Do They Remember? What Neuroscience Says About Infant Memory and the Voices in the NICU

One of the best parts of working in the NICU is the visits we receive from families who have graduated and come back with their growing, glowing babies. During these visits, I was always puzzled by the fact that some of them — especially the ones I worked more closely with — seemed to recognize my voice and would even award me with a smile. I do have the habit of chatting with my patients, and since I used to work every weekday in the mornings, those babies heard my voice a lot. But still: would it really be possible that they remembered it? This week I came across a study that made me think about this more seriously: https://www.science.org/doi/10.1126/science.adt7570

Published in Science in March 2025, Yates and colleagues scanned the brains of awake infants between 4 and 25 months of age while showing them photographs they had never seen before. Using functional MRI, they found that babies whose hippocampus — the brain region responsible for forming memories — was more active during that first exposure were more likely to recognize the image later. This pattern emerged reliably around 12 months of age. The surprising conclusion: infants do form individual memories. The reason we cannot recall our earliest years is not that memories were never made — it is that, as the brain matures, the mechanism for retrieving them stops working. Infantile amnesia, in this new framing, is a retrieval problem, not an encoding problem. The memories were there. We just lose access to them. This reframes something fundamental. The infant brain is not a passive, unrecording system. It is actively taking things in — and the hippocampus is already doing its job much earlier than we used to think.

And what about preterm babies (and their preterm brains)? Here is where it gets both more complex and more relevant to neonatology. Research in animal models shows that the basic cellular machinery for forming memories — including the synaptic plasticity mechanisms that underlie learning — is already functional in the hippocampus at the equivalent of the extreme preterm period in humans. The hardware is there, remarkably early. But it is also remarkably fragile. Even brief episodes of low oxygen, which are common in very preterm infants, can disrupt hippocampal maturation without causing the kind of visible brain injury we look for on imaging. The damage is functional and silent: changes in how neurons connect, how they communicate, and how efficiently they encode new information. Studies consistently show that very preterm infants have smaller hippocampal volumes by the time they reach term-equivalent age, and that these differences are associated with memory and learning difficulties that persist into school age.

So the preterm infant carries a double challenge: the universal retrieval barrier that affects all of us (infantile amnesia), plus a more specific vulnerability in the very encoding mechanisms that Yates showed are necessary for memories to form in the first place. But what about voice recognition? Here is the part that I find most compelling — and that may speak directly to what I have been observing in those follow-up visits. Not all memory depends on the hippocampus. Voice recognition and auditory familiarity are processed through circuits that develop earlier and are more resilient than the ones involved in episodic memory. The auditory cortex and surrounding regions mature substantially during the third trimester — precisely the period that preterm infants spend in the NICU. Recognizing something as familiar, without being able to explicitly recall when or where you encountered it, is a different and older form of memory, mediated by different brain structures. A baby who spent weeks hearing the same voice every morning is not retrieving an episode. They may simply be responding to something that feels, at some neural level, known. Safe. Familiar. That kind of recognition does not require an intact retrieval system. It may not even require conscious processing.

I am not suggesting that NICU babies consciously remember us. But the neuroscience does suggest that the voices in the room during those early weeks are part of an environment that a developing brain is actively processing — and that repeated, warm, familiar auditory input during a period of extraordinary neural plasticity is not inconsequential. The baby who smiles when they hear my voice at follow-up may not remember me in any way they could ever describe. But something in their brain might recognize that this particular voice was there — consistently, every morning — during one of the most intense experiences of their early life. That possibility alone feels like a good reason to keep talking to our patients.