Home recordings would probably be a more reliable source for an ethnographic study of this sort. Of course, the locations of those home recordings would need to be known, and the dates would be helpful. I have a couple I can think of - one taken in Camden NY in 1905, and one of the Welch family (I cannot remember offhand where they lived) taken on Christmas Day, 1912. They all sounded "normal" to me.
Talking records by companies such as Berliner (employing local talent) might give some insight. George Graham has a slight mid-Atlantic accent...
A study like this would be a huge undertaking if it were to be properly approached (unlike the fluff we read on the Internet).

George P.