Scholarlink

<p>With few exceptions, modern psychometric concepts and methods are not part of the “standard” toolkit used to analyze linguistic data. Data from <i>Kam</i> (or <i>Dong</i>), an under-studied language of China with a very complex tone system, is used to illustrate how a psychometric approach can help investigate inter-individual differences in the perception of tone contrasts in an AX task, and the factors that may influence them, including working memory, age, gender, and years of formal education. We show that these methods uncover an unexpected structure of the responses, as some items designed as “different” based on the language’s phonology were instead perceived as “same”, prompting a recoding of these items in line with their perception. The subsequent use of % correct responses, Signal Detection Theory, multiple regression, mediation, and path analysis, found a complex network of influences on the AX task: there are direct effects of age, education and working memory, but only an indirect effect of gender. Therefore, even such a “simple” AX task can benefit from this approach, and we argue, in this primarily methodological paper which also includes directly usable extensive computer code in <Emphasis FontCategory="NonProportional">R</Emphasis>, that modern psychometrics should be more widely used in linguistics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Psychometric methods and signal detection theory uncover subtle differences in the perception of tone contrasts in speakers of Kam (Dong)

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月9日

查看全文
  • Dan Dediu,
  • Luchang Wang,
  • Patrick C. M. Wong,
  • Manxiang Wu

摘要

With few exceptions, modern psychometric concepts and methods are not part of the “standard” toolkit used to analyze linguistic data. Data from Kam (or Dong), an under-studied language of China with a very complex tone system, is used to illustrate how a psychometric approach can help investigate inter-individual differences in the perception of tone contrasts in an AX task, and the factors that may influence them, including working memory, age, gender, and years of formal education. We show that these methods uncover an unexpected structure of the responses, as some items designed as “different” based on the language’s phonology were instead perceived as “same”, prompting a recoding of these items in line with their perception. The subsequent use of % correct responses, Signal Detection Theory, multiple regression, mediation, and path analysis, found a complex network of influences on the AX task: there are direct effects of age, education and working memory, but only an indirect effect of gender. Therefore, even such a “simple” AX task can benefit from this approach, and we argue, in this primarily methodological paper which also includes directly usable extensive computer code in R, that modern psychometrics should be more widely used in linguistics.