| Validity studies on language proficiency tests have attracted many researchers in the last decades. Most of the validation work try to investigate skills assessed by the items in the tests, hence the construct validity of the test. The investigation of construct validity on language proficiency has established well-recognized ways quantitatively and qualitatively. The present study takes factor analysis approach and questionnaire method to investigate the construct validity of the reading part of the Public English Test System Level 3(PETS3). PETS is now as a powerful alternative test for language level proof, and the domestic construct studies on PETS have focused on listening, speaking, reading, writing as general language competences. However, the construct studies on the reading part of this test seem quite few. The empirical investigation into reading construct may provide statistical proof of the reading construct validity, much insight of the reading process and feedback of the reading test for the considerations of the test designers.Consequently the research addressed the following questions: 1.Do the majority of the experts agree with the skills the test items intending to test, according to Weir’s taxonomy of reading skills? 2. Does exploratory factor analysis support that the PET reading parts(Part A and Part B) assess the same reading skills claimed by its test specifications? The conclusion was that in Part A only one factor was extracted; the reading skills the experts agreed upon the test items intending to test were included in Factor one’s item cluster, and 60% shared a common “latent variable”. It was reasonable to believe the skills the reading Part A testing were in accordance with PETS’ test specification. The second research question didn’t come to the expected results quantitatively. However, considering the test’s reliability and combined with Grabe’s reading process theory, Part B’s reading construct validity was relatively average. |
PDF Full Text Request