objectives—obtaining valid information while still testing all English-language learners—create a sizable policy tension for the design of assessment systems, particularly when they involve high stakes.
Recommendation: Systematic research that investigates the impact of specific accommodations on the test performance of both English-language learners and other students is needed. Accommodations should be investigated to see whether they reduce construct-irrelevant sources of variance for English-language learners without disadvantaging other students who do not receive accommodations. The relationship of test accommodations to instructional accommodations should also be studied.
Recommendation: Development and implementation of alternative measures, such as primary-language assessments, should be accompanied by information regarding the validity, reliability, and comparability of scores on primary-language and English assessments.
A sufficient number of English-language learners should be included when items are developed and pilot-tested and in the norming of assessments (Hambleton and Kanjee, 1994). Experts in the assessment of English-language learners might work with test developers to maintain the content difficulty of items while making the language of the instructions as well as actual test items more comprehensible. These modifications would have to be accomplished without making the assessment invalid for other students.
Recommendation: The learning and language needs of English-language learners should be considered during test development.
Various strategies can be used to obtain valid information about the achievement of English-language learners in large-scale assessments. These include native-language assessments and modifications that decrease the English-language load. Such strategies, however, are often employed inconsistently from place to place and from student to student. Monitoring of educational outcomes for English-language learners as a group is needed to determine the intended and unintended consequences of their participation in large-scale assessments.