How much attention should you pay to test scores? “A test score alone offers too little information to make meaning of it,” says Paul LeMahieu of the University of Delaware, who also directs research and development for Delaware’s education department and has written extensively about the purposes and techniques of different forms of assessment. Before rushing to actions aimed at raising scores, he suggests asking:
1. How does the testing instrument define what it sets out to measure? Nothing is more important. It may be a math test, for instance, but does that mean math computation or the cognitive processes of solving problems? You may care about only part of what is tested, about it all, or about very little of it.
2. What kinds of judgments, decisions, or actions do you need to make? What kind does the test legitimately support? This is the “validity” question: tests are built differently for different purposes. If you want to make decisions about what or how to teach a particular child, don’t go to a test designed to determine eligibility for a diploma-use an assessment designed for your purposes.
3. What’s the nature of the information provided by the test? Does it refer to whether students know particular material measured by the test questions? (These are criterion-referenced tests.) Or does it quantify how much students know about what is tested compared to other students who took the test? (These are norm-referenced tests.) Some tests score students as to whether they meet specific desired levels of performance (standards-referenced tests). Or a test score can chart its taker’s own change or growth (over time, for instance, or across categories). Depending on your questions about student learning, one or another of these kinds of information may be right for you.
4. How accurately and well does the test measure the abilities of all kids? Look at the questions it includes-they can greatly influence the kinds of scores it yields for students in particular circumstances. And changing various factors not relevant to the matter being tested-such as extending timing, using plainer language or translating assessment tasks, and accepting oral responses-could yield more trustworthy information.
5. What actions or system of actions, if any, does the test suggest? Only the best of tests provide a level of detail that can signal ways the system could do better by kids. Aim for an assessment system that inspires new conversation among professionals and the community, clarifying the expectations for students and disciplining the understanding of what quality is and what serves as its evidence.