| The study reported in this paper is an investigation of descriptors of spoken English in the questionnaire survey (Appendix 1) which is an essential part of the project of Social Science Foundation of China,"On Establishing a Unified Level Framework of Language Proficiency in China"(No. 06BYY027) led by Professor Yang Huizhong. Aiming at providing evidence and reference for the project, the present study investigates the descriptors from both macroscopic and microscopic perspectives.From the macroscopic perspective, the study explores the underlying structure of oral proficiency that the descriptors describe by using a statistic method, exploratory factor analysis, and then compares and contrasts the structure with the theoretically three-dimensional structure of the descriptors, which is based on the rating scale of CET-SET.From the micro aspect, we make attempts to confirm whether a single descriptor accurately reflects the feature described in the descriptor by correlating the descriptors'questionnaire ratings with the quantitatively measured values of the features in given descriptors. The procedure mainly consists of five steps: Step 1: examine the features in descriptors and identify features that can be quantitatively measured; Step 2: transcribe the oral data and segment oral data; Step 3: review operational definitions and measures of the identified features in previous studies and put forward our own definitions and measures which might be a modified version of those we review; Step 4: calculate the values of the index of each feature; Step 5: correlate the values of quantitative measurement of each feature with ratings on corresponding descriptor or descriptors. If the correlation coefficient is relatively high, then we could consider that the descriptor reflects the feature in it in a relatively accurate way; if the correlation coefficient is relatively low, then the descriptor is somewhat problematic.The major findings of the present study are as follows: Through exploratory factor analysis, we have found that the underlying structure is also three-dimensional, since three factors have been abstracted. Factor 1 contributes much more to the explanation of variances in test-takers'oral proficiency than Factor 2 and Factor 3. Factor 1 seems to describe both"Coherence and Cohesion"and"Flexibility, Appropriacy and Effectiveness", but"Flexibility, Appropriacy and Effectiveness"assumes a dominating position, since descriptors in"Flexibility, Appropriacy and Effectiveness"generally have higher loadings than those in"Coherence and Cohesion"on this factor; descriptors in Factor 2 come from three different parts of the questionnaire, all describing average or low level of oral proficiency; descriptors in Factor 3 are describing"Accuracy and range", which is in accordance with our expectation. The factor structure is then compared and contrasted with the theoretically designed structure of the descriptors. Two similarities and three differences have been summarized.Through correlation analysis, we have the following findings: First, some descriptors'ratings have very low correlations with quantitatively measured values of features while some have high correlations with them. Two descriptors of which ratings have very low correlations with values of the index of the features in given descriptors are D302 and D101. Second, high correlations occur between values of the index of Lexical Diversity (LEDIV), Speech Rate (SPR), Volume of Valid Production (VOVP), and Fluency (FLU) and the ratings of related descriptors. For each finding, a couple of interpretations have been provided.In addition, we have also discovered the"higher demands, higher correlations"phenomenon and there exists certain relationship between factor analysis and correlation analysis.
|
PDF Full Text Request