A Retrospective Study On Test-takers' Processes Of Taking Compound Dictation Test

Compound dictation is one variant of dictation, the purpose as to what compound dictation intends to measure is far from clear, and the description of it is fairly general and fuzzy in CET testing syllabi. Previous researches differ substantially on the view of what compound dictation actually measures. It is necessary to look into the test-taking process if one wants to understand what constructs the test actually measures, thus this study employs retrospective approach to investigate test-takers'thought processes of taking the compound dictation.The research questions are put forward as followings: 1) What is compound dictation of CET actually measuring? 2) Do different response formats of compound dictation exert any influence on test-takers'performance? If yes, what are they? To answer the questions, immediate retrospection and retrospective interview are employed and thirty-six non-English majors at the tertiary level are involved in the study, the subjects are divided into four groups according to different response formats.Based on the previous studies on dictation test validity, with the help of Bachman and Palmer's (1996) description and division of language ability, and by examining the first group test-takers'protocols about their test-taking process, a framework for analyzing test-takers'processes of taking the compound dictation test is developed. The test-takers'protocols are further categorized and analyzed according to this framework. The findings of the present study are mainly as follows:First, examination of test-takers'verbal reports finds that none of them conducted the compound dictation test in accordance with the requirements of the directions, which may pose threat to the rationality of directions. It also reveals there may be some problem with the scoring criteria of compound dictation test; the language abilities tested in the compound dictation can be divided into linguistic knowledge and strategy use. The linguistic knowledge includes phonology, vocabulary, spelling, written expression and semantics, test-takers attach great importance to the speech sound of words they hear while completing the compound dictation test. Next, the employment of various cognitive and metacognitive strategies in the test-taking processes is investigated. Test-takers apply metacognitive strategies more frequently and there also exist some differences of test-takers'application of strategies across proficiency levels; there are also some ability-irrelevant factors affecting test-takers'processes of taking compound dictation. From their report, the requirement of memory capacity has important influence on the task performance, internal factors that more relating to test-takers'characteristics are attention and psychological state, external factors that more relating to the task itself are speech rate and time factor.Second, from the test results, it can be found that different response formats of compound dictation don't have much influence on four group testees'performance on word dictation, but Group Three (answering in Chinese) and Group Four's (without a context) scores on sentence dictation are significantly lower than Group One. There is not much difference between testees of different proficiency levels'performance on word dictation, but proficient and less proficient testees of Group One and Group Three have significant differences on sentence dictation, indicating the sentence dictation that they took may be more effective considering the item difficulty and discrimination, whereas the validity of word dictation is questionable; from four group testees'verbal protocols about their test-taking processes, testees of Group Three consider more semantic meaning and main idea of what they heard, less obstructed by the spelling and written expression ability of L2, apply both the cognitive and metacognitive strategies more frequently than other groups, and are less affected by ability-irrelevant factors when there is a context, which means they are more efficient during the test-taking process. Whereas testees of Group Four seldom use semantic knowledge and apply the strategies least compared with other groups when completing the test. Which illustrate that the compound dictation test without a context lacks the difficulty and discrimination for testees, and the response format of testees answering in their native language appears to be more effective in testing listening comprehension ability than other response formats.Based on the above findings, this paper attempts to put forward some suggestions on CET compound dictation design and college English teaching and learning, at the same time provide the theoretical and methodological references for adopting retrospection approach in the area of other test type validity studies.