Department of Speech and Hearing Sciences, Portland State University, OR.
Hospital General Dr. Manuel Gea González, Instituto Mexicano para la Atención del Desarrollo del Niño, Mexico City, Mexico.
J Speech Lang Hear Res. 2019 Jan 30;62(1):70-83. doi: 10.1044/2018_JSLHR-L-18-0110.
Purpose The purpose of the current study was to examine construct validity evidence for the following 4 lexical diversity measures for use in Spanish narrative samples of children with developmental language disorder and typical language development (TLD): (a) type-token ratio (TTR; Chotlos, 1944 ; Templin, 1957 ), (b) hypergeometric distribution D ( McCarthy & Jarvis, 2010 ), (c) measure of textual lexical diversity (MTLD; McCarthy, 2005 ), and (d) moving-average TTR (MATTR; Covington & McFall, 2010 ). Method A multiple-group, confirmatory factor analytic stepwise procedure was applied to examine factorial invariance in a sample of 435 monolingual Spanish-speaking children (335 with TLD and 100 with specific language impairment). Results MATTR was the strongest indicator of the lexical diversity of a sample and demonstrated measurement invariance across children with TLD and specific language impairment. MTLD scores included some noise at different levels as a function of the length of a sample, especially for short language samples, but differences in relation to MATTR were relatively small. The validity of hypergeometric distribution D and TTR score interpretations was low, as the scores covaried systematically with length, with TTR being the poorest measure of lexical diversity. Conclusion Results support the use of MATTR and MTLD as measures of lexical diversity and should assist clinicians and researchers in the selection of lexical diversity measures to minimize bias in score interpretations when conducting language sample analyses.
目的 本研究的目的是检验以下 4 种词汇多样性测量方法在西班牙语发展性语言障碍（DDL）和典型语言发育（TLD）儿童叙事样本中的结构效度证据：（a）类型-令牌比（TTR；Chotlos，1944；Templin，1957），（b）超几何分布 D（McCarthy & Jarvis，2010），（c）文本词汇多样性测量（MTLD；McCarthy，2005），以及（d）移动平均 TTR（MATTR；Covington & McFall，2010）。
方法 采用多组验证性因子分析逐步程序，对 435 名单语西班牙语儿童样本（335 名 TLD 和 100 名特定语言障碍）进行因子不变性检验。
结果 MATTR 是样本词汇多样性的最强指标，且在 TLD 和特定语言障碍儿童中具有测量不变性。MTLD 分数在不同长度的样本中存在一定的噪声，尤其是在较短的语言样本中，但与 MATTR 的差异相对较小。超几何分布 D 和 TTR 分数解释的有效性较低，因为这些分数与长度系统相关，其中 TTR 是衡量词汇多样性的最差指标。
结论 结果支持将 MATTR 和 MTLD 作为词汇多样性的测量方法，这将有助于临床医生和研究人员在进行语言样本分析时选择词汇多样性测量方法，以最大程度地减少评分解释中的偏差。
