Studies of Psychology and Behavior ›› 2024, Vol. 22 ›› Issue (6): 814-821.DOI: 10.12139/j.1672-0628.2024.06.013

• ? • Previous Articles     Next Articles

Assessing the Reliability of Statistical Learning Measures: Effects of Modality, Material, and Task

Wenbo YU1, Hetong QI1, Tianlin WANG2, Dandan LIANG*,1,3()   

  1. 1. School of Chinese Language and Culture, Nanjing Normal University, Nanjing 210097
    2. School of Education, University at Albany, State University of New York, Albany 12222, United States of America
    3. Interdisciplinary Research Center for Linguistic Science, University of Science and Technology of China, Hefei 230026
  • Received:2024-09-29 Online:2024-11-20 Published:2024-11-20
  • Contact: Dandan LIANG

统计学习能力测量的信度研究:基于模态、材料特征和测试任务的证据

于文勃1, 亓鹤潼1, 王天琳2, 梁丹丹*,1,3()   

  1. 1. 南京师范大学文学院,南京 210097
    2. 美国纽约州立大学奥尔巴尼分校教育学院,奥尔巴尼 12222
    3. 中国科学技术大学语言科学交叉研究中心,合肥 230026
  • 通讯作者: 梁丹丹
  • 基金资助:
    高校哲学社会科学研究一般项目(2022SJYB0251);国家社会科学基金重点项目(22AYY013);国家社会科学基金青年项目(23CYY048)。

Abstract:

Statistical learning (SL) ability is often examined as an independent variable in predicting individual language development. However, traditional experimental tasks designed to assess group differences typically exhibit low reliability, failing to meet basic psychometric standards. This study aimed to address these limitations by synthesizing learning materials with mixed-length target structures and utilizing two testing tasks: a two-alternative forced-choice task and a familiarity rating task. Additionally, the study incorporated auditory speech and visual graphic modalities, calculating the reliability of the tasks using Cronbach’s alpha and split-half reliability. The results showed that experimental tasks utilizing mixed-length target structures demonstrated higher reliability compared to previous studies. Notably, tasks in the visual modality showed greater reliability than those in the auditory modality, and the forced-choice task yielded higher reliability than the familiarity rating task. These findings highlight the advantages of using mixed-length learning materials and visual stimuli when assessing SL ability.

Key words: statistical learning ability, reliability, two-alternative forced choice task, familiarity rating task

摘要:

统计学习能力常作为自变量预测个体语言能力的发展,但以组间差异为目标所编制的实验任务普遍信度不高,难以满足心理测量学的要求。为探究统计学习能力的测量方案,本研究使用混合长度的目标结构合成学习材料,在测验任务中对比了二选一的迫选任务和熟悉度评分任务;同时设计了听觉语音和视觉图形两种材料模态。信度指标计算了测验的内部一致性系数和分半信度。结果发现混合长度目标结构编制的实验任务的信度普遍高于以往使用等长度目标结构的任务;同时,视觉模态下的信度要高于听觉模态;前者迫选任务的信度也要高于熟悉度评分任务。综上,本研究推荐在视觉模态下,使用混合长度的目标结构合成学习材料、以迫选任务来考察被试的统计学习能力。

关键词: 统计学习能力, 信度, 迫选任务, 熟悉度评分任务