TESTING EFL READING COMPREHENSION IN ENGINEERING ENGLISH

This dissertation investigates the descriptive characteristics, reliability and validity of a norm-referenced test of engineering reading ability. The sixty-item (three-passage) test was built from the "best" items in three longer tests, previously developed at UCLA in a cooperative project between MA students in TESL and professors of engineering. The shorter composite version was administered to four groups of people: (1) American engineering graduate students, (2) American ESL graduate students, (3) Chinese engineering graduate students and (4) Chinese EFL graduate students.; The descriptive statistics were calculated for all four groups. Then, classical test theory reliability (K-R20 and Split-half methods) was calculated for each group individually, as well as for each nationality. Reliability was also considered from the generalizability theory (G theory) point of view. Using analysis of variance (ANOVA) techniques, G theory enabled the investigator to isolate variance due to differences among items and subtests as sources of testing error. In addition, G theory allowed prediction of reliability coefficients at different numbers of items and subtests. Construct validity was also experimentally demonstrated. Based on the data from the four groups above, ANOVA procedures were used to show that the test discriminates significantly between native and non-native speakers of English, as well as between engineers and non-engineers. The results indicate that the test was a reliable and valid test of engineering reading ability for the samples involved in the study. In addition, examination of the various item types used shows that all types of items are equally appropriate for this type of testing. Finally, correlational analysis of engineering reading ability and overall language proficiency indicates that as much as 62 percent of engineering reading ability may be accounted for by overall language proficiency.; These results are discussed in terms of using such a test for decision making in EST and in terms of the factors which may contribute to engineering (if not overall EST) reading ability.