How collaboration arrises and why it fails

How does a rat build a three-dimensional picture of the world using nothing but hair? Mitra Hartmann unpacks the biomechanics of whisker sensing, the distinction between active touch and passive sensation, and her vision for a tactile paintbrush that could scan objects in 3D.Subscribe for more from the Convergent Science Network podcast series.Hartmann begins by carefully distinguishing active touch from active somatosensation, drawing on James Gibson's original framework. Active touch requires purposeful exploration, not merely muscle engagement. A rat brushing its whiskers against objects to assess texture is performing active touch; a person grabbing a hot pan with a cloth is not. This distinction matters because it frames the central question of her research: how does an animal transform mechanical energy into a perception of the world?The rat whisker system serves as Hartmann's model for studying perception because its neural pathways are analogous to those carrying information from the human hand through the brain. Rats possess roughly 60 whiskers that sweep back and forth at speeds up to 1,000 degrees per second, with all sensory receptors located at the base of each whisker follicle rather than along the hair shaft. Multiple laboratories have demonstrated that rats can discriminate textures with whisker acuity comparable to human fingertips, navigate wall contours, and determine bar orientation. The open question remains why some animals actively whisk while others, like cats and dogs, do not.As an engineer, Hartmann argues that characterizing the mechanical input is the essential first step before understanding neural processing. Her group investigates whisker material properties, single-whisker mechanics, and natural whisking behavior to determine what signals reach the follicle during real-world exploration. She describes her concept of a whisker paintbrush, a tactile 3D scanner consisting of an array of instrumented bristles that could reconstruct object geometry through mechanical contact. Building such a device could also answer fundamental biological questions about whether independent whisker movement provides information advantages over passive arrays.

How does a dead fish swim upstream, and what does that reveal about the hidden intelligence of body design? Maarja Kruusmaa explores the surprising physics of fish locomotion, lateral line sensing, and why propellers may not be the last word in underwater engineering.Subscribe for more from the Convergent Science Network podcast series.Kruusmaa challenges naive biomimetics, the tendency to copy everything from nature without understanding which features actually matter. She draws parallels to early propellers with feathers and cars with horse compartments, arguing that the real engineering challenge is identifying which biological principles are worth extracting. While propellers remain a mature and powerful technology, fish outperform them in energy efficiency and acoustic stealth, leaving almost no wake behind them. The key advantage lies in distributed actuation across hundreds of muscle fibers, something current motor technology cannot replicate at practical scales.The conversation dives deep into fish swimming mechanics. Kruusmaa explains how swimming speed relates linearly to tail-beat frequency, while amplitude remains an independent variable. Fish control their propulsion primarily through stiffness modulation, which shifts resonance frequency and thereby changes amplitude. At cruising speed, fish use remarkably few anterior muscles while the rest of the body remains passive, explaining their extraordinary endurance. The discussion of a dead fish swimming upstream in George Lauder's lab at Harvard illustrates how body morphology alone can generate propulsion in periodic turbulent environments, a striking example of morphological computation.Kruusmaa introduces the concept of inverse biomimetics, using robotic fish as tools for biological discovery rather than just engineering products. Her work on artificial lateral line sensors demonstrates this approach: by selectively disabling parts of a robot's sensory array, researchers can generate hypotheses about biological function that are difficult or impossible to test in living fish. The lateral line's dual modality, measuring both flow velocity and pressure, enables fish to build complex hydrodynamic maps of their environment, a capability roboticists have barely begun to explore.

What if consciousness isn't generated by the brain at all, but is a way of describing how organisms interact with the world? Kevin O'Regan presents a radical sensorimotor theory that dissolves the hard problem of consciousness using the same conceptual trick that demystified life itself.Subscribe for more from the Convergent Science Network podcast series.O'Regan argues that searching for neural correlates of consciousness leads to an infinite regress: even if we found the exact neurons responsible for the feel of redness, we could always ask what makes those neurons produce red rather than green. His solution borrows from the history of biology, where vitalism was abandoned once scientists recognized that life is not a substance but a description of how organisms interact with their environment. Similarly, he proposes that feel is not something generated inside the brain but a characterization of the sensorimotor laws governing an organism's engagement with the world.The interview systematically addresses the three classical mysteries of qualia. Ineffability arises naturally because the low-level sensorimotor details constituting a feel are cognitively inaccessible, much like a whistler cannot describe their tongue position. The structure of feels, why red resembles pink more than green, falls out of the objective, measurable differences in sensorimotor laws governing interactions with colored surfaces. And sensory presence, the reason vision feels different from proprioception, relates to the richness and bodily engagement of the sensorimotor contingencies involved.O'Regan and interviewer Paul Verschure probe the relationship between this framework and Gibson's affordances, exploring whether qualia might be understood as the subjective dimension of affordance relationships. They examine how the sensorimotor approach partially overcomes interpersonal ineffability by grounding feel in observable behavior, and whether contortionists might experience richer tactile qualia due to finer motor control. The discussion culminates in the provocative claim that a sufficiently complex robot like the Terminator would genuinely feel pain, not because of any special ingredient, but because it would interact with the world in the ways we call feeling.

Can algorithmic control ever match the adaptability of a lobster navigating the ocean floor? Neuroscientist and roboticist Joseph Ayers reveals why DARPA abandoned traditional approaches and how chaos-based neural controllers are reshaping biomimetic robotics.Subscribe for more from the Convergent Science Network podcast series.In this episode, Ayers explains why conventional algorithmic robot control fails in unpredictable environments. Drawing on decades of studying lobster neurophysiology, he describes how animals use chaotic variations in their neural networks to escape situations no programmer could anticipate. The fundamental problem: you cannot pre-program escape strategies for every possible scenario an autonomous robot might encounter in the real world.Ayers walks through four generations of robotic lobsters built since 1998, each informed by biological discoveries. The latest generation replaces state machines with true central pattern generators built from discrete-time map-based neurons developed by Nikolai Rukov. These phenomenological neuron models capture spiking, bursting, and chaotic dynamics using just two control parameters, enabling hundreds of neurons and synapses to run on a single DSP chip in real time. The coordination between six walking legs emerges from governing and governed oscillators maintaining proper phase relationships.The conversation explores how building robots reveals gaps in biological knowledge. Ayers describes discovering that lobsters likely rely on simple bump sensing rather than sophisticated joint proprioception, and how accelerometry-based comparisons between expected and actual movement patterns can detect when the robot is stuck. He details the sensory architecture of the lobster brain, from Wiersma's classification of visual interneurons to the layered reflex systems that process optical flow, hydrodynamic flow, and obstacle contact. The discussion reveals how the robot-biology feedback loop generates new hypotheses about corollary discharge and motor control that can be tested in living animals.

Why does understanding the body matter as much as understanding the brain , and how do soft, squishy biomechanics simplify the control problems that nervous systems must solve? Hillel Chiel extracts general principles from tongues, worms, and sea slugs. Subscribe for more from the Convergent Science Network podcast series. Hillel Chiel's research program rests on a foundational claim: evolution selects not for brains or bodies in isolation, but for the coupled dynamical system of brain, body, and environment. This means that understanding neural control without understanding biomechanics is like studying software without knowing the hardware it runs on. His work systematically demonstrates how mechanical properties of soft tissues constrain and simplify the control problems that nervous systems face , often dramatically. The tongue provides a vivid entry point. Modeled as a muscular hydrostat (a "hot dog in a bun" of longitudinal and circumferential muscles), the tongue's geometry creates a massive mechanical advantage for the longitudinal muscle when the tongue is extended. A simple pulse of neural activation produces rapid shortening that the circumferential muscle cannot resist until the tongue is already retracted. The control implication is striking: for a single lapping motion, the nervous system can effectively ignore one of the two muscle groups. This simplification is invisible without biomechanical analysis and would never be predicted from neural recordings alone. Chiel then scales up to peristaltic locomotion, challenging the standard view that it is slow and energetically wasteful. His mathematical analysis of continuous (rather than segmental) peristaltic waves shows that, properly configured, the center of mass can maintain constant velocity without depending on external friction , meaning the energy costs come from internal tissue properties rather than ground contact losses. A one-meter robot built on this principle moves fast enough that you have to walk briskly to keep up. The Aplysia feeding system illustrates the principle that what a muscle does depends on its mechanical context. As the geometry of the feeding apparatus changes during a bite, swallow, or rejection, the same muscle can switch from protraction to retraction. This means that multifunctional behavior arises not from dedicated muscles for each action but from changing coalitions of muscles recruited according to the current biomechanical state. Chiel frames this as a general principle: the nervous system exploits context-dependent mechanics to achieve behavioral flexibility with minimal rewiring, ganging degrees of freedom together for simple movements and fractionating them when precision is needed.

How does a lobster find food in a turbulent ocean where chemical signals vanish for minutes at a time , and what can a robot lobster teach us about the strategies that work and fail? Frank Grasso explores the neuroscience of olfactory search through biomimetic robotics. Subscribe for more from the Convergent Science Network podcast series. Frank Grasso studies "crunchies and squishies", lobsters and octopuses, not because they resemble us, but precisely because they do not. These animals face the same physical challenges as vertebrates but solve them with completely different brain and body architectures, revealing the true design space for adaptive behavior. His laboratory at Brooklyn College builds robot models of these animals, tests them under identical conditions to the real organisms, and uses the animal's performance as a yardstick for evaluating hypotheses about neural control. The lobster work centers on olfactory search in turbulent water. Seventy percent of the lobster brain is dedicated to olfactory processing, and its chemical sensors are extraordinarily sensitive , capable of detecting concentrations equivalent to a teaspoon of rose water dissolved in Lake Champlain. But the real challenge is not sensitivity; it is intermittency. Even directly within a chemical plume, sensors may detect nothing for minutes at a time, punctuated by staccato bursts of odor pulses. Grasso's team measured these environments directly and found that the temporal structure of the signal, its rhythm and patterning, carries spatial information that pure concentration measurements cannot provide. When robot lobsters equipped with biologically matched sensors were tested using algorithms that biologists had proposed for a century, the robots failed dramatically. Simple chemotaxis strategies that work in smooth gradients collapse in turbulent plumes. Adding flow-sensing information improved performance, but only in specific regions of the plume. The robots revealed that different distances from the source present qualitatively different information landscapes, requiring different strategies , a finding that would have been difficult to establish from animal observation alone. The octopus work addresses a different frontier: controlling a body with no hard parts. Octopus arms are muscular hydrostats , enclosed bags of water that reshape themselves through muscle-against-muscle contraction, with no skeleton at all. Three-fifths of the octopus's half-billion neurons reside outside the central brain, distributed in a brachial plexus that may allow arms to negotiate with each other rather than requiring centralized executive control. Combined with the animal's exceptional learning abilities , possibly an adaptation to both a complex body and exponential growth from milligrams to 90 kilograms in a single year , the octopus represents a radically different solution to the problem of embodied intelligence.

What if the next generation of robot muscles were made of rubber, driven by static electricity, and could sense their own deformation? Federico Carpi introduces dielectric elastomer actuators , soft, lightweight, and already shipping in consumer electronics. Subscribe for more from the Convergent Science Network podcast series. Federico Carpi makes the case that conventional electric motors are fundamentally mismatched to the needs of robots that must interact closely with humans. They are rigid, noisy, energy-hungry, and made of materials nothing like biological tissue. His alternative: electroactive polymers, specifically dielectric elastomers , essentially sheets of insulating rubber sandwiched between compliant electrodes. When voltage is applied, electrostatic forces squeeze the rubber, causing it to expand laterally. The principle is pure Maxwell stress, not piezoelectricity, and the resulting actuators are soft, silent, lightweight, and remarkably versatile. What makes these materials particularly compelling is their intrinsic dual function as both actuator and sensor. Because the device is fundamentally a deformable capacitor, reading its capacitance during operation provides continuous information about its deformation state , no separate sensor required. This mirrors biological muscle, where actuation and proprioception are integrated in the same tissue. Carpi's group has demonstrated stacked actuators for larger displacements, membrane actuators, bubble actuators, and linear actuators, all from the same basic material platform. The technology has already reached the consumer market. A major mobile phone manufacturer has replaced conventional vibration motors with dielectric elastomer films , thinner, lighter, and more power-efficient because capacitive loads draw minimal current despite requiring kilovolt-range voltages. Carpi addresses the voltage concern directly: while one kilovolt sounds alarming, the currents involved are tiny and the energy stored is comparable to the static shock from a car door. Compact voltage multipliers a few cubic millimeters in size handle the conversion from battery voltage. Four application areas stand out. Variable-stiffness rehabilitation devices can provide customized resistance for post-stroke hand therapy. Energy harvesting systems can convert ocean wave motion into electricity at potentially lower cost than any competing technology. Haptic displays , including a Braille reader that could enable full-page tactile output for blind users , exploit the material's ability to create programmable surface textures. And biomimetic tunable lenses, inspired by the human eye's ciliary muscle, can change focal length by deforming fluid-filled membranes, with prosthetic eye applications on the horizon.

Is the neocortex really uniform , and do feedforward and feedback connections mean what we think they mean? Kathy Rockland challenges foundational assumptions about cortical organization with evidence that structure and function are more discrepant than textbooks suggest. Subscribe for more from the Convergent Science Network podcast series. Kathy Rockland is a neuroanatomist who has spent her career examining cortical connectivity at a level of detail that most theorists and imagers never encounter. In this interview, she delivers a series of provocative challenges to standard assumptions about how the neocortex is organized. Her first point is counterintuitive: the most striking property of cortical axons is not their specificity but their divergence. When you trace individual axons and examine their collateralization, the word that comes to mind is distributed , connections fan out broadly before any question of specificity arises. Rockland then dismantles the textbook story of ocular dominance columns. The functional columns are unambiguously about 500 microns in diameter, but the anatomical inputs that supposedly create them, thalamocortical axons from the LGN, come in at wildly different scales: parvocellular arbors at 250 microns, magnocellular clusters larger with multiple foci, layer 4A inputs at 100 microns, and layer 1 projections that are highly divergent. Some operation within intrinsic cortical circuitry must be converging these mismatched inputs into the 500-micron functional unit. This means that thalamocortical connections are not the basis of ocular dominance columns in any simple sense , the cortex itself is doing something essential to create the functional organization we observe. The conversation turns to feedforward and feedback pathways, which Rockland argues carry misleading temporal assumptions. She proposes reframing these as layer-4-biased connections and layer-1-biased connections, respectively , a description that captures the anatomical reality without implying a sequential relay. In rodents, the laminar scheme breaks down substantially, and even in primates there are abundant exceptions. Rockland also notes that the supposed unimodality of primary sensory areas is threshold-dependent: lower your detection threshold and cross-modal inputs appear even in monkey V1 and V2. Her overarching message is that structure-function correlation, long treated as a guiding principle, is more often the exception than the rule. The brain works, but not in the way our simplified models suggest.

Why do bacteria have more elegant network architecture than the internet , and what can both teach us about building robust, evolvable systems? John Doyle unpacks the universal design principle of layered constraints that biology and technology share. Subscribe for more from the Convergent Science Network podcast series. John Doyle is a control theorist and mathematician who found himself drawn to biology by a simple observation: the bacterial biosphere has one of the most robust and evolvable architectures on Earth. It evolved into us, yet continues to adapt with remarkable speed on every timescale , rearranging protein networks in seconds, swapping genes across species over generations. Doyle argues that this dual capacity for rapid robustness and rapid evolvability stems from a shared architectural principle: layered constraints that deconstrain. The concept, borrowed from biologists Gerhardt and Kirschner, holds that a few wisely chosen constraints , like ATP as a universal energy carrier or TCP/IP as a packet protocol , create platforms that enable enormous flexibility above them. In bacteria, core metabolic protocols have persisted for billions of years, yet they enable wildly dynamic responses to environmental challenges. In technology, operating systems sit between hardware and applications, enabling the plug-and-play modularity we take for granted. Doyle argues that layering is the highest-level expression of modularity, and that much of the scientific literature on modularity misses this point by focusing on component-level decomposition rather than the architectural constraints that make modularity possible. The interview draws a sobering contrast between biological and engineered systems. While bacterial biochemistry appears spectacularly well-designed from an engineering perspective, refined over billions of years of selection, human-built large-scale systems are profoundly unsustainable. Doyle is blunt: our energy, transportation, water, and food networks have recognizable design flaws, and the interplay between technology, markets, and policy is the least understood system of all. He uses a clothing metaphor to make architecture accessible: garments have both an inner-to-outer layering (skin layer, insulation, weather protection) and a compositional layering (fiber to yarn to cloth to garment), illustrating how different dimensions of constraint combine synergistically. Doyle also highlights a critical lesson from both biology and the internet: if you make a mistake in a core protocol and build extensively on top of it, correction becomes nearly impossible. The internet's early design choices, made by operating systems engineers who won a historical battle against information theorists, are now deeply embedded , brilliant in some respects, flawed in others, and extraordinarily difficult to change.

How do flowers deceive insects into pollinating them , and what does this reveal about how olfactory systems encode meaning? Bill Hansson explores the evolutionary arms race between plants and pollinators through the lens of insect chemosensory neuroscience. Subscribe for more from the Convergent Science Network podcast series. Bill Hansson studies olfaction in insects, and his entry point is one of nature's most elaborate deceptions. One-third of all orchid species are deceptive , they attract pollinators without offering nectar rewards, instead mimicking the chemical signatures of food, mates, or egg-laying sites with extraordinary precision. Hansson describes flowers that replicate the individual odor variation of female bees so accurately that males never learn to avoid them, and Mediterranean lilies that mimic both the volatile chemistry and the elevated temperature of rotting flesh to lure egg-laying flies into their chambers. These deceptive systems serve as powerful experimental tools. Because evolutionary pressure demands that the mimicry be nearly perfect, deceptive flowers effectively reveal which chemical features matter most to the insect brain. Hansson's laboratory uses this insight in reverse: by identifying the behaviorally relevant compounds through the deception, they can probe the olfactory system with precisely the stimuli it evolved to detect. The approach has been transformed by advances in single-neuron electrophysiology and optical imaging of the antennal lobe , the insect brain's first olfactory processing center. A surprising finding emerges from comparing input and output patterns in the Drosophila antennal lobe. At the receptor neuron level, there is no clear clustering of activation patterns by behavioral valence , attractive and repulsive odors look similar. But at the projection neuron output, attractive and repulsive patterns separate cleanly. This suggests the antennal lobe performs a valence-sorting operation, not just odor discrimination, before information even reaches the mushroom bodies traditionally associated with learning and memory. Hansson speculates that this early valence coding may serve the direct pathway to the lateral horn, which bypasses the mushroom bodies entirely and may mediate reflexive behavioral responses. The interview also examines a remarkable case of evolutionary specialization in Drosophila sechellia, a species that feeds exclusively on a single toxic fruit. This fly has sacrificed several receptor types used by its generalist relatives and massively expanded both the peripheral neurons and the central brain regions dedicated to detecting its host fruit , achieving detection thresholds rivaling moth pheromone systems at picogram concentrations.

How did a brain system for grasping objects become the foundation for human language? Michael Arbib traces the evolutionary path from mirror neurons to speech, arguing that schema theory provides the missing link between neural circuits and cognitive architecture. Subscribe for more from the Convergent Science Network podcast series. Michael Arbib has spent decades developing schema theory , a framework for decomposing complex behaviors into interacting functional units that can be mapped onto neural circuits. In this interview, he explains how this approach bridges the gap between high-level cognitive descriptions and low-level neural implementations, using two case studies: the visual control of hand movements and the evolution of language. The story begins with the premotor cortex, where Arbib's collaborator Giacomo Rizzolatti discovered mirror neurons , cells active both when a monkey performs a hand action and when it observes the same action performed by another. Brain imaging revealed that the human homologue of this mirror region overlaps with Broca's area, traditionally considered a speech center. This anatomical coincidence opened a research program connecting manual action to linguistic communication. Arbib outlines eleven evolutionary steps from our common ancestor with monkeys to the language-ready human brain, each representing what he calls a "small miracle" , a plausible transition requiring only modest genetic changes. The key transitions include: extending action recognition to imitation of novel actions, developing pantomime from practical object manipulation, conventionalizing gestures through social interaction, and finally recruiting the vocal apparatus for proto-speech. Arbib emphasizes that language likely did not evolve as a single package but was gradually discovered by human cultures exploiting brain capacities that evolved for other purposes. Sign language demonstrates that the linguistic capacity is not inherently vocal , it is a general-purpose system for structured communication. Schema theory serves as the computational backbone of this framework. Arbib positions schemas as intermediate-level descriptions, analogous to high-level programming languages, that capture the functional decomposition of behavior without committing to specific neural implementations. But unlike purely abstract computational theories, schemas are meant to be iteratively refined against neurophysiological data, creating a loop between cognitive-level hypotheses and circuit-level constraints. Arbib insists on causal completeness: a model must account for the full chain from sensory input through internal processing to behavioral output, not just correlate with isolated neural recordings.

How does evolution build brains without a blueprint? Terence Deacon reveals how self-organizing developmental processes, constrained by diffusible molecular signals, generate the neural architecture that natural selection then sculpts through competition and functional use. Subscribe for more from the Convergent Science Network podcast series. Terence Deacon brings the evo-devo perspective to brain science, arguing that understanding how brains are built during development is essential to understanding how they evolved. The core insight is that evolution does not modify adult brains directly , it modifies developmental programs. Since development relies heavily on self-organizing processes at every level, from gene regulatory networks to cell migration to axon guidance, the space of possible evolutionary changes is profoundly constrained by what development can produce. Deacon describes a two-phase process in brain construction. Early development is remarkably conserved across vertebrates: a neural tube divides into segments, segments differentiate into compartments, and generic form-production mechanisms generate a variety of circuits. This phase is so similar across species that embryonic brains of fish, birds, and mammals are nearly indistinguishable. The second phase involves selection: signals from sense organs, muscular systems, and inter-regional competition sculpt the generic architecture into species-specific functional circuits. Connections that are functionally validated persist; others are eliminated. The interview draws on Deacon's cross-species transplantation experiments, which produced striking results. Cortical cells transplanted anywhere in the brain grow axons only to targets appropriate for cortex , even in adult brains, even across species as different as pigs and rats. This means that molecular guidance cues persist throughout life, long after the developmental period when they were originally needed. Deacon suggests these cues serve ongoing plasticity and local synaptic maintenance rather than being mere developmental leftovers. He also describes clinical applications: pig fetal dopamine cells transplanted into Parkinsonian rats and eventually human patients found their targets, formed functional cross-species synapses, and produced measurable clinical improvement. Deacon challenges the common assumption that genes directly specify brain architecture. Instead, he describes cascading self-organization: genes regulate each other in network patterns, producing diffusible signals that create concentration gradients, which in turn activate or silence genes in neighboring cells. Even finger formation relies on this interplay , cells between digits are instructed to die by the intersection of multiple diffusion fields, not by a gene that says "build a finger here."

How do we move beyond "this brain region lights up" to genuinely understanding how neural circuits compute? Allard Roebroek argues that the future lies in merging bottom-up computational models with top-down neuroimaging analysis , and that neither community can succeed alone. Subscribe for more from the Convergent Science Network podcast series. Allard Roebroek tackles a fundamental tension in neuroimaging: the field generates gigabytes of whole-brain data per minute, yet most analyses reduce this richness to statements about which regions activate during which tasks. He distinguishes two modeling traditions that have developed largely in isolation. Bottom-up modelers build biophysically inspired simulations of neural circuits, from spiking networks to hemodynamic coupling, but face a crippling indeterminacy problem: infinitely many models can reproduce the same behavioral data. Top-down modelers invert observation models to go from fMRI or EEG signals back to inferred neural activity, but typically work with only a handful of pre-selected brain regions and stop at causal connectivity without asking what computations those regions perform. Roebroek's vision is to unite these approaches. He advocates for models that simultaneously perform the task (as bottom-up models do), are biophysically grounded, and are accountable to whole-brain neuroimaging data (as top-down models aspire to be). This triple requirement has not yet been achieved, but he argues it is the only path toward models that are both computationally meaningful and empirically constrained. The whole-brain coverage of fMRI provides a unique advantage over electrophysiology , not in spatial or temporal resolution, but in the ability to observe the entire system at once without invasive intervention. The interview engages seriously with criticisms of this program. Can correlation-based neuroimaging data really constrain causal models? Roebroek acknowledges that causality requires assumptions beyond correlation, but argues that computational models themselves provide exactly those assumptions , transforming observed dependencies into mechanistic explanations. He also confronts the common practice of restricting analyses to regions of interest, which discards the very whole-brain information that makes neuroimaging valuable, and calls for models that encompass at least all cortical regions plausibly involved in a given task.

What if patients diagnosed as vegetative are actually conscious but trapped , unable to signal their awareness through any behavioral channel? Adrian Owen describes how fMRI and EEG are revealing hidden minds and opening new paths to communication. Subscribe for more from the Convergent Science Network podcast series. Adrian Owen's research confronts one of the most unsettling problems in clinical neuroscience: the possibility that a significant minority of patients diagnosed as vegetative are in fact aware, experiencing the world much as we do, but completely unable to demonstrate it. Using fMRI, Owen and his team have identified patients who can reliably modulate their brain activity on command , imagining playing tennis to activate premotor cortex, or imagining navigating their home to activate parahippocampal regions , despite showing no behavioral signs of consciousness whatsoever. The breakthrough came with the realization that active imagery tasks, unlike passive sensory stimulation, cannot be performed automatically. When a healthy volunteer is told to imagine playing tennis and deliberately chooses not to, no premotor activation appears. This volitional quality is what distinguishes Owen's paradigm from earlier fMRI studies that showed brain responses in vegetative patients but could not rule out reflexive processing. The tennis task proved remarkably robust , it works in every healthy subject tested, likely because any version of tennis involves vigorous arm movements that reliably engage premotor cortex, regardless of individual differences in how people imagine the game. Owen has since used this approach to establish yes-or-no communication with a patient previously assumed to be in a vegetative state. By assigning tennis imagery to "yes" and spatial navigation imagery to "no," the patient answered autobiographical questions correctly, confirming not only awareness but access to long-term memory and language comprehension. The team is now working to transfer this technology from expensive, immobile fMRI scanners to portable EEG systems , a transition complicated by the fact that many of these patients have damaged or surgically altered skulls that disrupt standard EEG source localization. The conversation also probes deeper questions about the taxonomy of consciousness disorders. Owen suggests that the distinction between vegetative and minimally conscious states may be less about depth of consciousness and more about temporal intermittency , patients may cycle in and out of awareness rather than occupying a fixed intermediate state. He argues that the practical goal of building communication channels for non-responsive patients should take priority over philosophical debates about the nature of consciousness itself.

Why is the ability to hold something in mind, even briefly, the gateway to flexible cognition? Xiao-Jing Wang explains how attractor dynamics and slow synaptic reverberation in prefrontal cortex give rise to both working memory and decision-making. Subscribe for more from the Convergent Science Network podcast series. Xiao-Jing Wang begins with a deceptively simple argument: without the capacity to maintain information in the absence of direct sensory input, an organism is enslaved to its environment, reduced to reflexive responses. Working memory , sustained neural activity that bridges the gap between stimulus and action , is therefore the foundation of cognitive flexibility. Drawing on decades of lesion studies, single-neuron recordings, and computational modeling, Wang makes the case that prefrontal cortex is uniquely equipped for this role, thanks to its dense recurrent excitatory connections and distinctive neuromodulatory environment. The interview dives deep into the mechanics of attractor networks, which Wang uses as the theoretical framework for understanding prefrontal dynamics. He is careful to demystify the concept: attractor states are simply relatively stable states, not rigid black holes. What makes prefrontal cortex special is not persistence per se, even oculomotor circuits show persistent activity, but the capacity to maintain multiple stable states simultaneously and switch between them with brief inputs. This multiplicity is what a working memory system requires, and it emerges naturally from the nonlinear dynamics of strongly recurrent circuits. A key surprise from Wang's modeling work is that the reverberation sustaining working memory must be slow, mediated primarily by NMDA receptors rather than fast AMPA transmission. This was not a design choice but a computational necessity: fast positive feedback makes the network explosively unstable, while slow reverberation provides both stable memory states and the gradual ramping activity observed during decision-making. The same circuit architecture that holds items in working memory also integrates evidence over time, producing the reaction-time signatures seen in prefrontal recordings during perceptual decision tasks. Wang also addresses the frontier challenges: extending local circuit models to large-scale brain systems, understanding how mixed selectivity in prefrontal neurons supports combinatorial coding of sensory, rule, and motor information, and reconciling the role of neural oscillations and correlations with the stochastic firing of individual neurons. His vision is one of building blocks , elementary computational mechanisms that can be composed into increasingly realistic models of cognition.

Can neuroscience tell us what consciousness really is , even when introspection and behavior fall short? Viktor Lamme argues that recurrent neural processing, not global workspace activation, is the fundamental ingredient of conscious experience. Subscribe for more from the Convergent Science Network podcast series. Viktor Lamme opens with a challenge to the dominant paradigm in consciousness research: if we cannot reliably know what we are conscious of at any given moment, then searching for neural correlates of consciousness is fundamentally misguided. Instead, he proposes building a definition of consciousness from neuroscientific evidence itself , using neural arguments rather than behavioral reports to determine when and where conscious experience occurs. At the core of his theory are four stages of cortical processing. Stages one and two involve feedforward activation, shallow or deep, that can reach prefrontal cortex and trigger cognitive functions like attention and inhibitory control, yet remain entirely unconscious. Stages three and four involve recurrent or reentrant processing, where higher areas feed back into lower areas. Lamme's central claim is that recurrent processing is both necessary and sufficient for consciousness. Even localized recurrence between early visual areas produces a conscious percept, albeit a primitive one, while widespread recurrence incorporating frontoparietal networks adds reportability and cognitive access without adding consciousness itself. This directly challenges global workspace theory, which holds that prefrontal-parietal broadcasting is essential for conscious experience. Lamme argues that this conflates consciousness with attention, cognitive control, and reportability. He points to evidence that prefrontal activation can occur without consciousness and that consciousness can occur without prefrontal involvement. The distinction matters clinically and scientifically: if consciousness requires only local recurrence, then patients and experimental subjects may have rich conscious experiences that they simply cannot report. The interview takes a molecular turn when Lamme describes experiments showing that recurrent signals in monkey visual cortex depend on NMDA receptors, while feedforward signals rely on AMPA receptors. This dissociation suggests that conscious processing may be uniquely linked to synaptic plasticity , raising the provocative prediction that there is no such thing as truly unconscious learning. Every instance of learning in the literature that Lamme has examined involves conditions where local recurrent processing could plausibly be occurring, even if global reportability is absent.

What if the cerebellum works less like a learning machine and more like an interrupt handler , resetting circuits and gating sensory information depending on what the animal is doing? Sam Wang shares how advanced optical imaging is rewriting our understanding of cerebellar function. Subscribe for more from the Convergent Science Network podcast series. Sam Wang came to neuroscience from physics, drawn to the cerebellum by its deceptively simple architecture: a small number of cell types arranged in a circuit that seemed ripe for theoretical analysis. In this interview, he describes how his laboratory's optical imaging methods have revealed surprising dynamics in the climbing fiber system , the slow, one-hertz input pathway from the inferior olive that has long puzzled researchers. Wang reframes these climbing fiber signals as interrupt signals that can simultaneously reset ongoing cerebellar processing in real time and drive long-term synaptic plasticity. The key insight comes from synchrony. Individual climbing fibers fire so rarely that extracting meaning from their timing alone is a hard coding problem. But when populations of olivary neurons fire together, coupled by gap junctions, they produce what Wang calls "chords" across many Purkinje cells simultaneously. These synchronous events can be detected by the deep cerebellar nuclei as special signals, distinct from the background wash of simple spikes. Wang uses a musical metaphor: asynchronous firing is like random piano keys, while synchrony is like a chord that stands out against the noise. Perhaps the most striking finding is a gating phenomenon observed in awake, behaving mice. When a mouse is resting, climbing fiber populations respond robustly to external stimuli like air puffs or sounds. But when the animal begins walking, the same population switches to self-generated synchronous events and becomes insensitive to external input. This suggests a context-dependent gating mechanism, analogous to "don't talk to me, I'm tying my shoes", where the cerebellum dynamically routes either external or internal signals depending on behavioral state. Wang is candid about the limits of current cerebellar theory. While frameworks from control engineering, forward models, inverse models, adaptive filters, provide useful conceptual scaffolding, he suspects many will prove wrong when tested against well-designed experiments. His laboratory is pushing toward better temporal resolution in imaging, genetically targeted indicators, and optogenetic perturbation to move from observation to causal manipulation of these circuits.

What happens when engineered systems become too complex for humans to understand, let alone control? Riccardo Sanz argues that the path forward requires machines capable of controlling themselves , and that this leads, perhaps inevitably, toward machine self-awareness. Subscribe for more from the Convergent Science Network podcast series. Riccardo Sanz approaches consciousness not from philosophy or neuroscience, but from the hard edge of control engineering. In this interview, he explains why traditional control theory breaks down when the controller itself becomes so complex that it can fail in ways no human operator can diagnose. Modern countrywide electrical grids, flight control systems, and computing infrastructures already exceed human comprehension during failure states , leading to blackouts, crashes, and cascading breakdowns. Sanz's provocative claim is that the only scalable solution is to give these systems the capacity to model and manage themselves. This is not, he insists, an attempt to mimic human consciousness. Instead, his research group arrived at concepts of self-awareness and self-modeling from purely technical requirements for robust, adaptive control. The convergence with consciousness research was discovered after the fact, when they found that the competences they needed, self-monitoring, self-repair, cognitive flexibility, overlapped with properties that consciousness researchers attribute to sentient systems. The distinction matters: Sanz argues that copying the human brain would reproduce its evolutionary limitations, whereas extracting the underlying principles of self-awareness could yield systems that far exceed human capabilities in speed and information integration. The conversation probes the risk of infinite regress , if a controller needs a meta-controller, what controls that? Sanz proposes that each successive layer of self-representation compresses complexity, collapsing into increasingly compact models until the system converges on a unified self-description. He draws parallels to industrial process control, where hierarchies of control loops ultimately reduce to a single variable like profitability, but notes that current systems lack the self-awareness to handle their own failures. On the question of existential risk from superintelligent machines, Sanz is sanguine. He believes that by the time engineering reaches the sophistication needed to create deeply self-aware systems, the technology for bounding their behavior will be equally mature. His core message is a call for rigor: the fragmentation of control engineering, neuroscience, and philosophy into separate communities with incompatible vocabularies is the real barrier to progress.

How do you map the wiring of an entire brain when neuroscience is simultaneously drowning in data and starving for the right kind? Partha Mitra explains why he left theoretical physics to build a whole-brain mesoscale connectome , and what it reveals about the gap between data richness and genuine understanding. Subscribe for more from the Convergent Science Network podcast series. In this episode, Partha Mitra describes the paradox at the heart of modern neuroscience: half a million abstracts published on PubMed each year, yet no comprehensive wiring diagram for any mammalian brain beyond C. elegans. Trained as a theoretical physicist, Mitra recounts how his growing humility toward the complexity of the brain drove him from abstract modeling to the lab bench, where he now leads an industrial-scale neuroanatomy project at Cold Spring Harbor Laboratory. His goal is to systematically map the mesoscale connectivity of the mouse brain , the level at which developmental programs lay down the architecture that sits between single synapses and whole-brain function. Mitra draws a compelling analogy to Google Earth: just as geographic data remained fragmented until a unifying spatial framework existed, neuroscience data lacks a scaffold on which to hang its heterogeneous findings. His project aims to provide that scaffold by injecting tracers across the entire mouse brain and building probabilistic maps of where axons from any given region project. He argues that this mesoscale is uniquely important because it is genomically patterned , shaped by developmental genes rather than purely by experience , making it the natural bridge between molecular biology and systems neuroscience. The conversation also tackles deep methodological questions. How much individual variability exists between brains of the same species, and can meaningful regularities still be extracted? Mitra hypothesizes that brains occupy a low-dimensional manifold of variation , constrained enough to reveal common design templates, yet variable enough to be scientifically interesting. He envisions comparative studies across species that could uncover conserved architectural principles shaped by convergent evolution, not just shared ancestry. Perhaps most striking is Mitra's philosophical evolution. He advocates what he calls "ontological monism and epistemological pluralism" , one physical reality, but multiple legitimate theoretical frameworks for understanding it. He cautions against the assumption that all theories must reduce to one another, and urges neuroscientists to take engineering perspectives more seriously as a source of insight into how evolved systems solve functional problems.

What if we've been studying brain activity for decades without actually knowing how the brain is wired? Neuroscientist Olaf Sporns introduced the concept of the connectome, a complete structural map of the human brain's network, and explains why understanding connectivity is the missing foundation beneath all of functional neuroscience. Subscribe for more from the Convergent Science Network podcast series. Olaf Sporns, one of the pioneers behind the human connectome concept, joins Paul Verschure at the BCBT summer school to explain why neuroscience needs a comprehensive wiring diagram of the brain. The connectome describes the full set of structural connections between brain regions or individual neurons , the network architecture that generates all the dynamic activity researchers have been measuring for years. The conversation addresses a fundamental gap in neuroscience: everyone records brain activity, but without knowing the underlying connectivity, we cannot explain where that activity comes from or how it is generated. Sporns traces the idea back to Ramon y Cajal while emphasizing that modern diffusion imaging now allows us to infer connectivity in living humans non-invasively for the first time. The discussion explores the relationship between structural connectivity and functional dynamics. Sporns argues that the connectome is not just an anatomical catalog but a framework for understanding how network architecture shapes cognition, behavior, and brain disorders. The challenge is scale , mapping individual neurons remains impossible in humans, but region-level connectivity is now within reach. Key topics include the methodological limitations of current imaging techniques, the difference between structural and functional connectivity, how network science tools from physics and mathematics apply to brain organization, and why the connectome project represents a shift from studying isolated brain regions to understanding the brain as an integrated network. Part of the Convergent Science Network podcast series from the BCBT Summer School.