[1]There are some approaches in symbol processing AI (in the field of machine learning) which in fact do generalize to some extent but the ways in which generalization is achieved is typically ad hoc. See also below: "Integrated learning".

[2]The field of robotics is an exception.

[3]In one experiment a tape recording from a child was transcribed into English text and for each letter the phoneme encoding as pronounced by the child was worked out by the linguists. In a different experiment the prescribed pronunciation was taken from a dictionary.

[4]The term "functional" is used in different ways. Here the term is used to distinguish one level of explanation from a learning/developmental and an evolutionary one.