Seeing how difficult it is to normalize data got me thinking that in general, we misrepresent most observation as coming from a uniform distribution. This is the same idea that Nassim Taleb develops extensively, most events have long tails (fat tails).
Another way of saying this is that even things that seem continuous are discontinuous, the apparent linear relationships, causative relationships, are in reality non-linear.
For example, nutrients seems to increase linearly, more selenium is better than less, yet at a certain point selenium will become toxic. We perceive a continuous spectrum, measured in mcg, yet it is not continuous, it is two separate disjoint distributions, below toxicity and above toxicity. The artificial measurement of mcg is misleading, as Melanie said, just because we call it something does not make it that thing. Each disjoint distribution is continuous within its space due to a higher level semantic, toxicity.
Similarly, a series of frames are seen as a continuous movie, so it seems to me that the logic is the same, in truth they are discontinuous frames, yet there is a higher level semantic that joins the frames together. It is not that there is a continuous causative relationship that we fill in, rather there is a higher level semantic that project back onto the individual parts.
