Learning to see is teamwork
Seeing is more than light hitting the retina — it is the result of a finely tuned interplay between networks of neurons. A new study by researchers at the Frankfurt Institute for Advanced Studies (FIAS) and international partners shows how the brain learns to reliably process visual stimuli after the eyes open. With experience, incoming signals become more precise and increasingly aligned with internal circuits — a process that enables stable visual perception. These insights could inform advances in artificial intelligence (AI) and therapeutic approaches.
Learning to see is a complex interplay of neurons—and thus presumably a fundamental principle of the brain. Understanding this provides important foundations for research, therapies, and the development of artificial intelligence. (Image: Sigrid Trägenap, supported by NanaBanana)
Bernstein members involved: Sigrid Trägenap, Matthias Kaschube
How the brain learns to interpret the world visually has long puzzled researchers. Earlier studies suggested that neural responses at eye opening were immature and poorly coordinated. But new findings, led by scientists at the Max Planck Florida Institute for Neuroscience (MPFI) in close collaboration with FIAS, offer a different view: visual experience not only sharpens the input signals, but also aligns them with recurrent neural circuits. As a result, initially variable patterns give rise to a stable and consistent representation of the visual world.
The team in Florida studied this interplay in ferrets — an ideal model, as their eyes remain closed for weeks after birth. This delay allows researchers to observe how visual experience shapes the development of neural circuits with precision. Using a unique combination of modern techniques — simultaneous electrophysiology, calcium imaging, and whole-cell recordings — they captured the activity of individual neurons and entire networks at the same time.
For the first time, this approach enabled researchers to directly link single-cell activity with broader patterns across the brain — offering a multi-layered view of cortical development. Building on these data, researchers from the group of FIAS Senior Fellow Matthias Kaschube developed a computational model to disentangle the contributions of different biological processes. The findings show that clear visual perception only emerges when both the precision of incoming signals and their alignment with internal networks improve over time. This broad methodological framework — made possible through close international collaboration — provides a coherent picture of maturation, from individual neurons to entire networks and across cortical layers.
“Our results show that the brain doesn’t simply fine-tune existing structures when learning to see,” says FIAS doctoral researcher Sigrid Trägenap, who co-developed the model. “It actively adapts its internal networks to the world it perceives.” This adaptability, she adds, is one of the brain’s greatest strengths — and one that artificial intelligence has yet to match.
The researchers believe that this developmental process may reflect a fundamental principle of brain organization — one that extends far beyond the visual system. It could help explain how the brain achieves both precision and flexibility across different sensory modalities and cognitive functions. In addition to advancing basic research, the findings may inspire new approaches in artificial intelligence and inform medical therapies, such as rehabilitation strategies following stroke or other brain injuries.
