T NHE WIDESPREAD distrust of social statistics rests upon the experience that the most contradictory results are deduced from identical data by so-called experts. The public sees continuously that statistical figures, by whatever methods they may be gathered,' are handled carelessly, ignorantly, sometimes with a delusive design. It thus comes to believe that exact methods of statistical control and interpretation scarcely exist. It must be admitted that statistical procedures as used by many criminologists, sociologists, medical men, and in our official publications are rough and poorly conceived. Unreduced figures are often compared. This is, of course, an unscientific procedure. When reduced figures are used, the reduction rests on a chosen unit of the general population (per ioo,ooo population). This standard, while an improvement compared with the crude uncertainty of unreduced figures, is still distinctly defective. Two books of 500 pages each would never be regarded as equivalent and therefore comparable entities. Just so, a population unit of Iooooo is a most heterogeneous composite; it is equal in no respect to another Ioo,ooo of population except in the mere mechanical and quantitative relation of numbers. Thus, the polymorphous compound, called population, has to be broken down into more elementary and primitive constituents. Births have to be related to females only who might be very unequally distributed in a segment of the population. Senile psychoses, obviously, should be connected with the older sections of the population, rapes to the number of adult males in a given population,2 larcenies to the needy strata, and not to hobos and bankers alike. All these classes might vary very widely in two numerically equal units of population and such diversity would alter profoundly the result of a statistical parallel. In the course of many thousands of computations made by the writer as an associate editor of the Attorney General's Survey of Release Procedures and as director of the Colorado Crime Survey, he was struck by the fact that widely varying results, in comparing a group of states for instance, were caused by some faulty statistical procedure. The comparative basis seemed to be identical but it was in reality most disparate. Application of the right, or a revised standard, reduced the numeri-