When a low-level English-learner answers a long-division problem incorrectly on a state exam, is it because the student can’t do the math? Or is it because the student lacks the English proficiency to understand the directions?
With the tests used now, discerning the reason is difficult, if not impossible.
But as test designers work to craft the new, common assessments set to debut in most of the nation’s public schools in the 2014-15 school year, their goal is to provide all English-language learners, regardless of their language-proficiency levels, the same opportunities to demonstrate their content knowledge and skills as their peers who are native English-speakers or former English-learners.
Doing so, however, will take an unprecedented effort not only to better understand the types of testing accommodations that can give English-learners full access to the math and English/language arts exams without compromising the material being tested, but also to get states to agree on them. And it will require the reversal of a decades-long practice in many schools: using a wide range of testing accommodations for ELLs that were originally designed for students with disabilities.
The two groups of states working together to develop the new assessments that will measure how well students are mastering the Common Core State Standards—the Smarter Balanced Assessment Consortium and the Partnership for Assessment of Readiness for College and Careers—have been working for months to develop guidelines and policies on appropriate accommodations for English-learners and students with disabilities.
So far, neither group has released a concrete set of policy recommendations on the types of supports that will be available to ELL test-takers within each consortium, but both groups are working on drafts for their members to consider.
Smarter Balanced and PARCC officials say they are striving to make test items on the computer-based exams “accessible” to the widest possible spectrum of students and are drawing on evidence from research to inform their choices on what types of accommodations should be allowed.
Both groups also must reconcile accommodations policies that vary widely across states. Among PARCC’s 22 members, for example, Arizona and Massachusetts are English-only states that provide no native-language support to ELL test-takers, while in Illinois, another PARCC member, educators are given discretion to provide support such as test directions presented orally in a student’s first language.
“We are using research to drive this process,” said Magda Chia, the director of underrepresented students for the 24-state Smarter Balanced consortium. “We’re not making decisions based on what has been happening.”
Many Devices
Researchers estimate that states and school districts collectively have been drawing from as many as 75 different testing accommodations for English-learners, many of them designed originally for students with disabilities.
Both Smarter Balanced and PARCC have done surveys of their member states to collect the number and types of testing supports allowed for English-learners. Examples include giving ELLs extra time to take their tests and providing commercial dictionaries to look up unfamiliar words.
(Federal law requires that English-learners who have been enrolled in U.S. schools for at least one year take state content tests for accountability purposes.)
“The policies across the states are really all over the map,” said Tamara Reavis, PARCC’s senior adviser for assessment, accessibility, and equity.
Other fairly common accommodations being used for ELLs are giving tests in small groups and allowing students to write answers in test booklets rather than on an answer sheet, said Jamal Abedi, an education professor at the University of California, Davis, who is advising the Smarter Balanced states on accommodations for English-learners.
Those accommodations, he said, may not have “any relevance for ELLs.”
“It’s one of the major misconceptions in accommodations,” he said. “States use these tools for the sake of accommodations with no consideration of whether it actually makes the assessments more accessible to ELLs.”
At the behest of Smarter Balanced, Mr. Abedi and four other colleagues, including Ms. Chia, pored over the available research to develop a system for rating whether specific testing accommodations are effective in making a test item more accessible to an English-learner, as well as whether they are valid. To be valid, Mr. Abedi said, an accommodation cannot alter the construct of the item being tested, meaning it can’t change what the item is intended to measure.
“It’s not going to be valid to provide a test to a student in his or her native language if all their instruction has been in English,” he said. “It may not help them and could even hurt them.”
The Smarter Balanced team of experts rated all the accommodations in use for ELLs and students with disabilities in the consortium’s member states and made recommendations to “use,” “use with caution,” or “not use.” The accommodations that earned a “use” recommendation had to have sufficient evidence in the research literature that they both make the test items more accessible to ELLs and do not alter the construct of test items by providing an unfair advantage.
A commercial dictionary is one commonly used accommodation for ELLs that the panel rated as one to “use with caution,” Mr. Abedi said. While dictionaries make test items with unfamiliar words more accessible, they can also provide content information that would render a test-taker’s answer invalid.
To address that issue, both Smarter Balanced and PARCC have been developing customized, digital “pop up” dictionaries that would only provide definitions of words that are unrelated to the content being tested. All students, not just ELLs and students with disabilities, would be able to use those tools, said Ms. Chia. (“Test Designers Tap Students for Feedback,” December 7, 2012.)
Mr. Abedi is also urging the Smarter Balanced states to weigh the feasibility of accommodations as they consider what to adopt. One example of an accommodation that may not be feasible on a large scale, he said, is a one-on-one test setting for students to take their exams with only an adult educator present.
‘Language-Complexity Tool’
Smarter Balanced officials are conducting thorough evaluations of the level of difficulty of the language used in each test item they write to determine whether it can be made simpler without changing what items are meant to measure.
H. Gary Cook and Rita MacDonald, English-language-acquisition researchers at the Wisconsin Center for Education Research based at the University of Wisconsin-Madison, developed a “language-complexity tool” for test designers. It is intended to help them figure out how dense the ideas packed into the language of a test item are, and if the vocabulary, grammar, and syntax are more sophisticated than is necessary to measure whatever piece of knowledge or skill an item is seeking.
The tool is actually a rubric that test writers can use to assign numbers to the various language-complexity components of a test item.
“The question test designers can ask is, ‘Does the language need to be this sophisticated in order to get at what you are trying to measure?’ ” said Mr. Cook.
“And it will ultimately be very useful for accountability purposes, because right now, when I look at an English-learner’s content-test results, especially in math, I don’t know if I am looking at math or if I am looking at language,” he said. “This helps us untangle the language better.”
PARCC is also “looking very closely at issues of language complexity and how it ties into assessment,” said Ms. Reavis.
“We want the language level of a test item to be in balance with the cognitive level,” Ms. Reavis said.
Ultimately, both common-core-assessment consortia say they want to create assessments and accommodations policies that can be widely used by all students, whether they are English-learners, have a disability, or are in general education.
But even with the most accessible test design, ELLs will still struggle to demonstrate what they truly know if their teachers aren’t prepared to help them develop their academic-language skills across the content areas, said Mr. Abedi. The language demands in the common standards are much more sophisticated than what they have been in most existing state content standards.
“Just look at the kindergarten math standards, where students are being asked to ‘describe’ and ‘explain’ and ‘elaborate’ their answers in written language,” Mr. Abedi said. “If they are going to succeed, all teachers must teach them the language.”
