| This thesis reported a study of idiosyncratic features of disfluency (abbreviated as "DF") in normal EFL speech in the interest of enriching the understanding of speech production processing.DFs are thought to have considerable potential to shed light on the nature of the speech production processing system. They carry information on where in the processes the breakdown has occurred, and how, in the sense of possible mechanism involved. Previous studies have assumed that the evidence drawn from DF investigation can reflect what a particular system is possibly like and how, under what conditions, the system normally works. Based on the assumption, the study addressed the following questions: (a) What kinds of DFs do EFL speakers produce? (b) How often do EFL speakers produce DFs? (c) What in speech production system causes EFL speakers to produce DFs?The main part of the study began by selecting a naturalistic approach. Speech mode and topic domain were designed in accordance with the research questions. 10 graduate students and 10 undergraduate students were randomly selected from colleges to contribute spontaneous conversational speech in the setting of relaxed situations. A background information survey was administered to every subject after each conversation and one sample chi-square test was performed to verify the data validity. Transcription criteria specialized for the research were established in the light of transcription conventions. The surreptitiously recorded speech was then transcribed into a written corpus in XLS file, generating 23 117 words and 2 307 DFs. Annotation system and algorithm of DF taxonomy were developed to treat data in preparation for statistical analysis. The naturally displayed disfluent phenomena were labeled as hesitation, repetition, substitution, deletion, insertion, and same position multiple DF and multiple position multiple DF. Spearman rank correlation coefficient test and paired-samples Mest were performed respectively to analyze statistical results. The results of the investigation into DFs of EFL speech production were discussed from three angles: DF frequencies, DF distribution and DF types. It has been shown in the research that more frequently used words receive adequate activation for execution more quickly than less frequently used words do and that the English lemmas that do not have syntactical, semantic and prosodic correspondences in Chinese language receive less activation for execution than those that have syntactical, semantic and prosodic correspondences in native languageIn conclusion, the density of competing nodes at each level of speech production system and the automaticity of processing determine what an EFL speech production system is like and how it works.Based on the analyses of EFL spontaneous speech data, the study has suggested the possible causes of DFs in EFL speech production system. Implications for the speech production theory and spoken English teaching have been provided in the study and will contribute to further psycholinguistic research in the field. |