口语流利度，尤其是对话流利度，是口语能力的重要组成部分。Segalowitz（2010）提出的流利度三维框架包括认知流利度、话语流利度和感知流利度。其中认知流利度及其与话语流利度之间的关系仍属新兴研究领域。此外，大部分研究通过独白数据探讨口语流利度，而对话流利度方面的研究甚少。相应地，流利度研究多关注个体自身表现，忽略了个体间影响。基于此，本研究从个体内和个体间视角考察了对话中话语流利度与感知和认知流利度的关系。本研究采用了并行三角互证混合研究设计，量化数据来自理解性认知流利度测试及问卷调查，质性数据则来自产出性认知流利度测试的口头产出录音、中英文对话任务和刺激性回忆。60名中国英语学习者和40名美国英语母语者参加了本研究。研究发现一个人的话语流利度可预测自己和对话者的感知流利度得分,多个认知流利模块可预测话语流利度特征，主要研究发现如下：1）多个话轮内话语流利度特征（如停顿、修复、语速）和话轮间特征（如话轮间停顿）可以预测话轮内感知流利度；话轮间停顿、互动式倾听、副语言特征和话轮内停顿可以预测话轮间感知流利度，话轮间关联性仅对高水平说话者的流利度得分有预测作用；2）不同的认知流利度模块对话语流利度的预测作用差异较大；形式合成能力主要与停顿相关；监管与话轮内和话轮间的话语流利度均相关，而且语音监管比形态句法监管更可能预测话语流利度；听力理解与语速和话轮间停顿时长变化度有关；3）说话者的诸多话语流利度特征可预测对话者的话轮内和话轮间流利度得分（总效应）；对话者的停顿和重复频数对相应的效应有中介作用，且中介作用更可能在英语水平较高的说话者之间发生；说话者的话轮内停顿和语速可以通过对话者的多个话语流利度指标的中介作用预测对话者的感知流利度得分。研究结果有助于深化对于对话流利度的认识。从理论上讲，有助于厘清对话流利度中三个维度的概念定义和操作定义；从个体内和个体间视角拓展话语-感知流利度关系和认知-话语流利度关系研究的深度与广度。话轮内和话轮间指标的区分以及新的话语、认知和感知流利度指标有助于更全面地考察口语流利度的三个维度，为对话流利度提供分析框架。在教学上，确认有效的话语和认知流利度指标及对话者之间的影响，可以为流利度测评、二语口语教学设计与口语练习提供实证支撑和方向。

Speech fluency, particularly dialogue fluency, constitutes an essential component of oral proficiency. Segalowitz (2010) proposed a triadic framework of speech fluency, encompassing cognitive, utterance, and perceived fluency. Cognitive fluency, especially L2-specific cognitive fluency, and its relationship to utterance fluency remain emerging research areas. Moreover, the predominant focus on monologue data in fluency research has led to a dearth of studies investigating dialogue fluency. Additionally, research related to fluency often considers it as an aspect of individual performance. Few studies have adopted a between-speaker perspective of fluency in dialogues to examine interlocutor influence. Correspondingly, the present study explores the relationship of utterance fluency to perceived and cognitive fluency in dialogues, from both within-speaker and between-speaker perspectives.This study employed a concurrent triangulation mixed-methods design. It collected quantitative data through comprehension-oriented cognitive fluency tests as well as a questionnaire, and qualitative data through production-oriented cognitive fluency tests, dialogue tasks, and stimulated recall. Sixty Chinese EFL speakers and 40 American native English speakers participated in this study. Quantitative and qualitative analyses revealed that one’s utterance fluency can predict not only one’s own but also one’s interlocutor’s perceived fluency scores, and most cognitive fluency modules can predict various utterance fluency features. The major findings are as follows: 1) various within-turn (e.g., pause frequency and duration, repair frequency, and pruned articulation rate) and between-turn (e.g., turn pause length) utterance fluency measures significantly predicted within-turn perceived fluency; utterance fluency measures significantly predicting between-turn perceived fluency included turn pause length, interactive listening frequency, paralinguistic noise duration, and within-turn pause frequency and duration; extra utterance fluency features (e.g., between-turn contingency) is necessary in grading higher-proficiency speakers’ performance; 2) different cognitive fluency modules showed various prediction effects on utterance fluency; formulation was primarily related to pause-related utterance fluency measures; monitoring was associated with both within- and between-turn utterance fluency, with phonetic monitoring more likely to predict utterance fluency than morphosyntactic monitoring; comprehension was related to pruned articulation rate and turn pause variability; 3) total effects on the interlocutor’s within- and between-turn perceived fluency scores were found in a myriad of one’s utterance fluency measures; the interlocutor’s pause frequency, within-clause pause frequency, and repetition frequency mediated the effects, and the mediation effects were more likely to occur among speakers of higher proficiency; the effects of one’s various pause frequency measures and pruned articulation rate on the interlocutor’s perceived fluency score were mediated by multiple mediators (the interlocutor’s utterance fluency features).The findings provide some theoretical, methodological, and pedagogical contributions and implications. Theoretically, this study helps to clarify the conceptualization and operationalization of dialogue fluency in the three dimensions (cognitive, utterance, and perceived fluency) as well as broaden and deepen the research into utterance–perceived and cognitive–utterance fluency links from both within-speaker and between-speaker perspectives. Moreover, this study pioneered the distinction between within- and between-turn fluency and incorporated novel measures of utterance, cognitive, and perceived fluency to comprehensively explore the three dimensions of speech fluency. Pedagogically, the identification of effective utterance and cognitive fluency measures and interlocutor influence provides research-based evidence for fluency assessment and L2 speaking instruction and learning.