Here is a strange fact that we mostly do not find strange because it is so familiar. The period of your life in which you learned the most, by a very wide margin, is the period you cannot remember. You acquired a language, a body, a working model of physical causation and a working model of other people. Almost none of it is available to you as memory.
This is called childhood amnesia, and it has resisted a clean explanation for well over a century.
The easy explanation is wrong
The intuitive answer is that infants and toddlers do not form memories. This is straightforwardly false, and demonstrating it was one of the more elegant lines of work in developmental psychology.
Show a toddler a novel sequence of actions with objects. Come back weeks or months later, hand them the objects, and they reproduce the sequence. They have retained a specific episode over a long interval, and they can show you without needing to describe it.
Children of two and three can also talk about things that happened days or weeks ago, given the right prompts. They remember. What changes later is not the formation but the retrievability.
The indexing problem
One family of explanations concerns how memories are addressed. A memory has to be findable, and finding it means having a cue that leads to it.
Adults index episodes richly: by date, by place, by who was there, by what else was going on, by the narrative of a life. A young child has few of those. There is no sense of last spring, no house that was ours before this one, no calendar structure to hang anything on.
On this account the memories may be present and simply unaddressable, like a book in a library with no catalogue entry. It is consistent with the way early memories occasionally surface when a very specific cue, a smell, a room, arrives unexpectedly.
Language, and the way families talk about the past
A second explanation involves language, and it has some of the more interesting supporting evidence.
Memories that can be told can be rehearsed, and rehearsal maintains accessibility. Events that happened before a child had the language to narrate them may be less likely to be converted into the kind of durable, tellable episode that survives.
There is also a family effect. Families differ in how much they reminisce with young children, and in how they do it: some ask elaborative questions that build a narrative, others ask closed questions that check facts. Children in more elaborative families tend to produce richer and earlier autobiographical memories. Cross cultural comparisons point in the same direction, with earlier reported first memories in cultures that emphasise personal narrative.
This does not mean parents can install early memories by talking more. It suggests that the social practice of narrating a life is part of how autobiographical memory is built, which is a much more interesting idea than a technique.
You need a self to have a memory of yourself
Episodic memory is not merely a record of an event. It is a record of an event that happened to you. That requires a self to attach it to.
A sense of self as a continuous entity persisting through time develops through the second and third years, and the timing lines up suggestively with the boundary of childhood amnesia. On this account, memories laid down before there was a stable self to own them are not autobiographical memories at all, whatever else they may be.
Suggestive timing is not proof, and this magazine tries to be careful about that distinction. But the alignment is hard to ignore.
Neural turnover
A fourth line comes from biology. The hippocampus, central to forming episodic memories, is undergoing rapid change in early childhood, including high rates of new cell formation.
Work in animals suggests that high rates of new neuron formation in this region can disrupt previously stored memories while improving the capacity to form new ones. If something similar occurs in human infancy, early memories could be actively displaced by the very growth that makes learning so rapid.
This is an area where animal work is doing most of the causal lifting and human evidence is largely indirect, which is worth keeping in mind. It is a strong hypothesis rather than a settled account.
The boundary is a slope, not a wall
Adults asked for their earliest memory typically report something from the preschool years, with considerable variation between individuals and between cultures. Memories become steadily more available across the following years rather than switching on.
Two complications matter. First, people are poor judges of whether an early memory is genuine. Family photographs and repeatedly told stories generate vivid recollections of events the person did not encode, and there is no internal marker distinguishing the two. Second, reported earliest memories shift with how the question is asked.
So the boundary is soft, personal and partly constructed, which is characteristic of memory generally rather than special to childhood.
Why any of this matters
Partly because it is one of the genuinely puzzling facts about being human, and that is reason enough for a magazine like this one.
But it also corrects something. Parents sometimes conclude that experiences in the first years do not count because the child will not remember them. The inference does not hold. What is lost is the retrievable episode. What is retained is nearly everything else: language, expectations about people, motor skill, the shape of how a day goes.
Memory in the sense of being able to tell the story is one narrow product of a system that is doing far more than storing stories. For the neighbouring machinery, see working memory in plain words, and for the role of sleep in stabilising what has been learned, sleep and learning.