The Influence Of Oral Task Type On Test-takers' Performance

In language testing, if a feature affects test results to a significant degree, but is irrelevant to the ability being measured, it is indeed a source of measurement error and needs to be eliminated. It is then an important goal of research related to language testing to discover which test features constitute significant sources of true variance in testees' performance. Based on this idea, the present study is to investigate the impact of different tasks on the performance of test takers of both genders in oral proficiency test, and tries to find the suitable tasks for testees.To be specific, because the author chooses the three tasks in 2006 Oral test for postgraduate entrance exam for a college in Shanxi province as the research objects, which are respectively Questions and Answers, Discussions, and Oral Presentations. The present study focuses on the following research questions:·Whether the differences in the 3 tasks result in differences in testscores? If so, what and how big are the differences? Which or none of the task is best in view of test taker's performance?·Whether there is agreement between men and women on thedifferences? In other words, whether the 3 tasks are sex typed.·What causes the differences?Firstly, this paper deals with the theoretical background of performance, listing the elaborations from Hymes, Chomsky, Canale and Swain, Bachman(1990), and Bachman and Palmer (1996), and then identifies the three big aspects that may affect oral performance, which include the abilities and the knowledge we wish to measure, the means by which we observe and measure these abilities, and individual attributes that are not part of test takers' language competence. Subsequently, the author starts from the second aspect, e.g. TMF (Test Method Facets according to Bachman) and takes task as the research object and lists nearly all the possible tasks that appear in oral tests so far and oral relevant problems. During the empirical study, the author makes a quantitative and a qualitative analysis on the performances of test takers on the three tasks. Four primary approaches are adopted, including ANOVA, Correlation, and retrospective interview and triangulation. The research sample involves 30 in the quantitative study and 12 in the qualitative study with 2 invalid. The quantitative study uses ANOVA to examine the effect of gender and task on the scores and their interactive effect and the Correlation method to check the relationship between the 3 tasks in our test, while the qualitative study uses the retrospective interview to identify the reasons for these effects, i.e. the differences in scores in terms of the test takers' comments and the descriptions about each task characteristics, based on Bachman and Palmer's framework, and rechecks it by triangulation with questionnaires from teachers who also participate in this test.Results show that, on one hand, task is a definite factor that influences the performance of test takers. Right in this experiment, the differences between the three tasks (Questions and Answers, Discussions, and Oral Presentations) result in the variability of scores of test takers and they are the biggest factor that contributes to the score variability, which is a very important one in this study. On the other hand, these 3 tasks are not sex typed. There is a rough agreement between men and women on the score differences. As to the reasons for these differences, in this study, we could identify three points (easiness, popularity, familiarity of the task) and their different task characteristics in Bachman and Palmer's framework. However, we can not easily reach a decision that Questions and Answers is the best in view of test takers' scores, or Discussion is the best in views of test takers' comments in the retrospective interview. There are still other factors that determine the best or the most suitable task for the test takers and it is even very hard to find a balance between these points of tasks.At last, the results of this test can not be overgeneralized into writing, reading or other different test and instruction conditions, especially when the sample only includes 30 scores in the quantitative study and 10 test takers' comments in the qualitative study. The results are valid and valuable only in similar test conditions in oral tests or instructions, and hope to shed some insights into the continuous study on tasks.