An Empirical Study Of The Validity Of Objective And Subjective Reading Tests

Nowadays, the issue of the validity of test approaches of reading comprehension tests has been a heated topic in language testing. Yet, there are a number of myths quite widely held about them. So the aim of this study was to clarify, to some extent, these myths by systematic research.In this thesis, we elaborated the advantages and disadvantages of multiple-choice questions (MCQs) and short answer questions (SAQs), objective testing and subjective testing as well as the issue of validity. Based on and supported by them, the specific aim of this study was to investigate this issue, to conduct a comparative study of MCQs (the representative test approach of objective testing) with SAQs (the representative test approach of subjective testing) in terms of validity to verify our hypothesis that short answer questions would have more validity than multiple-choice questions in reading comprehension tests, by comparing the effects of MCQs and SAQs on testing language learner's reading comprehension.With 120 middle school students as its participants and the reading comprehension tests, questionnaire, interview and participants' scores of the English examination given in the last term as its material, this research had two factors in accordance with the design of it: (A) test approach (within-subject factor), that is MCQs and SAQs, and (B) proficiency level (between-subject factor), that is, we assigned our participants randomly into two parallel groups. Then each group was accordingly divided into three subgroups, namely, good, fair and poor students according to their scores of the English examination given in the last term. Thus, the purpose of this research was of three effects, two of which were main effects: first, test approach's main effect, that is, the effect of two different test approaches ontesting reading comprehension (if there was any difference between two test approaches in testing reading comprehension); second, the main effect of participants' English proficiency level on testing reading comprehension (if there was any difference among three subgroups of each random group in their reading comprehension performance); and third, the possible interaction between test approach and proficiency level due to the design of two factors. The result could be detailed as follows:Test approach's main effect was not significant, that is, there was no significant difference between the two test approaches, MCQs and SAQs, in participants' reading comprehension performance; second, the main effect of participant's English proficiency level on testing reading comprehension was significant, that is, there were significant differences among three subgroups of each random group in their reading comprehension performance. And third, there was a perceived significant interaction between test approach and proficiency level, that is, there was an interactive effect of proficiency level in relation to participants' reading test performance perceived in the two test approaches, MCQs and SAQs. From these, we could conclude that, within two test approaches, MCQs could not show significant difference in distinguishing good, fair and poor students while SAQs could show significant differences among good, fair and poor students. A post hoc showed that significant differences existed between good and fair, fair and poor, good and poor students in the employment of SAQs, that is, SAQs showed significant differences in distinguishing both good and fair, fair and poor, good and poor students in testing reading comprehension. From it we found out that SAQs stood a greater chance than MCQs to distinguish good, fair and poor students. In other words, SAQs were more valid than MCQs.We further analysed the data obtained to examine correlations between the scores of MCQs and the proficiency scores (participants' scores of the English examination given in the last term), SAQs and the proficiency scores. The result could be summarized as: participants' scores of MCQs were positively correlated (or consistent) with the proficiency scores, and on the...