I recall sitting in the 200 seminar at Harvard when James Kugal (completely off the cuff) quotes what must have been more than two dozen lines from Deuteronomy. While making a point a few seconds later, he stops, as he recognized he misvocalized a pausal form. I personally have known others from Yeshivas who have this capacity, as well as a number of Muslims who have memorized the Qur'an. Doubtless you are aware of accounts of Homeresque tellers of tales. The hypothesis of writing from memory is more than just vaguely logically possible. It could well be right. The array of new types of textual corruptions which could flow from this process are intriguing, and would be the determining factor in making this claim.
Mark Anthony Phelps
[Non-text portions of this message have been removed]