矩阵推理测验中的错误类型分析  

Error Analysis in Matrix Reasoning Tests

作　　者：李中权[1] 孔明[2] 张厚粲[3] 周仁来[3] 

机构地区：[1]南京大学社会学院心理系,南京210093 [2]苏州大学医学院临床心理学研究所,苏州215123 [3]北京师范大学心理学院,北京100875

出　　处：《心理科学》2010年第3期663-665,共3页Journal of Psychological Science

基　　金：南京大学人才引进培养基金(项目编号010922410108);应用实验心理北京市重点实验室(北京师范大学)开放研究课题的资助

摘　　要：计算机自动化项目生成已逐渐被用来应对矩阵推理测验曝光的问题,然而对诱选项该如何设置却没有很好的文献支持。现有与矩阵推理测验(如瑞文测验)相关的大量研究中,研究者都只使用正确作答数或者错误率,对个体所犯错误信息的深入探查较少。通过对340名大学生在矩阵推理测验作答时产生的错误答案的编码分析,发现可将所有被试的错误模型分成4种类型,其中出现最多的是不完整解答错误(IC),忽视项目涉及的部分规则。中高能力者的错误都集中在IC类型上,而低能力者在四种错误类型上的分布相对平均些。这一结果支持了矩阵推理测验的认知模型,也为计算机自动化项目生成的诱选项编制提供参考。Computerized automatic item generation （AIG） was gradually employed in dealing with item exposure in matrix reasoning tests. However, how to design better distracters was still unclear. Most studies related to matrix reasoning tests （MRT, represented by Raven Matrices） used total number correct or incorrect rates as the variables of interest, but few conducted analysis of the types of errors that individuals made. In the present study, 340 undergraduates were asked to give answers to 24 MRT items, and wrong responses were coded and analyzed. The results indicated that all of their errors could be classifled into four categories, among which the most frequent was the incomplete solution error （ i. e. , omitting some rules）. Errors made by participants witb high or medium ability were mainly IC while errors of low-ability participants were nearly averagely distributed in four types. The results not only partly supported the cognitive model of MRT, but also helped design distracters in computerized AIG.

关 键 词：矩阵推理测验 计算机自动化项目生成 错误分析 诱选项 工作记忆 

分 类 号：B841.7[哲学宗教—基础心理学]