This potential consequence argues for helping learners understand and share learning goals.

We have described ways in which classroom assessment can be used to improve instruction and learning. We now turn to a discussion of assessments that are used in large-scale contexts, primarily for policy purposes. They include state, national, and international assessments. At the policy level, large-scale assessments are often used to evaluate programs and/or to set expectations for individual student learning (e.g., for establishing the minimum requirements individual students must meet to move on to the next grade or graduate from high school). At the district level, such assessments may be used for those same purposes, as well as for matching students to appropriate instructional programs. At the classroom level, large-scale assessments tend to be less relevant but still provide information a teacher can use to evaluate his or her own instruction and to identify or confirm areas of instructional need for individual students. Though further removed from day-to-day instruction than classroom assessments, large-scale assessments have the potential to support instruction and learning if well designed and appropriately used. For parents, large-scale assessments can provide information about their own child’s achievement and some information about the effectiveness of the instruction their child is receiving.

Substantially more valid and useful information could be gained from large-scale assessments if the principles set forth in Chapter 5 were applied during the design process. However, fully capitalizing on the new foundations described in this report will require more substantial changes in the way large-scale assessment is approached, as well as relaxation of some of the constraints that currently drive large-scale assessment practices.

As described in Chapter 5, large-scale summative assessments should focus on the most critical and central aspects of learning in a domain as identified by curriculum standards and informed by cognitive research and theory. Large-scale assessments typically will reflect aspects of the model of learning at a less detailed level than classroom assessments, which can go into more depth because they focus on a smaller slice of curriculum and instruction. For instance, one might need to know for summative purposes whether a student has mastered the more complex aspects of multicolumn subtraction, including borrowing from and across zero, rather than exactly which subtraction bugs lead to mistakes. At the same time, while policy makers and parents may not need all the diagnostic detail that would be