Strategy Use In Answering Local-level Reading Comprehension Questions In Multiple-choice Format:A Multi-sample Sem Approach

This study examines test takers’ strategy use in answering LRCQ,i.e.,Local-level Reading Comprehension Questions,in the context of CET-6,and aims to explore the relationship between strategy use and test performance through the application of structural equation modelling(SEM).LRCQ is an important and common item type in high-stakes reading test,which requires test takers to retrieve detailed information located in specific text segments.In response to Anderson,Bachman,Perkins and Cohen’s appeal for gathering information on test-taking processes as part of construct validation,the study adopts a process-based approach to validate LRCQ through probing into test takers’ strategy use when they arrive at their answers.The study was done by examining a total of 217 Chinese EFL test-takers’ responses to a 39-item strategy questionnaire,a 15-item LRCQ test,a 10-item banked cloze and a 20-item multiple-choice cloze.The banked cloze and multiple-choice cloze were included to investigate the role lexical-grammatical ability plays in local-level reading comprehension test performance.Half of the sample are “test-wise”students and half are “test-na?ve” students,i.e.,half of the test-takers have received instructions in coaching schools,which specialize in the teaching of test-taking strategies,and the other half have not.The questionnaire was developed by drawing on Cohen’s test-taking strategy taxonomy as well as theories about metacognitive and cognitive reading strategies.Main findings of the study include: first,a third-order strategy use model for LRCQ was found to fit the data.Secondly,strategy use was found to have a relatively small effect on test-takers’ performance in answering LRCQ.Specifically,test management strategies were found to have a very small positive effect(β = 0.10)on test performance,and the heavier use of Test-wiseness strategy led to poorer performance(β =-0.09).In addition,there was no significant difference between the test-wise group and test-na?ve group in the way test-taking strategies affected their performance.Thirdly,lexico-grammatical ability had a direct and significant effect on test performance(β = 0.72),indicating language ability is the dominant contributor totest takers’ success in answering LRCQ.Findings of the study provide evidence for the validity of LRCQ in the context of CET-6.