Learning to walk is not remembering the lesson
Your first years contain a strange mismatch: you learned an enormous amount, yet you probably cannot replay most of the experiences in which that learning happened. The mismatch becomes less mysterious once we stop treating memory as a single recording system.
A learned skill can guide behavior without bringing back a particular afternoon. General knowledge is different again: knowing what a spoon does does not require remembering the first spoon you saw. An autobiographical memory makes a more specific claim. It places an event in your own past, with people, a setting, and a sense that you were there.
Infantile or childhood amnesia mainly concerns the scarcity of those personal recollections from the earliest years. Adults commonly have few from before about age three or four, followed by a thinner set from early childhood than from later life. This is a broad pattern, not a deadline on which everyone’s memory system switches on.
How do you test a baby who cannot tell you?
Asking an infant to describe last Tuesday is a poor experiment. Researchers need behavior that can reveal memory without requiring a spoken account.
One approach is deferred imitation. An adult demonstrates actions with an object, then later offers the object again. Reproducing the demonstrated actions provides evidence that something about the earlier event was retained. Researchers can vary the delay, the objects, or the order of actions to investigate what survives.
Recognition tasks provide another route. Researchers present familiar and unfamiliar pictures and measure looking behavior. Neither method turns a baby into an adult witness. They do, however, make “babies cannot remember anything” a very difficult position to defend.
The distinction between the test and the claim matters. Remembering an action sequence after a delay is evidence for memory of an experience. It does not demonstrate that the same infant will later narrate that experience as part of a life story.
The hippocampus participates earlier than once assumed
The hippocampus works with other brain regions to help form memories linking the pieces of an event. The selected illustration turns those pieces into a puzzle: who was involved, where it happened, when, and what happened. It is a conceptual drawing, not an anatomical map or a claim that each piece occupies a separate compartment.
A 2025 study by Tristan Yates and colleagues used functional MRI while awake infants viewed photographs. Greater hippocampal activity during initial viewing was related to later memory-based looking behavior, beginning around one year of age. That finding supports the presence of an encoding mechanism during a period that is largely absent from adults’ autobiographical recall.
Three stages need to be distinguished:
- Encoding: forming a memory of an event.
- Retention: preserving a trace over time.
- Retrieval: accessing the retained information.
Evidence for infant encoding directs attention to the later stages. An entirely inactive hippocampus cannot explain the adult gap if this encoding mechanism is already operating in infancy.
It does not prove that every early experience was stored, or that any particular memory remains available decades later. The experiment measured an infant memory task, not an adult’s access to the same events.
Some childhood memories disappear while children grow
Adults looking back cannot reliably tell researchers everything they once remembered and then forgot. Prospective studies approach the problem from the other direction: establish an early memory first, then ask about it later.
Patricia Bauer and Marina Larkina studied conversations about events when children were three. Different subgroups were interviewed again at ages five through nine. The younger follow-up groups remembered at least 60% of the early events; the eight- and nine-year-olds remembered fewer than 40%.
Those results concern particular events discussed in a particular study. They are not a universal timetable or the percentage of all childhood experiences anyone retains. Their value is more direct: children had remembered events that became harder to retrieve years afterward. Explaining childhood amnesia therefore requires an account of forgetting, alongside improvements in the ability to form memories.
The developing network is changing as well. Connections between the hippocampus and other regions mature, and memory traces must be stabilized within that network. One hypothesis, supported largely by animal experiments, is that high early rates of new-neuron formation can alter circuits in ways that weaken access to existing memories. That is a possible contributing mechanism, not an established explanation for every missing human memory.
Do the memories survive somewhere?
Experiments in mice offer evidence that apparent forgetting can sometimes be a retrieval problem. Researchers can label neurons involved in an early experience and later activate them artificially. In a 2023 study, activating labeled infant engram cells restored memory-related behavior in mice after ordinary access to the memory had failed.
An engram is a physical memory trace, not a miniature film. The experiment shows that a trace can outlast the behavior normally used to detect it under those experimental conditions. It does not show that an adult human contains a complete, recoverable account of infancy.
It also supplies no method for recovering personal memories through conversation or imagination. Artificial activation of selected neurons in a controlled animal experiment is very different from recognizing a family story as something that feels familiar.
Words and a personal story add structure
Language, concepts of time, and a developing sense of self help children describe events as parts of their own history. These capacities develop together, rather than arriving as one finished package. Family conversations also give children practice in telling others what happened and how they experienced it.
Their contribution should not be overstated. In Bauer and Larkina’s study, overall maternal narrative style related to children’s contributions at age three, but did not predict later recall of the early events across the follow-up ages. Turning the conversational turn back to the child did relate to how much information children later reported. The evidence does not support a simple promise that enough storytelling preserves every memory.
Language cannot explain the whole phenomenon either: young rodents show rapid forgetting without having human language to develop. Biological changes and the construction of an autobiographical narrative address different parts of the problem.
If you picture an early birthday from a photograph, the picture and your knowledge that the birthday happened are not automatically evidence of firsthand recall. It is useful to distinguish what you independently remember from what you have been told, without assuming either that every early memory is false or that vividness proves its source.
The missing years were full of learning. What is missing is reliable access to many individual episodes, through a memory system that was still developing when those episodes occurred. Skills can remain, knowledge can remain, and some traces may remain, without a complete personal account remaining available.
For another example of information becoming difficult to retrieve, see why doorways can make you forget what you wanted. That short-term effect is a different problem from childhood amnesia, but it helps separate losing access from proving that information was erased.
