The development of the English Speaking Test: An investigation of reliability and validity

This study focuses on the development of a simulated oral proficiency interview (SOPI) of English language proficiency and a subsequent investigation of its reliability and validity. In this study, the development of an English SOPI included investigations of test content, interrater reliability and validity. Test development included a survey of 119 students of English as a second language studying in the United States from which test tasks were developed.; The test was administered to 30 students of English as a second language in the United States. Two raters rated each test according to the ACTFL Guidelines, which yielded a Pearson correlation of 0.90.; The study centered on this issue: do the ACTFL Guidelines elicit the kind of language they claim to elicit by speakers at three proficiency levels (Intermediate, Advanced and Superior) on tasks associated with those proficiency levels? To measure this, the researcher identified twelve linguistic features, three for each level of proficiency. The ACTFL model predicted that features at the ACTFL Intermediate level would show no significant differences in speakers, features at the Advanced level would show differences between speakers at the Intermediate and Superior levels and Intermediate and Advanced levels of proficiency, while features at the Superior level would show differences between the speakers at the Superior level and the Advanced level and between the Superior and Intermediate levels.; An ANOVA conducted on the mean frequencies of linguistic features associated with each of the three proficiency levels on groups of subjects rated at the three proficiency levels showed mixed results. The study also showed that the number of words produced by speakers was the best predictor of differences between proficiency levels.; This study has produced a test of English language proficiency consisting of tasks based on the needs of the target population. The study shows mixed support for some distinctions among the three proficiency levels; but more research is necessary to fully validate or discredit the Guidelines.