摘要：This study was primarily aimed at developing an English-speaking proficiency test
and analytic rubrics designed to measure speaking proficiency of Malaysian
undergraduates. On the basis of Littlewood’s Methodological Framework and
Long’s Interaction Hypothesis, the researchers derived three speaking tasks from
four sources: (a) syllabus of the English language courses at the relevant university,
(b) Kathleen Bardovi-Harlig’s operationalizing conversation speech acts, (c)
IELTS part B speaking test, and (d) task B speaking section of Malaysian
University English Test (MUET). A total of 96 undergraduates with four levels of
the language proficiency (i.e., low performers, intermediate performers, upperintermediate
performers, and high performers) from a public university in Malaysia
voluntarily participated in the study. While two TESOL experts were invited to
validate the content of the tasks and the rubrics, two raters rated students’ test
scores. Construct validity was established through known-group validity (construct
validity) for a known-group comparison of the task performance at the three
difficulty levels namely, elementary, intermediate and advanced. The test scores,
having good internal consistency (a= .89) and inter-rater reliability (ICC= .84),
yielded speaking proficiency descriptors. This result showed that the test is reliable
and valid to diagnose speaking proficiency of Malaysian undergraduates in pursuit
of improvement.
关键词：speaking assessment; analytic rubrics; validation; reliability; test
development
