions, including auditory, memory, motor, and dopaminergic and reward-related areas, among many others. In this vein, researchers have leveraged music to probe the human brain, with the cognitive neuroscience of music emerging as a new field of scientific interest in the last 20 years. In this talk, I will discuss recent findings from our lab in which we leverage music to assess the neural mechanisms of reward and its intersection with memory formation, motor processing, and social interactions. First, I will show how music can be used to assess the effects of both reward and novelty on long-term memory formation at the behavioral and neural levels. This first study includes the use of computational modelling to measure novelty in music. I will then turn to the use of music to understand motor processing, introducing the results of a study assessing music improvisation. In this work, we measured behavioral, physiological, and neural responses while people improvised with a music therapist and results will highlight the important connection between motor and reward-related processes while music-playing. Finally, I will also talk about a new wearable device developed at the lab that allows for the tracking of reward-markers of music‒the goosebumps we experience while listening to music‒in naturalistic environments. I will end by showing how this device can be potentially used in experiments assessing emotion in real-world scenarios.

Cognitive neuroscience provides a rich account of how brain systems give rise to diverse forms of learning and memory. However, these theories and models are based mostly on adult data and often neglect early development, the greatest period of learning in life. Key challenges for studying this period include that there are limited behavioral measures available and that infant-friendly neural measures (EEG and fNIRS) have coarse spatial resolution and lack access to deep-brain structures like the hippocampus that are critical for adult learning and memory. I will present our recent work adapting fMRI, a technique that addresses some of these limitations, for studying learning and memory in awake human infants. These studies reveal how hippocampal function develops over the first two years of life and shed light on the mysterious amnesia for this period all humans come to experience.

I propose that a field’s diversity is affected by what its members believe is required for success: Fields that value exceptional intellectual talent above all else may inadvertently obstruct the participation of women and (some) racial/ethnic minority groups. The environment in these fields may be less welcoming to women and minority groups because of the cultural stereotypes that associate intellectual talent -- brilliance, genius, etc. -- with (white) men. This proposal is supported by observational and experimental data from a wide range of fields in the sciences and the humanities, as well as by developmental data that reveal how early these stereotypes take hold.

Our understanding of the neurobiology of both human and non-human primate behavior largely derives from artificial tasks in highly- controlled laboratory settings, overlooking most natural behaviors their brains evolved to produce. How these animals navigate the multidimensional social relationships that structure daily life and shape survival and reproductive success remains largely unexplored at the single neuron level. Here, we combine ethological analysis with new wireless recording technologies to uncover neural signatures of natural behavior in unrestrained, socially interacting pairs of rhesus macaques within a larger colony. Population decoding of single neuron activity in prefrontal and temporal cortex unveiled robust encoding of 24 species-typical behaviors, which was strongly modulated by the presence and identity of surrounding monkeys. Male-female partners demonstrated near-perfect reciprocity in grooming, a key behavioral mechanism supporting friendships and alliances, and neural activity maintained a running account of these social investments. When confronted with an aggressive intruder, behavioral and neural population responses reflected empathy and were buffered by the presence of a partner. Surprisingly, neural signatures in prefrontal and temporal cortex were largely indistinguishable and irreducible to visual and motor contingencies. By employing an ethological approach to the study of primate neurobiology, we unveil a highly-distributed neurophysiological record of social dynamics, a potential computational foundation supporting communal life in primate societies, including our own.

If you are reading this abstract, you probably have a tremendous amount of practice with this particular skill – moving your eyes across lines of tiny characters and decoding a message from them. It may even feel effortless to you. Indeed, many theories of literacy postulate that as children learn to read, word recognition becomes ‘automatic.’ In this talk I will argue that nonetheless, word recognition by skilled adults requires a great deal of selective attention and top-down control. I will focus on a region in the ventral temporal cortex known as the “visual word form area” (VWFA), which plays a critical role in reading. In several fMRI experiments we have investigated what drives activity in this region: seeing words, attending visually to words, or explicitly trying to read words? The results do not show an automatic activation by words, even when they are attended visually. Rather, we found a complex interaction of bottom-up selectivity for letter strings and top-down enhancement of word processing that is contingent on voluntary effort to read. We conclude that reading depends on the willful activation of a cerebral network that is exquisitely specialized.

There is a fundamental tension between storing discrete traces of individual experiences, which allows recall of particular moments in our past without interference, and extracting regularities across these experiences, which supports generalization and prediction in similar situations in the future. One influential proposal for how the brain resolves this tension is that it separates the processes anatomically into Complementary Learning Systems, with the hippocampus rapidly encoding individual episodes and the neocortex slowly extracting regularities over days, months, and years. But this does not explain our ability to learn and generalize from new regularities in our environment quickly, often within minutes. We have put forward a neural network model of the hippocampus that suggests that the hippocampus itself may contain complementary learning systems, with one pathway specializing in the rapid learning of regularities and a separate pathway handling the region’s classic episodic memory functions. This proposal has broad implications for how we learn and represent novel information of specific and generalized types, which we test across statistical learning, inference, and category learning paradigms. We also explore how this system interacts with slower-learning neocortical memory systems, with empirical and modeling investigations into how the hippocampus shapes neocortical representations during sleep. Together, the work helps us understand how structured information in our environment is initially encoded and how it then transforms over time.

Although ‘representation’ and ‘inference’ are central to cognitive neuroscience, they are often used in loose and contradictory ways. In other fields, new definitions have triggered major breakthroughs. At the beginning of the 19th century, there wasn’t a proper definition of ‘continuity’. As a result, mathematicians were unable to prove the consistency of calculus, or even to investigate its basic properties, such as whether a function can be everywhere continuous but nowhere differentiable. But after a proper definition was introduced, progress accelerated, giving rise to a new field, real analysis. Likewise, in the early 20th century, new definitions of ‘information’, ‘simultaneous’, ‘computable’, and ‘computationally tractable’ triggered major breakthroughs in engineering, physics, and computer science. The results were communication theory, the theory of relativity, computability theory, and complexity theory. I believe that cognitive neuroscience is at a similar point in its development. In this talk, I will survey some candidate definitions of ‘representation’, list their problems, and then sketch my own proposal. In brief, I think that there’s a necessary connection between representing and learning.

How do attentional networks influence conscious perception? I will present data from two studies: one featuring magnetoencephalographic recordings, and the other featuring intracerebral recordings assessing the effects of supra-threshold peripheral spatial cues on the conscious perception of near-threshold Gabors. Behavioral and neuroimaging results converge on the importance of lateralized front-parietal networks in shaping our visual conscious perceptions. I'll then discuss the relevance of our findings with respect to current theories of consciousness and conclude by relating them to a less-studied form of visual perception: visual mental imagery. I will briefly review the literature regarding human imagination, and then show recent neuroimaging evidence obtained using 3T and 7T fMRI, pointing at the role of the frontoparietal networks in supporting imagination. I will conclude by bridging between the fields of visual perception and visual imagination, pointing at the frontoparietal networks in these two processes.