Transfer down the Hierarchy

An essential feature is that memories can be transfered down the hierarchy.
For instance, I have describe the Red-Square problem so many times that they are now bound together in my mind.
All this means is that I have a CA for Red-Square, and the part of the CA in Semantic Memory is directly accessible.
This was supported by the initial binding. The binding cuased the Red and Square CAs to be active simultaneously.
Due to Hebbian learning, the connections between them were strengthened.
I can still have Red active without Square, but Red will tend to activate Square.
(These are overlapping CAs.)
After recounting the event of where you were when Kennedy was shot for the 100th time, this may be transfered to Semantic memory.
This is enabled by connections between CAs within Semantic Memory, and may make the Episodic memory superfluous.
Higher levels can support learning at the lower levels.