Complexities of test fairness are described in nontechnical language, and their implications for the selection procedures practiced in our society are discussed. Four clearly distinguishable models of fair selection are presented: the Cleary, or traditional model; the Cole model; the Thorndike model; and the Darlington model. A distinction is made between the use of tests in a manner which is "fair," and the concept of "test bias," which most frequently refers to the content of the items of the test, regardless of any particular use to which the test is being put. It is possible to conceive of a biased test being used in a fair manner, and also possible to imagine an unbiased test being used unfairly. This discussion concerns the use, rather than the content of tests.