This research study employed a mixed-methods approach to investigate the test takers’ perceptions and anx- iety in relation to an English language proficiency test called Community English Program (CEP). This study also evaluated the direct and semi-direct modes for speaking module of this test. To this end, 300 Eng- lish as Foreign Language (EFL) students were recruited in the study as test takers. They were invited to take the CEP speaking test using five tasks of Description, Narration, Summarization, Role-play and Exposition in both direct and semi-direct test modes. Their perceptions and evaluations of both test modes, through questionnaires, interviews and observations were examined. The results of the factor analysis revealed that test takers’ evaluations of both direct and semi-direct speaking modes were quite similar, yet not exactly identical. On the other hand, although test takers’ anxiety was shown influential, the findings showed that the most determining factor in test takers’ oral performance was their capability level. Capability level was the main reason why some test takers out-performed the others. The findings also demonstrated that test difficulty identification was complex, difficult and at the same time multidimensional. The quantitative re- sults displayed that the raters were scoring speaking performances differently; the qualitative results also provided logic for the reasons of these differences on the side of the test takers. Finally, the impact of test takers’ gender differences on their perceptions was found nonsignificant. Manuscript profile

This study investigated students’ oral performance ability accounting for various oral analytical factors including fluency, lexical and structural complexity and accuracy with each subcategory. Accordingly, 20 raters scored the oral performances produced by 200 students and a quantitative design using a MANOVA test was used to investigate students’ score differences of various levels of language proficiency groups with respect to their oral scores in each analytical factor. The findings showed that students, in each level of language proficiency, were different from each other regarding various measures of fluency, lexical complexity, structural complexity and accuracy when performing the five oral tasks. Besides, the findings showed that language planning, perspective and immediacy were the determining dimensions in oral task difficulty. The findings demonstrated the usefulness of analytical approaches to rater training programs in detecting rater effects and demonstrating the consistency and variability in rater behavior. The analysis confirmed that the nature of second language oral construct is not constant, thus different results are achieved using different oral task dimensions. Consequently, the outcomes have constructive implications in the use of feedback as a reliable indicator of task difficulty and specifically as a basis for test design and validation. Manuscript profile