An Investigation Into The Construct Of Gapped Text Task In New HSK Level 6

The gapped text task was used as a measure of text reading comprehension for the first time in a large-scale language test of Chinese by new HSK. Aimed to provide warrants for score interpretation and justify the construct validity of the task, this thesis investigates the test-takers'responding processes from the perspective of cognitive psychology via mixed method of introspection and eye-tracking technique.The research was composed of two parts:Part I attempted to define the construct of the task via think-aloud and immediate retrospective method. Nine CSL volunteers completed 4 text gap tests with 20 items, the introspective data of successful and unsuccessful candidates was analyzed qualitatively first to explore the behaviours, knowledge, skills and abilities employed during the process and then analyzed quantitatively to make preliminary validation. Part II attempted to validate the construct defined in Part I via a mixed method of stimulated recall and eye-tracking. Two groups of Malaysian students (n=22/group) at different language proficiency levels took an on screen test consisting of 2 text gap tasks, one of which contained an additional distractor. A variety of data is analyzed quantitatively:(1) the performance of the candidates, (2) the stimulated recall interview data, and (3) 76 eye-tracking traces data.Part I defined the task construct from both behavioral and trait perspective via the analysis of responding processes of candidates. Behavioral definition of the construct was the ability to establish the framework of the text based on meaning proposition encoding and inference the relationship of key sentences logically and rhetorically. To accomplish such activities, one should have proficient lower-level language processing skills, enhanced sensitivity to structural centrality and the ability to establishing proper standards of coherence, which was the trait definition of the construct. The paper also clarified the operational definition of the construct mentioned above based on Part I research and previous studies. Part II validated the behavioral construct from statistical analysis of the test-takers'verbal protocols and partially validated the trait construct by correlation analysis of eye-tracking data and response efficiency. The findings of research provided backups for warrants to support the score interpretation claim generally, but rebuttals were also found. Through the examination of the rebuttals, the paper identified several task characteristics which devalidated the task and thus refined the score interpretation. In addition, the paper found that what the task measured varied according to language proficiency to some extent. For those from the advanced group, the score variation reflected the higher-level comprehension ability such as sensitivity to structural centrality, for those from the lower group it reflected the basic language processing skills to a greater degree. By adding distractors, the construct irrelevant behaviours were restrained in both groups, and better discrimination about higher-level comprehension ability could be made but only in advanced group.The findings of the paper could be considered to inform inferences of construct validity of gapped text task as well as language instruction. The theoretical framework for defining task construct and the research method to justifying the construct validity could shed light on future HSK validation study. And this paper is amongst the first pieces of research to apply eye-tracking in the area of Chinese language testing. Thus it serves as a methodological exploration in the use of information from eye-tracking traces.