In linguistic equivalence, the intent is to ensure that performance requests are identical across cultures, or where regional linguistic variations are observed (Drasgow & Hulin, 1987; Hulin, 1987). Or, as Helms (1992) has suggested, is the language used in test items equalized such that it holds the same meaning in the cultural groups being examined? In terms of methods to increase the likelihood of linguistic equivalence, several have been identified in the literature. Possibilities for test translation include (a) literal translation of the instrument; (b) adaptation of parts of the instrument; and (c) assembling of an entirely new instrument (Van de Vijver & Hambleton, 1996). One method frequently utilized involves translation of the test into the language of the culture, followed by back translation (Brislin, 1970, 1980, 1986; Ellis, 1989). It is of interest, however, that in the field of professional translation, the translationback translation procedure, unfortunately, is rarely used. Professional translators, typically, first translate, then in teams of competent bilinguals, utilize judgmental methods to assess the accuracy of the new language version (Van de Vijver & Leung, 1997). However, the translation-back translation approach is particularly valuable to researchers and clinicians who do not have proficiency in the translated language, as it can assure some control over the eventual product when the back translation occurs to the source language.
Additionally, in completing a translation, one that is "generic" in nature and broadly reflects the culture of interest would appear most potentially useful (Arnold, Cuéllar, & Guzmán, 1998). Different cultures and subcultures make use of various terms in referring to the same concept. For example, a child may be called 'chilpayate,' 'escuincle,' or 'cabro' in different regions of Mexico, Central, and South America; however, the term 'niño/a' would be a more generic and universally acceptable word choice. This translation approach assumes a rationale that the less complex the verbal performance requests the greater the likelihood it will be meaningful to a larger number of potential patients in the target population. Even so, at times, selection of appropriate words for test items may prove difficult. To illustrate, in some Central and South American countries, corn is referred to as 'choclo,' whereas, in Mexico, 'choclo' is a type of shoe, and the term for corn is either 'elote' or 'maíz-' To complicate test translation issues even further, a simple object such as a pen may have a variety of names, such as, 'pluma,' 'lápiz pasta,' 'lapicero,' or 'bolígrafo.' Following translation, an independent back translation is typically involved. The back translation evaluates translation quality, specifically comparing the original English and the back-translated English (Brislin, 1986; Hulin, Drasgow, & Komocar, 1982; Sinaiko & Brislin, 1973). Subsequent to the back translation of the instrument, review by a panel that adequately reflects the diversity of the targeted population is an important next step. Revisions to the translation are reiterated until reviewers are satisfied that a reliable test has been adapted. When reviewers can agree on adequate similarity of performance requests or statements representing personality traits, field trials with a sample of consumers who represent the population to be tested is conducted.
During these field trials, another method for validation of the translation can be applied that involves establishing functional equivalence by performance testing and knowledge testing (Sinaiko & Brislin, 1973). Functional equivalence refers to dissimilar behavioral competencies that are learned, but aimed at coping with similar problems (Dana, 1993). That is, individuals in dissimilar cultures may develop different coping styles to deal with similar problems. Performance testing includes observing the consumer perform a task using translated instructions, and to the extent that the consumer demonstrates competence in completing the task, the instrument demonstrates equivalency (Sinaiko & Brislin, 1973). The translated test should elicit completely accurate scores in the target culture if the test is sound, and should be independent of individual differences (Sinaiko & Brislin, 1973). Performance testing would appear most appropriate for tests measuring functional abilities that are observable. The second approach to validating translations seems most compatible with tests of functional abilities that are not directly observable, for example, reading comprehension. This approach, labeled "knowledge testing" by Miller and Beebe-Center (1956; Sinaiko & Brislin, 1973), involves targeted consumers reading a translated passage, followed by answering questions concerning the passage. An accurate translation is produced when the target group is able to answer all the questions on the test.
Twelve general guidelines for producing good translations of tests were listed in Brislin, Lonner, and Thorndike (1973), including (a) constructing items using brief, clear and simplified sentences, at times involving elimination of words from the original test to avoid compound thoughts in an item; (b) grammatically phrasing items in the active tense; (c) using nouns in a repetitive manner, if necessary, rather than resorting to pronouns; (d) avoiding the use of metaphors, or regionally specific phrases, idioms, or colloquialisms; (e) not using the subjunctive tense in translated items; (f) using additional phrases to insure comprehension of key meanings in item content; (g) not utilizing items that include adverbs and prepositions; (h) avoiding item content that includes possessive forms of words; (i) being specific and concrete in item content; (j) not using vague descriptors in item content; (k) familiarizing the translator with test item content; and, (1) avoiding inclusion of more than one verb in item content where the two suggest differing actions. Behaviors that have a very low probability of occurring in the target group, or have never been exhibited, should not be included in item content (Hulin, 1987). Translated items that are spelled identically or similarly but have different meanings may cause confusion in careless translators or bilingual test takers and contribute to inequiva-lencies. By way of illustration, translating 'to assist' from English to Spanish could be easily mistaken for 'asistir,' which means to be present at a function. Similarly, translating "to attend" from English to the Spanish verb 'atender,' meaning to help in a situation, may present problems. Also, poor translations may be due to language that is unique in meaning to a particular culture or region. In the Mexican culture, a "china" is a female of Chinese ethnicity, whereas, in Puerto Rico a "china" is an orange fruit.
Translation of a test from one language to another has the potential for many sources of error, which reduce accuracy and diminish the reliability of the measurement. Potential problem sources include (a) a sound translation can be misinterpreted by an incompetent back translator; (b) a poor translation can be completed with subsequent problems with the back translation. In the latter case, a good translator during the back-translation process may make adjustments to make the back translation more valid, even though the original translation is not a good one (Bontempo, 1993). These potential problems with the translation-back translation procedures require that more measures of equivalence be completed (Sinaiko & Brislin, 1973).
Three important, but rarely utilized designs have been proposed for determining the usefulness of test translations (Hambleton, 1993, 1994; Manuel-Dupont et al., 1992): (a) both target and source versions of the translated test are taken by bilinguals; (b) original and back-translated versions of the test are taken by source language monolinguals; and (c) monolinguals in both target and source language take the test in their respective language. Following implementation of these designs, to determine linguistic equivalence of tests, statistical methods are applied.
Was this article helpful?