The Multiscale Problem

This essay reflects my view of the current limitations of neuroscience and my belief that a more rigorous epistemological approach, one that integrates information across multiple scales, may offer deeper insight into the field’s most important problems. It is a living document and does not exist in isolation from my broader thinking, though I have tried to keep the argument as self-contained as possible in explaining why the brain presents a fundamentally multiscale problem.

The pursuit of knowledge in scientific domains exists to order, arrange, and make sense of phenomena, transforming the chaotic and unknown aspects of nature into structured knowledge. The aim is pragmatic: to predict, control, and manipulate the world. Authority over a physical domain is the mechanism through which individual agents and collective societies instantiate change and drive progress. Our universe has provided an inexhaustible number of complex systems. As a result, we experience reality as almost mystical, even as science brings order to its depths.

Few systems challenge our understanding as profoundly as intelligent, arguably conscious biological entities. Brain is a system shaped by eons of evolutionary pressure, composed of a dense stack of interacting cellular processes, proteins, and genetic machinery. Semipermeable membranes regulate molecular flow in a soup of transmitters and intracellular communication processes; neurons come to form circuits; circuits form networks; and these networks come to form cognition, behavior, and emotion. Across these interacting layers, dynamics unfold continuously, and the interactions come to form what we recognize as mind, consciousness, and experience.

Yet, complexity across scales is not unique to neuroscience. Biological processes, such as cancer, resist simple explanation because their origins span multiple levels, from genetic to cellular to environmental. Historically, cancer treatments focused on eliminating rapidly dividing cells, prevention of their origination is beyond our current ability. This 'brute force' approach reveals the limitations of scientific understanding when knowledge appreciates but fails to address the system as a whole.

By necessity, we have fallen into equating useful assumptions with knowledge, squeezing understanding into a Procrustean Bed of sorts. We chop reality into convenient slices to fit our limitations. Physics claims particles, chemistry molecules, biology cells, psychology behavior, and sociology groups. Each level provides insight but remains incomplete in isolation. Layers are entangled: lower-level processes constrain higher levels, while higher-level organization influences lower-level dynamics.

The Multiscale Problem is the challenge of understanding how these levels and systems relate to, interact with, and generate one another. Brain represents one of, if not the most, complex biological systems, where understanding at any single level is inherently limited without understanding the system across scales.

Brain as a Multiscale Problem

The brain is composed of a vast number of cells and types, with some consensus (though recent discoveries have challenged this) on a fundamental computational unit: the neuron. These cells function with a degree of independence, their behavior determined by local environmental conditions and a stream of electrochemical inputs from proximal and distant neighbors. Collectively, these units are organized into a hierarchy of anatomical structures, each specializing in unique roles in physiological or cognitive functions.

Our grasp of these structures is unevenly distributed. We possess some robust understanding of the ‘primitive’ structures of the brain- medulla, pons, midbrain. These areas have been extensively mapped through a combination of comparative neuroanatomy and clinical lesion studies, revealing the pathways and networks underlying life-sustaining processes such as rhythmic respiration and sympathetic nervous system activation. Despite this, higher faculties of mind, the ‘executive’ aspects of cognition, remain elusive. The mechanism by which the brain manipulates tools, navigates language, and extracts meaning from ambiguous data represents a significant jump in complexity that simple hierarchies struggle to explain.

Historically, our primary window into this complexity has been ‘tragic serendipity’: focal brain damage. In these instances, a specific deficit in skill or function serves as reverse-engineered evidence that the damaged region was a critical node in the network responsible for that behavior. For figures like Broca, such lesions became catalysts for exploring functional localization, the idea that specific cognitive abilities could be tied to discrete cortical regions. Later, Brodmann extended this effort by partitioning the cortex based on cytoarchitectonic differences, creating a structural atlas that would later be interpreted in functional terms. While this provided a useful map of the brain’s ‘geography’, it also set the stage for the modern challenge of reconciling fixed anatomical borders with the fluid, multiscale dynamics we now know underlie cognition.

Technology has evolved, providing greater access to the brain and its behavior, though it is still admittedly limited. What once could only be accomplished through tragic serendipity is now available through transcranial magnetic stimulation. Structural-anatomical injuries can now be observed indirectly through imaging. Functional-behavioral activity can be observed through hemodynamic changes and electrophysiological activity. Even psychophysics has evolved to capture information previously unavailable with greater validity. Our understanding of neurons, their types, their behavior, and their modulatory effects on each other has evolved significantly.

Despite all our advancements, how we study the brain remains fragmented. Cognitive science, behavioral neuroscience, computational neuroscience, and affective neuroscience each describe the same brain using different languages and frameworks. We can capture neural activity in a single neuron using Neuropixels. Yet, the significance of that activity to a cortical column, anatomical region, or the whole brain is lost. Cognitive models may partly explain our intelligence and blind spots, but these measures fail to capture the computational processes required to implement them at the cellular level.

Take a system such as our primary visual cortex- the relationship between vision and behavior is evident in our experience, but not so clear within the brain. Decision-making, perception, maybe all things we know to experience have this quality of self-evidence without explainability. The levels between reality, perception, behavior, and experience cannot be cleanly separated without reducing a highly dimensional system to an oversimplified representation. Causality in these systems is dynamic: attention, regulated by cortex, alters neural firing; receptive fields constrain what may be attended to; learning reshapes circuits; and circuit architecture constrains learning. Perception itself appears to be a superposition of prior knowledge encoded in the brain and sensory input, reflecting the highly dynamic exchange across scales.

Scientific explanations of the brain have historically oscillated between two poles: reductionism and emergence. The reductionist approach seeks to explain complex phenomena entirely by decomposing them into their lower-level components- suggesting that if we understand the neuron, we eventually come to understand the mind. Conversely, the emergent perspective emphasizes that novel properties arise at higher levels of organization and cannot be predicted from individual parts in isolation.

Brain embodies and defies both extremes. While scientists have sought a physical basis for subjective consciousness, neural activity alone does not capture the full richness of cognition or subjective experience, and there is still no observable brain structure that serves as a definitive seat of consciousness or explains the nature of qualia. Yet, we must reconcile this with the biological fact that every aspect of our identity, memory, and experience exists as a manifestation of cellular form and interaction.

The more rigorous "gradient" view suggests that brain, consciousness, and intelligence are results of constrained emergence. In this view, higher-level properties are strictly bounded by physical phenomena, yet they are expressed as organizational patterns that transcend the individual cells themselves. By viewing the brain this way, we can describe the system without resorting to "handwaving" or untestable metaphysical claims. It allows us to treat the mind not as a ghost in the machine, but as the dynamic, multiscale result of a system that is both the sum of its parts and the novel complexity emerging from how those parts are woven together.

These interactions are occurring, whether we have a lens on them or not. Neuromodulatory systems can shift the state of entire circuits across vast anatomical distances. Neural oscillatory activity, ‘brain waves’, acts as a temporally sequenced coordination of activity across distributed populations of cells. Simultaneously, plasticity allows the environment to physically reshape connectivity over time, while attention dynamically gates which sensory signals are amplified or suppressed. These processes demonstrate that the relationship between scales is a two-way street of top-down and bottom-up causation, grounded in phenomenal interactions across spatial and temporal dimensions.

Outlook on Science

Approaching the Multiscale Problem of Neuroscience is a central challenge for the 21st century. Integrating data across scales will not immediately link neuron activity to conscious thought, but it defines the space between scales. With a sufficient grasp of these scales, we can refine our understanding, leading our models to higher resolution and enabling us to ask better questions.

Understanding more of how the neural, neuroanatomical, cognitive, and totality of the brain works defines our ability to intervene in challenges with mental health (psychiatry and psychotherapy), to understand and develop intelligent systems (artificial intelligence), to resolve health concerns of neurodegenerative diseases and cognitive deficits, of learning disorders, and of all things embedded in the brain. Humanity has landed on the moon, mapped the genome, and split the atom; yet the system through which we perceive and understand those very achievements remains largely a mystery. The central challenge of the 21st century is not merely to collect more data at each scale, but to decode the laws of interactions that allow these scales to generate a unified, conscious whole.