A translated test should always provide information on the reliability and validity of the translation (Geisinger, 1992). Establishing the reliability and the validity of a translated test follows the same statistical format as that provided for any psychological instrument. Reliability statistics aimed at determining the internal consistency of the translated instrument must be included. Internal consistency coefficients should approximate those observed in the test construction of the original measure. Test-retest measures of reliability can be used to determine the stability of the measure over time. Particular interest may be focused, however, on certain types of validation issues. For example, a translated test may be administered to groups with known characteristics in the target population. If the test accurately predicts the known characteristics to a comparable degree as its performance in the normative group, it demonstrates sound criterion validity. In this manner, the translated test is evaluated on its basis to predict behavior in the target population to the same degree as observed in the norm group (Butcher et al., 1998).
Another important validation approach applied to translated tests focuses on determining the factor structure of the translated instrument. In established tests with well-documented factor structures, construct validation of the translated test is possible. In this approach, a factor analysis is first completed with the translated version of the test. A comparison is then made between the resulting factor components derived from the translated test, and those of the original instrument. To the extent the translated test factor structure is similar to that of the original, greater confidence is assured that the translated test and the original test are measuring similar constructs (Dana, 1993). A similar type of factor associated with cultural equivalence is conceptual validity, which is defined as the familiarity of different cultural groups with item content of a test and similar comprehension of their item meanings (Helms, 1992). If dissimilar factor structures are observed, construct bias may be present. Construct bias has been described as being more likely with translations of instruments from a source to a target language or culture, than when an instrument is developed simultaneously for different cultures or languages (Van de Vijver & Hambleton, 1996).
Finally, convergent validity may be used when a test is measuring a trait. A convergent validity approach can be taken where bilingual populations are available. This approach evaluates test scores on the translated and original instruments to determine if they seem to be measuring the same underlying trait, but not in the same manner as does factor analysis with construct validation.
Was this article helpful?