中文摘要：

本文分四部分：第一部分从事实与理论两方面说明构式理论有用但不能包打天下。第二部分从外在语言和内在语言两方面说明构式的产生。从外在语言方面来看，语言的变异是常态，语言变异被频繁使用并泛化，就会逐渐固化为新的语法构式。从内在语言方面来看，“构式源于认知”，这可以从两方面来加以假设说明——“由内到外运作”的假设和“由外到内运作”的假设。第三部分以实例说明构式理论需与语块理论相结合才能发挥更好的作用。第四部分提出一个可以进一步探究的问题：构式与语言信息结构该是什么关系？

英文摘要：

The paper consists of four parts. The first part argues from the perspectives of fact and theory that though useful, construction theory does not explain everything. The second part tries to explain the generation of constructions from the perspectives of external language and internal language. From the perspective of external language, language variation is a common phenomenon. Linguistic variations generalize after frequent use and become new grammatical constructions. From the perspective of internal language, construction originates from cognition, which can be ap- proached with two hypotheses： ＂internal to external mechanism＂ and ＂external to internal mechanism＂. The third part uses examples to argue that in order to play a better role, construction theory needs to be combined with chunk theo- ry. The fourth part proposes a question for further study： What kind of relationship should construction and finguistic information structure have？