多维度多特征语域分析在语言测试效验中的应用——以多任务写作测试为例

Applying multi-dimensions multi-feature register analysis to language test validation:The case of multiple-writing-task assessment

多维度多特征语域分析是基于语言词汇—语法特征以及语义属性等,以因子分析的统计手段对文本语域加以研究的研究方法。该方法通过提取正负因子,可对文本的文体属性加以分类。在写作测试中,出于构念效度的考虑,命题者会利用多个写作任务来诱发足够的书面表达产出,以此更全面地、多层次地测量受试者的写作能力。然而,利用多个写作任务能否诱发不同文体属性的写作文本材料需要加以验证。多维度多特征语域分析则给这类写作测试的效度验证带来了新的视角和方法。本文从多维度多特征语言分析的简介出发,提出将这种方法应用于写作测试效度验证的必要性和可行性,并通过一项实证研究加以举例说明。本文最后提出多维度多特征语域分析方法在其他语言测试效验中的一些设想。

With factor analysis as a statistical method, the Multi-dimension Multi-feature （MD-MF） register analysis investigates texts based on the lexico-grammatical features and semantic properties, etc. Researchers, therefore, may categorize texts in light of genres by extracting positive and negative loadings of latent factors. For an enhancement of writing assessment validity, multiple writing tasks are employed by a host of English testing batteries to evaluate the multi-facets of test-takers＇ writing proficiency in an all-round manner. However, it would be hard to build validity ar- guments for whether, if so, how different writing tasks are able to elicit texts of different genres. The MD-MF register analysis provides a new perspective and approach for test validation of this kind. With an introduction to the MD-MF register analysis as a point of departure, this paper highlights the necessity and applicability of this approach in test validation, followed by an empirical study using this method as an example. In the end, how the MD-MF register a- nalysis can be prospectively applied to future test validation is briefed.