Test Adaptation/Translation Methods

Introduction

During the decades the interest in cross-cultural studies has increased. These studies involve groups without a common language. The traditional approach of making close (‘literal’) translations of (usually Western) source instruments for all target languages, though still widely used, has been challenged. It is appreciated that an exclusive focus on the linguistic aspects of a translation does not address the question, to what extent the item contents are adequate in all target languages. Similarly, it is increasingly appreciated that in order to translate psychological instruments linguistic competence is necessary though insufficient (Geisinger, 1994; Hambleton, Merenda & Spielberger, 2001). Also needed are knowledge of the target cultures [Page 961]and expertise in designing psychological instruments. Modern translation projects pay much attention to the question of how cultural, psychometric, and linguistic knowledge can be combined so as to optimize their adequacy in each of these domains (e.g. Schroots, Fernández-Ballesteros & Rudinger, 1999).

Adopt or Adapt?

Linguistic, psychometric, and cultural criteria may lead to different translations. Linguistic criteria involve equality of, among other things, semantic meaning, comprehensibility, readability, and style. Psychometric criteria involve the need to follow good practice of item writing and to assess the adequacy of test translations using statistical means (analysis of bias and equivalence: (see entry on ‘Item Bias’). Cultural criteria involve the appropriateness of item contents and compliance with local norms and habits. Depending on the degree of convergence of the three criteria, two different options in test translation may be appropriate (Van de Vijver, Fons & Leung, 1997; Van de Vijver & Poortinga, 2001).

The first option is called adoption. It amounts to a close translation of an instrument in a target language. This option is the most frequently chosen in empirical research because it is simple to implement, cheap, has a high face validity, and retains the opportunity to compare scores obtained with the instrument across all translations. The aim of these translations often is the comparison of averages obtained in different cultures (Does culture A score higher on construct X than does culture B?). Close translations have an important limitation: they can only be used when the items in the source and target language versions have an adequate coverage of the construct measured on items showing bias. Standard statistical techniques for assessing equivalence (e.g. factor analysis, see Behling & Law, 2000, and Van de Vijver, Fons & Leung, 1997) should be applied to assess the similarity of constructs measured by the various language versions. However, even when the structures are identical, there is no guarantee that the translations are all culturally viable and that a locally developed instrument would cover the same aspects.

The second (and more modern) option is labelled adaptation. It usually amounts to the close translation of some stimuli that are assumed to be adequate in the target culture, and to a change of other stimuli when a close translation would lead to linguistically, culturally, or psychometrically inappropriate measurement (e.g. a coping questionnaire has the item ‘watches more television than usual’ to express the idea of seeking distraction. In areas without electricity or a low density of televisions this item should be adapted. A behaviour could then be identified that comes close to the original in terms of psychological meaning.). The validity of the new measure can be examined by scrutinizing the nomological network of the instrument (e.g. by correlating scores obtained with the instrument with scores on locally developed measures). Comparisons of scores obtained with adapted instruments, using a t test or analysis of variance, are impossible. However, recent advances in psychometrics such as Item Response Theory (see entry on ‘Item Response Theory’) and Structural Equation Modelling (e.g. Byrne, Shavelson & Muthén, 1989) allow for numerical score comparisons across language versions even in the cases of test adaptations.

...