The aim of the sampling design for stereology is to obtain the maximal amount of quantitative structural information at a given total cost or effort. Principles of such optimal designs are discussed and methods for generating them are illustrated by a biological example. In general, the variation between different individuals--the biological variation--is the major determinant of overall efficiency, whereas the variation between single microscopic features is unimportant. It follows that the expenditure of time and/or money in order to increase the precision of the individual measurements is irrational in almost all studies where the emphasis is on the biological results.