The findings reviewed above do not apply to translations of tests. Use of a test in a new linguistic culture requires that it be redeveloped from the start. One reason for the early success of the Stanford-Binet Intelligence Scale was that Terman reconceptualized it for the United States, reexamining Binet's theory of intelligence, writing and testing new items, and renorming the scales (Reynolds, Lowe, et al., 1999).
Terman's work was an exception to a rule of simple translation of the Binet Scales. Even today, few researchers are versant in procedures for adapting tests and establishing score equivalence. Nonetheless, the procedures are available, and they increase the validity of the adapted tests (Hambleton & Kanjee, 1995). Adaptation of educational and psychological tests most frequently occurs for one of three reasons: to
facilitate comparative ethnic studies, to allow individuals to be tested in their own language, or to reduce the time and cost of developing new tests.
Test adaptation has been common for more than 90 years, but the field of cross-cultural and cross-national comparisons is relatively recent. This field presently focuses on development and use of adaptation guidelines (Hambleton, 1994), ways to interpret and use cross-cultural and cross-national data (Hambleton & Kanjee, 1995; Poortinga & Malpass, 1986), and especially procedures for establishing item equivalence (Ellis, 1991; Hambleton, 1993; Poortinga, 1983; Van de Vijver & Poortinga, 1991). Test items are said to be equivalent when members of each linguistic or cultural group who have the same standing on the construct measured by the tests have the same probability of selecting the correct item response.
The designs used to establish item equivalence fall into two categories, judgmental and statistical. Judgmental designs rely on a person's or group's decision regarding the degree of translation equivalence of an item. Two common designs are forward translation and back translation (Hambleton & Bollwark, 1991). In the first design, translators adapt or translate a test to the target culture or language. Other translators then assess the equivalency of the two versions. If the versions are not equivalent, changes are made. In the second design, translators adapt or translate a test to the target culture or language as before. Other translators readapt the items back to the original culture or language. An assessment of equivalence follows. Judgmental designs are a preliminary approach. Additional checks, such as DIF or other statistical analyses, are also needed (Reynolds, Lowe, et al., 1999).
Three statistical designs are available, depending on the characteristics of the sample. In the bilingual examinees design, participants who take both the original and the target version of the test are bilingual (Hambleton & Bollwark, 1991). In the source and target language monolinguals design, monolinguals in the original language take the original or back-translated version, and monolinguals in the target language take the target version (Ellis, 1991). In the third design, monolinguals in the original language take the original and back-translated versions.
After administration and scoring, statistical procedures are selected and performed to assess DIF. Procedures can include factor analysis, item response theory, logistic regression, and the Mantel-Haenszel technique. If DIF is statistically significant, additional analyses are necessary to investigate possible bias or lack of equivalence for different cultures or languages.
A study by Arnold, Montgomery, Castaneda, and Longoria (1994) illustrates the need to evaluate item equivalence. The researchers found that acculturation affected several subtests of the Halstead-Reitan neuropsychological test when used with unimpaired Hispanics. By contrast, Boivin et al. (1996) conducted a study with Lao children and identified variables such as nutritional development, parental education, and home environment that may influence scores on several tests, including the K-ABC, the Tactual Performance Test (TPT), and the computerized Tests of Variables of Attention (TOVA). These results suggest that tests can potentially be adapted to different cultures, although the challenges of doing so are formidable. Such results also show that psychologists have addressed cultural equivalence issues for some time, contrary to the view of Helms (1992).
Was this article helpful?
Enchanted Learning Experiences -Why They Should Be The Norm For Our Children. The latter part of the twentieth century has seen more discoveries about the human brain than in all previous history of mankind. It is as though we have been paddling in the shallows of a vast ocean hitherto unaware of its existence.