In this question two memories are important to separate: the procedural and the episodic. Procedural is acting and is not really a very good memory: it is the way we have done each of our actions the last time they took place. It is our spontaneous programming to act in one way rather than another. Our behaviors are not too repetitive because this programming suggests to us a certain number of alternatives, which the context helps to sort out. This is not a very good memory as it erases the old ways of acting to keep only the most used.
Episodic memory is more like the memory of a computer: it is capable of preserving a large number of memories, without the most recent erasing the oldest. On the other hand, it is not active. Your memories do not spontaneously take control of your behavior. It is your conscience that convokes them and makes choices to act accordingly.
Episodic memory is activated for high-profile events. Criteria: unusual, incomprehensible, emotionally charged, having upset habits, or having recognized a new concept. This type of association awakens the engraving of the event by the hippocampus, the key center of episodic memory, richly interconnected with the cortex (interpretation of the facts) and the amygdala (emotional management). The stimulation of these neurons is not so frequent; when they awaken the events surrounding the memory also benefit, even if they are more ordinary: everything that is close to the remarkable moment is attached to the memory.
This memory is little sought in the infant for a simple reason: the entire brain is being engraved. Few significant events because the newborn lacks the elements of comparison. It does not matter if adults feel that exceptional events have affected the child; they will eventually make a memory of it in the child only if their own behavior was, at the moment, very unusual. Many pseudo-memories of childhood were in fact learned afterwards, by the stories heard or the pictures seen.
The episodic memory slowly takes hold, as the child’s brain becomes able to discern the unusual because it already has a sufficient basis of habitual things. There is no absolute age for this beginning. The first memories can go back before 3 years; for the majority of us they are later, and much more for adolescence than early childhood.
As for procedural memory, we could say that it starts even before the differentiation of the first neuron. Our programming, in fact, is included in the genetic heritage. The development of the nervous system and the first instinctive reactions are wholly contained in this code. The beginning of procedural memory is lost in the interminable procession of our forefathers. The first cry, awaited so impatiently by the parents and characteristic of our youngest age, is in fact the most archaic of our memories…