The Effects of High-Stakes Testing on Achievement: Preliminary Findings about Generalization across Tests

Detailed evidence is presented about the extent of generalization from high-stakes tests to other tests and about the instructional effects of high-stakes testing. Data are from grade 3 of a large, high-poverty urban district with large numbers of Black and Hispanic American students. The district's results in 1990 for two tests, designated Test B and Test C, were compared to the researchers' results for Test C (N=840 students in 36 schools). Student-level comparisons were made for three study tests and Test B. For mathematics, all comparisons, at district and student levels, support the primary hypothesis that performance on the conventional high-stakes test does not generalize well to other tests for which students have not been specifically prepared. Evidence in reading is less consistent, but suggests weaknesses in generalizing in some instances. Even the preliminary results presented in this paper provide a serious criticism of test-based accountability and raise concerns about the effects of high-stakes testing on instruction. Teachers in this district evidently focus on content specific to the test used for accountability rather than trying to improve achievement in the broader, more desirable sense. Five references are listed, and four tables of study data and six illustrative figures are included. (SLD)