Effects of an oral testing accommodation on the mathematics performance of secondary students with and without learning disabilities.

This study compared the performance of students with and without learning disabilities (LD) on a mathematics test using a standard administration procedure and a read-aloud accommodation. Analyses were conducted on the test scores of 625 middle and high school students (n = 388 with LD) on two equivalent 30-item multiple-choice tests. Whereas mean scores for students both with and without LD were higher in the accommodated condition, students without disabilities benefited significantly more from the accommodation (ES = 0.44) than students with LD (ES = 0.20). In addition, effect sizes from the present study were combined meta-analytically with those of previous studies. Results of the meta-analysis revealed that for elementary students, oral accommodations on a mathematics test yielded greater gains for students with LD than for students without disabilities: for secondary students, the converse was true. Findings of the study are discussed in relation to the question of the validity of an oral accommodation on mathematics tests for students both with and without disabilities.

**********

One of the most important accomplishments of recent U.S. federal education legislation has been to promote the full participation of students with disabilities in state educational accountability systems. The Individuals with Disabilities Education Act (IDEA) Amendments of 1990, the No Child Left Behind Act of 2001, and the most recent reauthorization of IDEA as the Individuals with Disabilities Education Improvement Act (IDEIA; 2004) have affirmed the principle of including students with disabilities in statewide assessments, as well as the need to offer appropriate accommodations or alternate testing procedures, as necessary, to support students' participation.

Over the past 15 years, the National Center on Educational Outcomes (NCEO) has been documenting states' policies and practices regarding the participation of students with disabilities on statewide assessments. In its latest report, Thompson, Johnstone, Thurlow, and Altman (2005) stated that one of the six key factors cited by states as contributing to positive trends in the participation and performance of students with disabilities has been the development and provision of accommodation guidelines and training. In addition, recent studies have catalogued the rapidly evolving use of accommodations on statewide tests (e.g., Johnson, Kimball, Brown, & Anderson, 2001; Thurlow, House, Scott, & Ysseldyke, 2000).

Given the serious consequences of test outcomes for states, districts, schools, and individual students, the validity of interpretations of test scores when students are given particular accommodations has been a critical question in both the research and policy arenas (Thurlow & Bolt, 2001; Thurlow, House, et al., 2000; Thurlow, McGrew, Tindal, Thompson, Ysseldyke, & Elliott, 2000; Tindal, 2002; Tindal & Fuchs, 1999). There is general consensus that to be considered a valid accommodation, a modification in test administration should remove disability-related variance without affecting construct-relevant variance. For example, allowing students with motor difficulties to dictate their solutions to mathematics problems to a scribe addresses the students' specific disability without affecting their mathematics skills. This accommodation would be expected to improve the test performance of students with motor impairments only. If the accommodation were given to students without motor impairments, no impact on test performance would be expected to result.

One test of the validity of a testing accommodation is whether it changes the meaning of test scores as evidenced by variance in factor structure or differential item functioning across tests administered with and without accommodations. Pomplun and Omar (2000) investigated the factorial structure of a fourth-grade state mathematics assessment administered to three groups of students: general education students taking the test without accommodations, students with LD taking the test without accommodations, and students with LD taking the test with a read-aloud accommodation. Results indicated the invariance of the test's factor structure across all three groups, providing support for the comparability of scores under both testing conditions and the validity of aggregating the scores of students with and without disabilities. A similar finding of invariance in factor structure was reported by Huynh, Meyer, and Gallant (2004) for an eighth-grade mathematics test administered to general education students and also to students with disabilities, with and without an oral accommodation.

Fuchs (2000) examined whether various testing accommodations were associated with differential item functioning (DIF) for students with disabilities compared to students without disabilities. In general, item functioning should be invariant in regard to characteristics of test-takers that are not related to the construct being measured, for example, gender and ethnicity. However, item functioning is expected when an accommodation removes construct-irrelevant variance. Thus, if reading ability is irrelevant to a measure of mathematics ability, then a read-aloud accommodation should change item functioning only for those test-takers with poor reading ability. Results showed that 50% of concepts and applications items showed evidence of DIF such that students with LD had improved performance on those items with a read-aloud accommodation.