Scholars Seek Best Ways to Assess English-Learners

Simplifying test questions so that they avoid unnecessarily complex English is the best way for states to include English-language learners in large-scale testing, according to the most prominent researcher on testing accommodations for such students.

But other scholars say not enough studies have been conducted to know how states should alter standardized tests, or the conditions under which they are given, to most accurately assess children with limited English.

Jamal Abedi, an education professor at the University of California, Davis, and the most-published researcher on testing accommodations for English-language learners, recommends that states devise separate versions of their academic tests using modified English. In that process, the wording of test items is changed so that students aren’t tripped up by complicated grammar, difficult vocabulary, or other linguistic obstacles.

“That’s the only accommodation that narrows the performance gap between English-language learners and non-English-language learners,” Mr. Abedi maintains.

At a time when the federal No Child Left Behind Act has spotlighted issues surrounding the assessment of English-language learners, he considers modified wording the best accommodation available for such students.

Some states have employed Mr. Abedi as a consultant on accommodations for English-learners, and in April 2003, the American Educational Research Association gave him its professional-service award for his contributions to relating research to practice.

Research Base Limited

Still, Charlene Rivera, the director of the Arlington, Va.-based Center for Equity and Excellence in Education, cautioned that the number of studies that support Mr. Abedi’s recommendation is “extremely limited.”

She writes in a research paper scheduled for publication this spring that only 15 reliable studies were conducted from 1990 to 2003 on the effectiveness of various testing accommodations for English-language learners.

Each of those studies examined one or more of five accommodations: simplifying English wording; providing customized dictionaries or glossaries; testing in students’ native languages; reading test items or directions aloud; and providing extra time.

To Margo Gottlieb, the director of assessment and evaluation for the Des Plaines, Ill.-based Illinois Resource Center, which specializes in helping schools serve English-learners, the findings to date suggest that the whole subject needs more study.

“No research on accommodations [for English-learners] is conclusive,” she said.

In addition to tests with modified English, it makes sense to assess ELL students in their native languages, Mr. Abedi’s research has shown, but only if the students are being taught the tested subjects in those languages.

If Spanish-speaking students who take mathematics in English are assessed in that subject with tests translated into Spanish, for example, they may do worse than if the exams had been in English because of unfamiliar Spanish math terminology.

Mr. Abedi says translation challenges make developing valid native-language tests much harder than creating reliable versions in modified English.

He and his research colleagues have identified a list of 48 linguistic features that can be changed without affecting the validity of a test. He recommends modified-English versions in subjects such as math, science, and social studies, but not reading. That’s because an obvious purpose of a reading test is to assess language skills.

In evaluating the research on testing accommodations, Ms. Rivera’s center selected only studies that used experimental or quasi-experimental research designs and examined the impact of accommodations on students’ test scores. Mr. Abedi was a researcher for seven of those studies.

“Although research on accommodations for ELLs is inconclusive,” the center says on its Web site, “two kinds of accommodations appear to have potential to support ELLs’ access to test content: native language and linguistic simplification.”

Charles Stansfield, a language-acquisition researcher who now runs his own test-development company, Second Language Testing Inc., in Rockville, Md., agrees with Mr. Abedi that modifying the wording of test items is the best accommodation.

Mr. Stansfield, who is married to Ms. Rivera, developed a modified-English test in math for Maine. (Maine has since stopped using that test and instead modifies the English of some items on tests for all students.)

“It’s unfortunate few states are doing it,” Mr. Stansfield said. “Somehow the word just isn’t getting out there.”

Mr. Stansfield’s company also translates tests into other languages, but he says such tests can serve only a small number of students in most states.

NCLB a Factor

Before the No Child Left Behind Act became law four years ago, many states exempted students with limited English skills from testing for the first three years they attended U.S. schools.

Now, all English-learners must take state math tests the first time they are given after the students enroll in the United States. During the first year they’re enrolled, they can take English-proficiency tests instead of state reading tests. But in subsequent school years, they must take the regular state tests in both reading and math.

The federal law says English-learners must be tested in a “valid and reliable” manner. It says that states must provide accommodations for such students and, “to the extent practicable,” test them in a “language and form most likely to yield accurate data on what students know and can do.”

Much of the legal language on including English-language learners in testing is not new in the No Child Left Behind Act. But the requirements have received more attention because of the stepped-up accountability the law imposes on schools for the performance of such students.

Ms. Rivera and Mr. Abedi say many states have provided English-learners the same accommodations they give students with disabilities, a group with a longer record of using testing accommodations. But many of those accommodations may not be appropriate for ELL students, the two experts say.

Ms. Rivera has identified 44 kinds of accommodations that states make available to English-learners, including giving tests in familiar rooms by familiar people and providing clarifications of directions.

State policies on test accommodations have become controversial, as some schools have received sanctions for not making adequate yearly progress under the federal law for their ELL-student subgroups.

In both California and Pennsylvania, school districts have sued their states over testing accommodations for English-language learners. Those lawsuits are pending.

State Practices Vary

Illinois, Kansas, Oregon, and Virginia administer modified-English versions of their state math tests for some English-language learners. Delaware and Minnesota are also preparing math tests that will contain test items with simplified English.

Oregon and Kansas offer versions of their reading tests in modified English, despite researchers’ warnings that simplifying the wording of reading tests may defeat their purpose.

Native-language tests are the commonest form of testing accommodation for English-learners, according to Mr. Abedi. As of last summer, 11 states provided such tests statewide and used them to meet NCLB requirements.

Most of those exams are translations of the regular tests. Minnesota, one of the 11 states, stopped providing tests translated into students’ native languages as of this school year and plans to develop math tests in simplified English by spring 2007.

In the end, says Mr. Stansfield, state officials base their policies about accommodations on more than research. In some states, he says, it’s politically sensitive to provide versions of tests in students’ native languages because of sentiments that English should be the country’s official language.

“In the final analysis,” Mr. Stansfield said, “cost becomes an issue, and staying out of trouble becomes an issue.”