Alas, this is pure conjecture and highly contoversial.
Piaget for example was of the view that a newborn infant had no concept of a distinction between "self" and "world", and that such a distinction arises later through a process of "de-centration". Even after the acquisition of "I" the child remains "ego-centric" to the extent that the child's viewpoint is assumed by the child to be that of others.
Other writers claim that that a concept of "self" as an "actor" amongst others requires socialization through language. In the early stages of this, no significance is attached to idiosyncratic variants of utterances like "Me do it"/"I do it"/"Peter do it". Concepts of "personal pronouns" are deemed to belong to the realm of abstract grammatical analysis rather than that of developmental psychology.
In anticipation of coberst's rejoinder that such points are "negative", he would be correct in as much they are dismissive of his perpetual source
Becker as the "fount of all wisdom". Concepts of "self" have been discussed in great detail on this forum and I invite coberst expand his knowledge by investigating them.
e.g.
http://www.able2know.com/forums/viewtopic.php?t=14696&highlight=