一个基于综合印象评分法的作文分事后调整模型  

A Writing Score Adjustment Model Based on Holistic Approach

作　　者：朱正才[1] 杨惠中[1] 

机构地区：[1]上海交通大学外国语学院,上海200240

出　　处：《心理科学》2005年第6期1459-1462,共4页Journal of Psychological Science

基　　金：国家社科基金项目(03BYY016)的资助

摘　　要：对大学英语四、六级考试作文评分进行了详细的描述,重点介绍了作文分的事后调整的原理和方法,并且给出了一个基于线性等值原理的数学推导。认为其数学模型主要是一个运用了“极大似然估计法”和“正态分”概念的统计模型。“评分标准的制定”、“用参照样卷来校准阅卷员对作文评分标准的掌握尺度”以及“阅卷员的培训和考核”构成了大学英语四、六级考试作文信度的基石;而作文分调整中基于“随机分发作文卷”、“客观题分数与作文分相关”以及“评分前后一致性”的统计方法则对出现系统性误差的阅卷员的评分结果进行了事后的校正。还提出如果拥有往次考试总体作文分均值数据,使用“加权移动平均法”可以实现对作文分的跨考次平衡。This paper presents a detailed description of a writing score adjustment model of CET and a pilot study based on the CET real data. There are generally two approaches to marking writing： a holistic one and an analytic one. In analytic marking, a performance is broken down into several separate components such as content, organization, cohesion, grammar and vocabulary, etc., which may be given different weightings as well. The over mark given to the performance is the combination of the ratings on the different components. In holistic marking, examiners are asked to evaluate an essay thurough the overall impression rather the than particular aspects of the examinee＇s production. Specifications of the marking scheme, the careful selection of range-finders, and the training of the markers, all form a firm basis to ensure the scoring reliability in the CET operation. The writing score adiustment model, which depends mainly on the random distribution of writing scripts, correlation between scores for objective items and scores for writing, and consistency in marking, filters out system errors which may occur in the marking process.

关 键 词：信度 作文分调整 正态分 

分 类 号：B842[哲学宗教—基础心理学]