While it is an effective method in ANNs, rehearsal is unlikely to be a realistic
model of biological learning mechanisms, as in this context the actual old
information (accurate and complete representations of all items ever learned
by the organism) is not available. Pseudorehearsal is significantly more likely
to be a mechanism which could actually be employed by organisms as it does
not require access to this old information, it just requires a way of
approximating it. (Robins, 1996)