Jim,
Of course the aural process has a great deal to do with the variation, but not as much as most people think. As I said, a large slice is down to the printer's hacks altering and mixing and matching. It would take a mammoth study to actually quantify this or come up with percentages, but I have spent a long time studying and comparing stall versions and I'm convinced their intervention is well understated. In Ireland for instance the likes of Brereton, Haly, Baird, Birmingham were similarly involved in this.

Most of those ballads that came down to us from the 17thc hadn't existed purely in oral tradition for 300 years (although a few had). Hacks and printers themselves were going back to the longer ballads and shortening them for more recent tastes. They often did this by splicing 2 stanzas together to make one, and similar processes. William Taylor for instance had about 30 double verses before it was pared down to the version we are familiar with.