Reliability And Validity Study On Task-based Communicative Spoken English Test(TC-SET)

Throughout the history of language testing,the reform of language testing is closely related to the development of linguistics.When structuralism was challenged by sociolinguistics,language testing was transformed from psychometrics-structural testing to psycholinguistic-sociolinguistic testing.In late 1970s,after Bachman put forward the communicative competence testing theory,communicative language testing stands for the new trend of testing development.At the beginning of 1980s,based on communicative language testing,task-based language testing is becoming a hot topic. The communicative language testing is concerned primarily with how language is used in communication,while the task-based language testing is concerned with testees' language ability by accomplishing the task.Task-based communicative language testing combines the advantage of communicative and task-based language testing,aiming at the accomplishment of tasks with a focus on communicative competence.It tests testees' communicative competence by examining the process of task accomplishing. Meanwhile,with the development of society,the skill of fluent spoken English has been vital in social and international exchanges.Therefore,in order to meet the current requirements of the language teaching and social development,this paper focuses on the study of task-based communicative spoken English test(TC-SET).Reliability and validity is the primary concern in test interpretation and use with its significance running through the entire process of test construction and evaluation, while the empirical study on reliability and validity of TC-SET is deficient both abroad and at home.Based on the theory of task-based communicative language testing,81 sophomores of non-English major in Jiangsu University are required to evaluate reliability and validity of TC-SET.The reliability study focuses on three aspects,that is, the factors that affect TC-SET,the external and internal reliability of TC-SET.The validity study focuses on the analysis of construct validity,content validity,face validity and predictive validity.Quantitative and qualitative fonns of data are gathered and SPSS 13.0 is employed to do statistical analysis.Correlation analysis and paired-sampled T test are taken to estimate external reliability;the analysis of Cronbach alpha Coefficient is employed to estimate internal reliability.Correlation study is conducted to test construct validity;qualitative analysis is used to evaluate the adequacy with which the test content represents the content domain;students' and raters' responses to the questionnaire are analyzed to see the face validity;the linear regression is employed to evaluate predictive validity.Results show that TC-SET has high reliability and validity.In general,TC-SET reliability is satisfactory.Most of students are satisfied with the length and difficulty of TC-SET.Only a few students consider that TC-SET should add new tasks and extend test time so as to reflect their oral communicative competence better.The reliability of inter-rater correlation coefficient is 0.891 which is significant at the 0.01 level(1-tailed),and there is no significance between two scorers' rating in the paired-sampled T test.The internal reliability of TC-SET is 0.8434(January) and 0.8718(June),which shows that TC-SET has high internal reliability.The construct validity of TC-SET is high.In the correlation analysis,students' scores on each task correlated significantly,indicating that the tasks of TC-SET measure the same construct only each of them focuses on a different aspect of the construct.Also,each task correlates more highly with the whole test.All the evidence shows that TC-SET has its construct validity,i.e.it tests an independent construct.As to the content validity,researches are conducted on 10 TC-SET papers,and the analysis shows that TC-SET tests what the syllabus claims to test.With regard to the face validity,responses to candidates' and raters' questionnaires show that most of the candidates and raters find TC-SET satisfactory.The predictive validity of TC-SET is also high.In the linear regression,the January score can predict 69.4%of the June score.In this study,a small-scaled university-based TC-SET is proposed,which is not only helpful to the reform of college English oral test,but also fills up the deficiency of National College English Test- Spoken English Test.