Abstract

To challenge the commonly made assumption in cross-national survey projects that close translation yields more comparable data than adaptation, we implemented a translation experiment in the CROss-National Online Survey Panel. The English source questionnaire was split into three batches of 20 items each and was translated by three translation teams into Estonian and three teams into Slovene. The teams received specific instructions on how to translate each batch (either closely or adaptively) so that, by design, the teams translated two batches following one approach and one following the other approach. Respondents in the two countries (Estonia and Slovenia) were randomly assigned to three distinct questionnaire versions based on the same source questionnaire, each consisting of translations by all three teams and including close and adaptive translations. We developed an analytical framework to assess the translation potential of the source items (i.e., all theoretically possible translations of a specific item) and the actual translation scores (i.e., the degree of closeness vs. adaptiveness of a specific translation). We show that some items are more sensitive to the wording (small linguistic changes result in a different response behavior) while others are more robust (the meaning of the concept is retained despite linguistic changes).

摘要

为了挑战跨国调查项目中的普遍假设，即封闭翻译产生的可比性数据比适应性更强，我们在CROss-National Online Survey Panel中实施了翻译实验。英文来源调查表分为三批，每批20项，由三个翻译小组翻译成爱沙尼亚语，三个小组翻译成斯洛文尼亚语。团队收到了有关如何（紧密地或自适应地）翻译每个批次的具体说明，因此，根据设计，团队按照一种方法翻译了两批，而另一种方法则翻译了一批。根据相同的来源问卷，两个国家（爱沙尼亚和斯洛文尼亚）的受访者被随机分配到三个不同的问卷版本中，每个翻译由三个团队的翻译组成，包括紧密翻译和自适应翻译。我们开发了一个分析框架来评估源项目的翻译潜力（即，特定项目的所有理论上可能的翻译）和实际翻译分数（即，特定翻译的亲和度与适应性）。我们显示，有些项目对措辞更敏感（小的语言变化会导致不同的响应行为），而另一些则更加健壮（尽管语言有所变化，但概念的含义仍然保留）。