A Study On The Validity Of Writing Test In NMET Hunan2011

Writing ability is one of the most important components of language ability. Compared with listening and reading, writing has long been considered as the most difficult and important one in language testing. Researches on the writing test of NMET mainly focus on defining the construct of writing and scoring reliability while research on the designing of writing tasks, especially on item types is relatively infrequent. The direct essay-task approach has high validity, but its drawbacks are obvious. It is subject to such factors as "content variance" and "scoring variance".The objective approach (mainly refers to multiple-choice item type) though is not easily affected by raters, its validity has been doubted by many psychometricians. What kind of writing test on earth can better measure candidates’writing construct? With the progressive development of reform in English curriculum and the gradually perfecting of NMET in some provinces, several new item types of writing have come into being since2004. Bearing in mind the shortcomings and strengths of above methods, with the purpose to improve the quality of the writing test in NMET, Hunan province designed two read-to-write writing tests. They have come along with the conventional composition item to form a combined style of writing test, aiming to lay particular emphasis on the measurement of test takers’ability to get and process information, then to express.This study theoretically based on Linguistics and Language Testing was designed to investigate the validity and reliability of this writing test by methods of SEM and correlation analysis. By answering the following3questions:(1). How valid is Section C?(2). How are the Section A and Section B related to Section C?(3). Is the inclusion of Section A and Section B in the writing part reasonable? It hopes to find some evidences to support this writing test type as a more valid and effective way of testing writing ability. Finally, according to the data results, we found that: Section C has good construct validity, there is a strong correlation between each two sections, this combined way of writing test is reasonable.