Administering standardized achievement tests to young children: How mode of administration affects the reliability and validity of standard- ized measures of student achievement in kindergarten and first grade.*

This paper reports on an experiment conducted to examine the consequences of assessing kindergarten and first-grade students’ academic achievement in group versus individualized assessment settings. In the experiment, 442 students blocked by classroom and grade level were randomly assigned to one of two assessment modes—a small group setting with 8 other students from their classroom versus an individualized setting. Students in both settings were administered the grade appropriate form of the CTB McGraw-Hill Terra Nova Tests of Achievement, Form A by trained assessors from the Institute of Social Research at the University of Michigan. Assessment results were then scored by the publisher. The results of the experiment showed that in both kindergarten and first grade, group assessment settings were more likely than individualized settings to be characterized by behavior that assessors coded as disruptive or districting for students, and that students at both grade levels who were assessed in the group setting omitted more test items and made more multiple marks on items than did students assessed in the individual setting. The study also found that kindergarten students assessed in the group setting had lower Reading, Language, and Mathematics scale scores (as estimated by the publishers’ three parameter IRT model) and that these scale scores had higher standard errors of measurement for kindergarten students assessed in the group setting. However, there were no differences in measured achievement or standard errors of measurement across assessment modes among first grade students. We argue that the differences in assessment environments and item-response patterns of students in group settings call into question the validity of assessment results for young children assessed in group settings, even when such results do not result in observable differences in the measured outcomes of these children compared to students assessed individually.