Recent empirical evidence concerning sex and racial bias in testing is discussed in terms of three primary sources of bias: (1) content of the test itself, (2) atmosphere in which the test is administered, and (3) the use to which the test results are put. Test content that is demonstrably more difficult for one group than another should be (1) eliminated in any setting in which equal difficulty is assumed or (2) perhaps more important, the biased content should be examined closely for possible causes of the difference, leading to modification of educational practices for the low-scoring groups. Special care should be taken routinely to see that minority groups are made to feel comfortable and are not intimidated by their surroundings. Pertaining to fairness in test use, methodological developments undermining the traditional statistical model of fairness previously accepted without question are described in some detail. The "new measures" approach to test bias is seen as essentially an abandonment of, or a reduced emphasis on, the traditional measures of status of aptitude and achievement. (Author/RC)