An Empirical Study Into Lexical Inference And Strategy Use In EFL Reading

The purpose of this study is to investigate, from the perspectives of foreign language proficiency, majors and gender, the differentiations and homologies of undergraduates in terms of Lexical Inference Ability (LIA) and Lexical Inference Strategy Use (LISU) on the theoretical basis of reading comprehension theory. It puts forward and addresses four research questions accordingly. 148 subjects are randomly sampled, according to two score ranges: high and low, from 7386 undergraduates who took part in the previous CET-4 in Sichuan Agricultural University to participate in this experiment. Of the 148 subjects, 20 participate in the pilot study of the examination paper so as to guarantee the test' s internal reliability and construct validity by way of reliability analysis and factor analysis; the other 128 attend a test on Lexical Inference Ability (LIA) and Lexical Inference Strategy Use (LISU), with only 112 valid samples. Descriptive statistics, independent sample t-test, non-parametric test, linear regression analysis, and structural equation model are employed to analyze the data.The results indicate that: (1) 65% of the subjects employ the intended (right) strategies to infer unknown word meanings successfully in the reading process, 35% of the subjects guess randomly. The frequencies of their using strategies successfully to solve problems are sequenced in the following order: Morphological Clues (Mor) 83%, Background Clues (BK) 77.7%, Paraphrasing Devices (Para) 73%, Logical Clues (Log) 63.4%, Monitoring Devices (Mon) 28.6%. (2) Students of high Language Proficiency (LP) perform, statistically speaking, significantly better than those of low LP in LIA, Strategy Score (SS), Inference Score (IS) and most of LISU except Mon and Log. As to Log and Mon, no significant difference can be found between students with high and low LP statistically. (3) Statistically speaking, no significant difference exists between the performances of students with different majors in terms of LIA, SS, IS, and most of LISU except Mon and Log. Students majoring in arts perform significantly better than those majoring in science and technology do in Mon, but the other way round in Log. (4) Statistically speaking, no significant difference exists between the performances of boy students and girl students with respect to LIA, IS, and most of LISU except Mon and Log. Boy students perform significantly better than girl students do in terms of SS and Log, but the other way round in Mon. (5) Linear regression analysis indicates that various LISUs are positively and significantly correlated with learners' LIA and IS. Structural equation model reveals that the LISU of students can effectively predict their LIA and SS.Therefore teachers, after realizing the existence of differential factors such as: gender, characteristics, context, LP, and majors, should take different measures to teach students in accordance with these factors in the process of teaching and learning practice, which can, on one hand, facilitate learners' strengths and avoid their shortcomings so as to develop learners' LISU as well as to, on the other hand, improve learners' LIA in the process of EFL reading in general.