the mix of activities, the distribution of time spent on various tasks, and the assessment measures to which they attend and respond. This work must be coupled with research to test the core hypotheses experimentally, so that the causal mechanisms are clarified.

As with reading instruction in the earlier years, features are likely to differ by grade and by the average achievement level and language development of the students in the classroom. Designing the research to look at various levels (e.g., third, sixth, and ninth grade) and at classrooms chosen to represent a wide variety of demographic factors (e.g., in suburban high-scoring schools, urban low-scoring schools, schools serving language minority learners) would again be required to draw implications for practice with specific attention to differences for students in different demographic groups and in different grades.

The next phase of R&D would involve the design of interventions that incorporate those components into instruction systematically, in an effort to verify their effectiveness experimentally, and to assess their efficacy with a wider array of students and reading curricula or subject matter. The interventions should address both teacher learning and student learning.

The design and testing phases of the development projects would look much like those in the initiative on early reading interventions. An important benefit of an R&D network is that the expertise in doing this type of work begins to accumulate in an organizational setting in which what is learned—both in research outcomes and in the conduct of research and development—has a continuing influence on future R&D projects and designs.

Simultaneously with the efforts to explore, identify, then rigorously test best practices in comprehension instruction, it is crucial that the educational research and practice communities collaborate with stakeholders, such as future employers, faculty in community colleges and universities, and others interested in educational outcomes, to define what adequate reading comprehension is for readers of various ages. The SERP networks are in a unique position to engage in this kind of stock-taking on