Reading is a basic skill. It has commonly been taught and tested in American public schools and is almost always a focus of concern in basic skills testing at any level, including higher education. The first reading tests were developed early in the twentieth century, and reading tests continue to reflect the design decisions made at that time ( Pearson & Valencia, 1987). Despite the fact that reading theory has changed a great deal since then and even the most basic conceptions of what takes place during reading have been altered, our reading tests have not changed much over the past sixty years ( Farr & Carey, 1986). We have become prisoners of test design efforts that have been aimed more toward establishing efficient measures with high levels of concurrent validity than toward establishing theoretically sound measures that possess construct validity. With each new test we have attempted to duplicate, in some fashion, the earlier test design efforts in an extensive and elaborate daisy chain that has not necessarily resulted in either a valid or a useful measurement of reading ability ( Pikulski & Shanahan, 1980).
Any test is an attempt to describe the whole on the basis of a part. We cannot possibly observe an individual trying to accomplish all reading tasks with all texts, so, instead, we select a significant and representative sample of behaviors on which to evaluate individuals and programs. On the basis of the performance on these samples, we then draw conclusions about how well somebody can read. From these sample test behaviors, we predict how individuals might
Timothy Shanahan is professor of Curriculum and Instruction in the College of Education at the University of Illinois at Chicago.