Problems with performance assessment (PA) and multiple-choice tests (MCTs) are outlined, with reference to the literature on accountability. PA for individual teachers who should integrate their assessments with their instruction; PA as a supplement to more traditional examinations for licensure decisions; and some limited, experimental tryouts of PA for other accountability purposes are supported. The anti-MCT demagogues, and making PA the latest fad are not supported. Reasons for PA's popularity include: old (but inaccurate) criticisms of MCTs in terms of bias, irrelevant content, and measurement of only recognition; cognitive psychologists' belief that many parameters that they want to study require formats other than MCT questions; increased concern that MCTs delimit the domains that should be assessed; wide publicity of the Lake Wobegon effect of teaching too closely to MCTs; and claims that teaching to MCT formats has deleterious instructional/learning effects. PA problems vary depending on several dimensions, such as secure versus non-secure assessments, matrix versus every student assessment, and accountability versus instruction. PAs have difficulty meeting the five "apple" criteria required of high-stakes tests used for accountability purposes: administrative feasibility, professional credibility, public acceptability, legal defensibility, and economical affordability. It is concluded that MCTs measure some things very well and efficiently; however, they do not measure everything and their use can be overemphasized. PAs can measure important objectives that cannot easily be measured by MCTs. A 52-item list of references is included. (RLC)