What types of linguistic information do people commonly use to determine sentence meaning? Within linguistic theory, the traditional view is that the verb is the main determinant of the syntax and semantics of sentences and it has been supported by innumerable empirical evidence particularly from the perspective of sentence comprehension studies. However, two latest experiments are reported in the Journal of Memory and Language by Guilia M. Bencini and Adele E. Goldberg (2000), suggesting that argument structure constructions are better predictors of overall sentence meaning than the morphological form of verbs.In this thesis we have investigated the role of verbs or argument structure constructions in determining sentence meaning, to test the assertion made by Bencini and Goldberg (2000). We have conducted two sorting experiments similar to Bencini and Goldberg (2000) in a context of Chinese EFL learners at three language proficiency levels. Participants were asked to sort sixteen sentences, which were created by crossing four verbs (got, took, threw, and slice) with four different constructions (transitive, ditransitive, resultative and caused motion construction), into four groups according to their overall meaning. Subjects could sort equally well by verb (i.e. verb-based sorts), or by construction (i.e. constructional sorts). The results suggested that in contrast to a uniform conclusion made by Bencini and Goldberg (2000) that sorts were closer to a constructional sort than to a verb-based sort on average, the sorting in the present study differs with language proficiency levels. Specifically, our conclusion is that EFL beginners are more influenced by shared morphological form of verbs, while intermediate and advanced learners are more influenced by shared constructions in sentence interpretation.This thesis is essentially an extended argument in favor of Bencini and Goldberg (2000), but it takes a further step claiming that in an L2 context, the language proficiency level makes a difference on the determinants that participants rely on in constructing sentence meaning. Meanwhile, it somehow contributes to the theory development of Construciton Grammar by testing empirically in an L2 context.<WP=7>AcknowledgementsI am especially grateful to my supervisor, Dr. Dong Yanping, a respectable scholar, a smart and graceful lady, who brought me into the world of cognition and led me approach the wildness of research. Her enduring insights, meticulous and logical thinking have profoundly influenced this work in innumerable ways. Without her greatest patience, incredible generosity in sharing her time and her deep insights, this thesis could never have reached its present form.I’d like to thank Professor Gui Shichun, Professor Qian Guanlian, Professor He Ziran, Professor Wang Chuming, Professor Wu Xudong, Professor Xiao Huiyun, Professor Chen Jianping, Dr. Zeng Yongqiang, Dr. Wen Binli and Dr. Ran Yongping for instilling in me a deep respect for the complexities of academic research, and for sharing their wisdom.I owe much gratitude to Charles J. Fillmore, Paul Kay, University of California, Berkeley, Adele E. Goldberg, University of Illinois, Mirjam Fried, University of Princeton, for their great generosity in providing me with their own articles and books, and sharing me with their wisdom. I am particularly grateful to Goldberg, for her consistent willingness to lend an ear and a critical eye, and to Fried for his brilliant intelligence in making the informative website of Construction Grammar and his consistent help and encouragements in updating my knowledge in related area.Special thanks go to Guilia Bencini, University of California, Berkeley for having me share her wisdom in data analysis, a highly demanding task on mathematics and computer science.I am also grateful to Willa Xiao-jing Wang, my former roommate now in the University of California, San Diego. During the middle and final stages of my writing this thesis, I was fortunate enough to get all those materials that