The research conducted in this area to date has, for the most part, been inconclusive (Sireci et al., 2003). This is not to say that the existing research is not useful. Indeed, many of the studies of accommodations on NAEP and other large-scale assessments have been well designed and implemented. However, as described in Chapter 5, they have for the most part focused on differences in scores for students taking tests with and without various accommodations. They have not investigated the comparability of scores obtained from accommodated and unaccommodated administrations, nor have they illuminated the extent to which similar inferences can be drawn about scores obtained under different conditions. In our view, research should more specifically address the central validity concerns that have been raised about the inferences that can be drawn from scores based on accommodated administrations, and this process should begin with articulation of the validation argument that underlies performance on the assessment.
One purpose that can be served by this report, therefore, is to suggest an alternative approach to research on the validity of scores from accommodated administrations. We do this by suggesting ways in which an inference-based validation argument for NAEP could be articulated. Such an argument would provide a basis for conducting validation research that systematically investigates the effects of different accommodations on the performance of students with disabilities and English language learners. Such a validation argument would also inform assessment design and development, since the effects of different alterations in task characteristics and in test administration conditions caused by accommodations could be better understood using this approach.
In this chapter we lay out a procedure for making a systematic logical argument about:
the skills and knowledge an assessment is intended to evaluate (target skills),
the additional skills and knowledge an individual needs to demonstrate his or her proficiency with the target skills (ancillary skills), and
the accommodations that would be appropriate, given the particular target and ancillary skills called for in a given assessment and the particular disability or language development profile that characterizes a given test-taker.
We begin with an analysis of the target and ancillary skills required to respond to NAEP reading and mathematics items. We then discuss procedures for articulating the validation argument using an approach referred to as evidence-centered design investigated by Mislevy and his colleagues (Hansen and Steinberg, 2004; Hansen et al., 2003; Mislevy et al., 2002, 2003). We illustrate the application of this approach with a sample NAEP fourth grade reading task. This example demonstrates how the suitability of various accommodations for