This report describes an experiment in constructed response testing undertaken in conjunction with the National Education Longitudinal Study of 1988 (NELS:88). Constructed response questions are those that require students to produce their own response rather than selecting the correct answer from several options. Participants in this experiment took constructed response tests in mathematics or science along with a battery of traditional multiple choice tests. Data were also collected on students' backgrounds and school experiences. The experiment, the High School Effectiveness Study (HSES), was designed to explore practical and psychometric issues in using constructed response test formats in large-scale, voluntary national surveys. After the eighth grade, NELS:88 participants dispersed to many high schools. For this study, a probability subsample of 247 urban and suburban NELS:88 first-followup schools in 30 metropolitan areas was chosen, and within-school cluster sizes of about 30 12th-grade students were also selected. The constructed response tests had somewhat lower levels of reliability than the multiple choice tests, but the greatest restraint on use of constructed response tests was the cost of scoring. Constructed responses must be read by individual human scorers. Although constructed response tests are time consuming to administer and expensive to score, they may provide diagnostic information and measurements of skill that are difficult to evaluate with multiple choice questions. Five appendixes present technical information, sample questions, descriptions of results, and data files. (Contains 13 tables and 13 references.) (SLD)