But if you're going to copy the byte anyway, why not copy it into a bit-addressable byte, so that the required bit can then be tested directly - without all that tedious messing about with masking, etc?

Spencer twice stated that he wants to detect a bit change. How does copying the byte to bit addressable memory help? I'm probably missing something, but it seems like doing it that way one would have to make two bit-check paths -- one path if the original bit is set and you'd check to see when if the new sample cleared, and another path if the original bit is clear and you'd check to see when if the new sample became set.