I know a lot of other bloggers have spoken about this, and I am not going to say much, except READ Judy Redman's article on memory in psychological research as it bears upon the gospels. It is quite thought-provoking and, frankly, a breath of fresh air on what seem to become rather stale debates.
Just one completely unimportant note: Her discussion of schema reminded me of Albert Lord's definition of "theme" as something of a standard outline to fill in with detail (with each new remembering for Redman; with each new performance for Parry/Lord). She does speak of Lord briefly to discuss the memory of trained tradents, although I think I read Lord a little differently. Whereas she indicates the oral tradents' usage of verbatim speech--what Lord would call the Formulae--my reading of Lord (which is something he highly emphasizes in his introduction to the Singer of Tales) is that the moment of performance is itself the moment of composition. The entire narrative is not repeated verbatim, but constantly changes from context to context in length, in elaboration, etc. Certain formulae may be remembered verbatim (but these are limited to repeated clauses, like "rosy-fingered dawn," "manslaughtering Hektor," "godlike Akhilleus," etc.), but the smaller units of narrative, the particular imagery used in simile, and even larger chunks of narrative change with each new performance, since each new performance is a new composition. This is neither here nor there, however, for Redman's main point, since the earliest "earwitnesses" as Redman calls them were hardly highly trained bards.