Quantitative reanalysis of multitrait-multimethod studies

为了更好地概念化语言测试中的结构，我们重新分析和定量综合以前的多特征多方法（MTMM）的研究。我们专门寻找模型最有可能适合的数据，以及在L1和L2的文献，以调查的首选方法来定义的结构和结构的语言能力，是最支持的。我们使用验证性因素分析重新分析了来自36项研究的44个数据源。最实证支持的模型定义的结构是跨L1和L2的特质和方法模型，这表明有必要包括特质和方法定义的结构时。此外，最常见的能力结构是高阶的L1，它是可变的（单一的，相关的，或高阶）在L2。这些结果与一些调节变量有关。

To better conceptualize the construct in language testing, we reanalyzed and quantitatively synthesized previous multitrait-multimethod (MTMM) studies. We specifically searched for models most likely to fit the data well in L1 and L2 literature to investigate the preferred approach for defining the construct and the structure of language ability that is most supported. We reanalyzed 44 data sources from 36 studies using confirmatory factor analysis. The most empirically supported models for defining the construct were the trait-and-method models across L1 and L2, suggesting the necessity to include trait and method when defining the construct. Additionally, the most frequently occurring ability structure was higher-order in L1, and it was variable (unitary, correlated, or higher-order) in L2. These results were related to some moderator variables.