Feedback, however, is not "the answer"; rather, it is but one powerful answer. With inefficient learners, it is better for a teacher to provide elaborations through instruction than to provide feedback on poorly understood concepts. If feedback is directed at the right level, it can assist students to comprehend, engage, or develop effective strategies to process the information intended to be learned. To be effec- tive, feedback needs to be clear, purposeful, meaningful, and compatible with students' prior knowledge and to provide logical connections. It also needs to prompt active information processing on the part of learners, have low task complexity, relate to specific and clear goals, and provide little threat to the person at the self level. The major discriminator is whether it is clearly directed to the task, processes, and/or regulation and not to the self level. These conditions highlight the importance of classroom climates that foster peer and self-assessment and allow for learning from mistakes. There are major implications for the design of assessments. Too often, assess- ments are used to provide snapshots of learning rather than providing information that can be used by students or their teachers to address the three feedback ques- tions. Certainly, a critical conclusion is that teachers need to seek and learn from feedback (such as from students' responses to tests) as much as do students, and only when assessment provides such learning is it of value to either. Most current assessments provide minimal feedback, too often because they rely on recall and are used as external accountability thermometers rather than as feedback devices that are integral to the teaching and learning process. It is the feedback informa- tion and interpretations from assessments, not the numbers or grades, that matter. In too many cases, testing is used as the measure to judge whether change has occurred rather than as a mechanism to further enhance and consolidate learning by teachers or students. The costs of these thermometer-related accountability tests are high, and the feedback returns are minimal (Shepard et al., 1996).