1. Introduction With the development of economic and political globalization, English has become the language of the world. More and more non-English speakers study English as their foreign language. Oral proficiency is now one of the most important principles to identify the English levels in EFL learners. As for the oral fluency, pause is a significant index for measurement in the existing domestic and foreign researches. What#8217;s more, the features of pauses in EFL learners with different oral proficiency are also different. 1.1Need for the study Pause is a basic feature of natural language output. For native speakers, it is necessary for the stop of natural verbal communication. Pause has many functions, such as ease of online speech load control or direct turn in discourse, and discourse markers of sentences and sentence boundaries (Yang, 2004). However, as for second language learners, pause or L2 unfluency performance, often means that the second language output encounters difficulties, so pause is also used as a standard to judge whether an output is fluent in second languages (Riggenbach, 1991). The type of pause includes filled pause and unfilled pause. Because many studies point to use cut-off point to define the pause, this study will define the words of a filled pause is equal to the cut-off point (filled pause) and intermittent (unfilled pause) (Cruttenden, 2002; Lennon, 1990; Towell et al., 1996). Wenzhong Zhang and Xudong Wu#8217;s research proved that the time threshold of pause was set at 0.3 second and above, was more appropriate for the English output fluency of Chinese students. Therefore, this study defined the pause in students' monologue as 0.3 second and above filled and unfilled pause. In this paper, English majors with different English oral proficiency are selected as the subjects to investigate the features of Chinese EFL learners#8217; non fluent pauses in L2 output. 1.2 Research purpose This paper aims to study the features of the different English levels of second language learners in non fluent pauses, and comparison of their non fluency pause difference in frequency, duration and distribution, including the relationship with oral proficiency and different pauses. 2. Literature review In this chapter, literature is reviewed in two parts, first on the theoretical foundation，second on related empirical studies on pause. 2.1 Theoretical foundations 2.1.1 Levelt#8217;s speech production model Levelt#8217;s speaking model, an oral production model, theoretically explains the whole process of discourse production from intake to output. The function of the conceptual mechanism is selecting the expression appropriate to the declarative knowledge concerned. And the formulator mechanism, by a certain encoding process, organizes the message into a pre-verbal form. The pre-verbal message with content of the assigned topic then passes through the formulator mechanism, whose function is to convert pre-verbal messages into phonetic form by correctly selecting appropriate lemma, lexeme, and using grammar and phoneme lemma and morphology stored in a language corpus. Once lemmas are selected, some relevant syntactic patterns are immediately picked out. Grammar encoding forms the surface structure of the discourse. Phoneme encoding occurs simultaneously with the formulation of this surface structure of discourse. The formulator mechanism converts the surface structure of discourse and phoneme encoding into phonetic form and then sends it into sound-production mechanism, and then the language production can be heard by listeners. In fact, as language comes out very fast, the whole process can occur almost instantaneously. Central to this framework is the concept that the development of oral ability consists of three stages (Levelt, W. 1989). In the first stage, knowledge about the target language is gained. In the second stage, ”declarative knowledge” is applied to oral use; the result of oral practice is ”procedural knowledge”, the development of which will gradually allow the learners faster, more unattended, automatic and simultaneous access to their ”declarative knowledge” about the target language during oral communication. In the third stage, the ”procedural knowledge” undergoes a ”restructuring”. According to Levelt, the more practical a speaker#8217;s declarative knowledge is, the more fluently his or her procedural knowledge is capable of producing language. 2.2 Related empirical studies on pause In China, Wenzhong Zhang, Xudong Wu, etc make the earliest research to make a definition and classification on pause according to foreign studies. On this basis, more professors make a deeper related research on pause such as a quantitative study; the pause under different tasks, researches combined corpus, etc. Wenzhong Zhang(2002) has also done some research on the reasons for the pause. From a standstill, no longer change the language stream, marks and words such four factors, they make an analysis and comparison on the past 12 subjects before and after the recording of the two texts and they found that two types of pauses serve the purpose of conception: what to say and how. And the pause used to conceive the content of the utterance (what to say) seems to be longer than the pause used to conceive the language expression (how to say). Mu (2005) hopes that some of the obvious features of Chinese students#8217; extraction and generation of vocabulary in the process of language generation will be linked to the students' overall language expression ability to find some of the reasons that affect the fluency and naturalness in oral production. The corpus used is the dialogue of the third part of the 2004 TEM 4. They extracted 13 of them and 26 samples as the object of specific analysis according to the recording quality and other factors. They mark on the supersonic features, lexical phenomenon, speech errors appeared in the dialogue and mainly focus on four aspects: difficult pause, pause strategy, word pronunciation, rhythm. Then they use the Software Speech Filing System (SFS) 4. 6 Edition 2 to make records of the material they collected to make a research. Some significant results are obtained: there is an inverse relationship between the frequency and pause, but there are exceptions, that in addition to pause frequency, there are other factors that affect the overall effect of spoken words; extraction and generation is indeed most difficult reason appears in most pauses; pause before in various forms, and these forms is an important part of the features of L2 language learners. By using the PACCEL-S corpus of Chinese English interpretation, Zhaohui Dai (2011) makes a research of the disfluency in Chinese to English interpretations of Chinese EFL learners. First, they study the features of non fluency phenomenon in interpretation of Chinese students with all kinds of labeled statistics by using the search pattern software. Then they extract the and information from the file information, and divide into low and high packet according to student achievement. Then they import them into SPSS 17, finally make description statistics and sample test research based on the non fluency phenomenon. The non fluency phenomenon in interpretation of learners in pauses (F_P) accounted for 43% and (U_P) for silent pause 28%. The most commonly used word pauses is ER (1083 times), eh (195), um (161), en (153), EM (130) or. Under the influence of native language, the use of words such as ”I mean#8217;, ”well#8217; was not widely used. Students in interpreting mostly repeated words is the definite article the, this shows that students from the memory accurately interpreting noun phrases are stressed, likewise, it is not easy to correctly interpret a qualified clause; it also explains why the students mostly repeated verbs are ”is#8217; and ”are#8217;. In Chinese students#8217; interpretation of non fluency phenomenon, pause and repeat account for the majority of the proportion. Vocative pause and repetition have the same pragmatic function, which is the hesitator in interpreting, so that students can get the time to pause in discourse output. References Cenoz, J. (1998). Pauses and communication strategies in second language speech. Reports-Research. Fillmore, C. S. (1979). Individual differences in language ability and a language behavior. New York: Academic Press. Krivokapic, J. (2007). Prosodic planning: Effects of phrasal length and complexity on pause duration. Journal of Phonetics, 35, 162-179. Liu, Q. [刘芹], 2008, 中国大学生英语口语水平研究. 现代外语（1）：84-89。

Liang, S. W. [梁砾文], 2010, 英语专业学生口语独白中不自然停顿的界定方法. 海外英语 (3)：66-68。

Levelt , W. J. (1989). Speaking: From intention to articulation. Cambridge, Mass: MIT Press. Mu, F. Y. [穆凤英], 2005, 中国学生英语口语实词提取与生成特征分析. 外语教学与研究(4)：250-258。

Miao, H. Y. [缪海燕], 2009, 第二语言口语非流利产出的停顿研究. 解放军外语学院学报 (4)：56-60。

Miao, H. Y., Sun, L. [缪海燕 孙蓝], 2006, 句子构造框架的使用与二语口语流利性发展研究. 外语教学与研究(5)：265- 271。

Sajavaara, K. (1987). Second language speech production: Factors affecting fluency. In Psycholinguistic Models of Production. (eds.) New Jersey: Ablex. Wood, D. (2006). Uses and functions of formulaic sequences in second language speech: An exploration of the foundations of fluency. Canadian Modern Language Review, 4, 573-589. Yang, J. [杨军], 2004, 口语非流利产出研究述评.外语教学与研究( 4)：278-284。

Yin, N. [尹楠], 2011, 不同任务下二语口语产出中的停顿现象研究. 牡丹江师范学院学报 (5)：67-69。

Yan, H. Y. (2005). Students#8217; self-willing motivation impacts on second language oral fluency. CELEA Journal, 28, 16-23. Zhang, W. Z. [张文忠], 2000, 第二语言口语流利性发展的定性研究. 现代外语(3)：273-283。

Zhao, Y. K. (2011). Rethinking pause and fluency in L2 oral performance. Teaching Research, 7, 108-114.