Great bit of research. I’d be very interested to see how this sort of experiment works with non-Latin character sets.