Randomization and the analysis of variance

This paper attempts to strengthen faith in randomization, first by defining it in general and second by showing how it is the fundamental means, in many experiments, of generating the probability space. It is defined by the natural structure of the experimental units (E.U.'s) not by a particular experimental design imposed on that structure. Three examples are given, with the same design but different E.U. structures and randomizations, to show that models developed from experimental designs are superficial, but that their randomization is fundamental.