A while ago, when thinking about the boundary between long and short, decided that the problem was that we use discrete labels for continuous phenomena.
I suppose this may be related to the Anchor Bias (http://www.overcomingbias.com/2007/09/anchoring-and-a.html). Something that is not a heap, after a grain of sand is added, is still a heap. Something that is a heap, after a grain of sand is removed, is still a heap.
An interesting instance of this arose when determining how numbers are described in the Piraha language, a language with only three words for quantities. Seeing one battery, the Piraha called it "ho'i". When they added one more (a large increase percentage-wise), they immediately switched to another word, so therefore "ho'i" means "one." But when they started with ten batteries and started removing them, one of them started calling it "ho'i" at six batteries, so therefore "ho'i" really means "few." (http://en.wikipedia.org/wiki/Pirah%C3%A3_language#Numerals_a...) I'd explain that as: when deciding whether to call something "few" or "many," they anchored off the initial judgement, and looked at the percentage change. So one politician could convince voters a 10% tax increase will have no effect by starting off "Well, what would a 0% increase do? How about 1%?", while another could convince them it would be devestating by starting off at 20%.