My issues with the simple classifications is noted well by Leo's article This is why we did this on Savant. The problem here is that while it is a huge step forward analytically, it's a step backward in trying to use it conversationally. We basically need something between the two