OBJECTIVE To investigate the factor structure of the verbal paired-associates (VPA) subtest in the WMS-III using a theoretically driven model of semantic processing previously found to be well-fitting for the WMS-IV version of the test. METHOD Archival data were used from 267 heterogeneous neurosciences patients and 223 seizure disorder patients who completed the WMS-III as part of a standard neuropsychological evaluation. Confirmatory factor analysis was used to test theoretically driven models for VPA based on principles of semantic processing. Four nested models of different complexities were examined and compared for goodness-of-fit using chi-squared difference testing. Measurement invariance testing was conducted across heterogeneous neuroscience and seizure disorder samples to test generality of the factor model. RESULTS After removing items with limited variability (very easy or very hard; 12 of 40 items), a four-factor model was found to be best-fitting in the present patient samples. The four factors were "recreational", "functional", "material", and "symbolic", each representing semantic knowledge associated with the function of the target word referent. This model subsequently met the criteria for the strict measurement invariance, showing good overall fit when factor loadings, thresholds, and residuals were held to equality across samples. CONCLUSIONS The results of this study provide further evidence that "arbitrary" associations between word pairs in VPA items have an underlying semantic structure, challenging the idea that unrelated hard-pairs are semantic-free. These results suggest that a semantic-structure model may be implemented as an alternative scoring in future editions of the WMS to facilitate interpretation.

中文翻译：

---

WMS-III 语言配对关联的潜在语义结构。

目的使用先前发现非常适合 WMS-IV 版本测试的语义处理的理论驱动模型来研究 WMS-III 中语言配对关联 (VPA) 子测试的因子结构。方法 使用来自 267 名异质神经科学患者和 223 名癫痫症患者的档案数据，他们完成了 WMS-III 作为标准神经心理学评估的一部分。验证性因素分析用于测试基于语义处理原理的 VPA 的理论驱动模型。使用卡方差异检验检查并比较了四个不同复杂性的嵌套模型的拟合优度。在异质神经科学和癫痫症样本中进行了测量不变性测试，以测试因子模型的一般性。结果 在删除可变性有限的项目（非常容易或非常困难；40 个项目中的 12 个）后，发现四因素模型最适合目前的患者样本。这四个因素是“娱乐性”、“功能性”、“物质性”和“象征性”，每个因素都代表与目标词所指功能相关的语义知识。该模型随后满足了严格测量不变性的标准，当因子载荷、阈值和残差在样本间保持相等时，显示出良好的整体拟合。结论 本研究的结果提供了进一步的证据，即 VPA 项目中词对之间的“任意”关联具有潜在的语义结构，挑战了不相关的硬对是无语义的想法。