This is one of the things that did in Jan Hendrik Schön.

I can't remember the exact measurement but he was showing a set of curves that were simply too neat for a relatively small sample size - roughly 100 or so plastic transistors. In reality you cannot get that neat an effect without a huge sample population