Sixty years ago, Harvard scientists David Hubel and Torsten Wiesel famously showed that altering visual input in young cats irreversibly changed the organization of neurons in the visual cortex, inspiring the concept of developmental critical periods—time windows in which experience was thought essential for refining brain connectivity. Now, in Nature Reviews Neuroscience (PDF), with Florian Engert and André Ferreira Castro we reexamine the prevailing assumption that patterned neuronal activity—whether stimulus-evoked or spontaneously generated—is necessary and instructive for shaping neural circuits. Conventional wisdom posits a two-phase model: an initial, genetically driven “coarse wiring” stage, followed by experience-dependent pruning and refinement of essential connectivity.
Yet this standard view falters in light of innate behaviors. Many animals exhibit remarkably complex problem-solving skills at birth—long before experience could modify their circuits. In our own experiments, we showed that larval zebrafish raised without any brain activity during development could still assemble functional neuronal pathways and execute visually guided behaviors at near-normal levels (once anesthesia was washed out). These findings call into question whether activity-dependent plasticity is indeed critical for building core sensorimotor networks.
Our Perspective proposes a resolution to this nature–nurture puzzle by distinguishing three “systems” of circuit formation. (1) Developmental Maturation, (2) Eureka Moments, and (3) Staying Tuned. We argue that the majority of the knowledge required for daily survival is realized genetically in System One, truly novel information from System Two is rarely essential for standard operations, and the plasticity of System Three mainly serves as a homeostatic stabilizer rather than a universal instructive force.
Why, then, does it feel as though everything is learned? One reason is that human infants complete their neurodevelopment after birth, whereas many other species are “born ready.” Another is our uniquely rich language and semantic memory, which magnifies the sense that all knowledge must be acquired by experience. Finally, the modern AI narrative of “learning from scratch”—despite evolution having pre-trained much of our circuitry—further entrenches the idea that all neural systems rely heavily on experience-driven rewiring.
In summary, we propose that most circuits are genetically hardwired (System One), while rapid learning (System Two) occurs infrequently but can be exceptionally potent. Meanwhile, continuous plasticity (System Three) primarily refines or stabilizes existing wiring. Crucially, our framework contextualizes—rather than dismisses—decades of developmental studies (including eye sutures, sensory deprivation, and genetic silencing) that have revealed important roles of neuronal activity. By questioning how much of our behavior is truly learned versus what is already specified by our genes, we hope to spark fresh inquiry into the interplay among innate wiring, plasticity, and behavior.
