Fully standardized assessments evoke performances from examinees under the same conditions, appraise them with the same evaluation procedures, combine the information over tasks with the same model, and interpret scores in the same way for all. Inferential problems can arise when students’ performance is affected by differences in understanding, interacting with, or responding to tasks in ways unrelated to the targeted capabilities. Providing students with assessments that differ as to construct-relevant demands but are matched to students’ capabilities as to construct-irrelevant demands can yield data that are less comparable, yet evidence that is more comparable. The ideas and approaches draw on recent work in international assessment, cross-cultural psychology, and testing of special populations. The discussion is mainly conceptual and concerns argumentation and adaptation at the level of tasks. Chapter 10 will bring in test-level argumentation and measurement modeling.