The question of how standardized tests can be better developed to improve educational program evaluation is probed in this paper. After the first section's brief introduction, section 2 explores the thesis that tests developed in terms of selection and inference may not serve current social functions of educational testing. To clarify this thesis, section 3 recounts an example of instrument development from the history of Project Follow Through, suggesting that the value of an instrument may be overlooked because the instrument is judged by criteria inappropriate to the original motivations behind its development effort. Section 4 attempts to go beyond the statement of the problem to suggest how thinking of a test as a source of individual learning might guide test development in nontraditional ways. Section 5 sums up some of the possible connections between testing and various social functions, pointing to some alternate ways in which standardized testing may serve goals of evaluation. (RH)