That babies learn early is something every parent knows. But that this learning goes far beyond recognising a voice in the womb, mirroring a smile, or gradually gaining control over their own body can still be surprising.

Years ago, research by scholars such as Stanislas Dehaene already showed that babies can distinguish between quantities at a very young age. “More” and “less” are not empty concepts, even without language. Long before there is any talk of counting, let alone arithmetic, the brain already responds differently to different amounts.

A new study published in Nature Neuroscience takes this one step further. It is not only about how much, but also about what.

A large team of researchers led by Cliona O’Doherty used fMRI to study the brains of babies just two months old. Awake, which in itself is technically impressive. What they found is perhaps even more striking: at this age, babies already have rich visual categories in their brains. Their visual system distinguishes between different kinds of objects. Animate and inanimate. Large and small in the real world. Not a collection of loose impressions, but a structured organisation.

Until now, research in this area largely relied on looking behaviour. Babies look longer at novel stimuli than at familiar ones, and from that we infer that they notice differences. But looking at time tells us little about how those differences are organised in the brain. This study shows that such an organisation is already present very early on, even before it becomes clearly visible in behaviour.

Importantly, this is not a simple story of “first low-level perception, later meaning”. The classic idea that learning proceeds neatly from simple to complex does not hold here. In babies’ visual brains, features at multiple levels are already present very early. Not fully developed, but present. What follows is refinement, not a complete rebuild.

The researchers even compared infant brain patterns with artificial neural networks. Not because babies are little computers, but because such models help clarify what kind of information is being learned. The patterns found in infant brains resemble systems that learn from statistical regularities in visual input, without explicit labels. That fits well with how babies experience the world: no explanations, no instructions, just a continuous stream of comparisons.

Together with earlier work on quantities and numerical intuition, this shifts our understanding of learning once again. Learning does not begin with words, with school, or even with conscious understanding. It begins in a brain that actively structures the world from the very start.

Perhaps this is the biggest misconception that this kind of research undermines: the idea that development starts from a blank slate, regardless of the old nature–nurture debate. Babies do not come equipped with the knowledge adults do. But their brains are already organised in ways that make knowledge possible. And that, it turns out, begins even earlier than we thought.

Image: Cusack Lab, © 2026, CC BY-NC-ND 4.0 (Found here)