This study explores the relationship between the prosodic features for time delay, self-monitoring in speech production, and perceived dysfluency. In this study, twenty native and non-native speakers of Chinese took a speech test. Each speech was transcribed, prosodic features were assigned symbols, and the coding system traced self-monitoring. An additional twenty-eight native speakers assessed the fluency of the speech samples, and then the researcher matched assessment results with symbols and coding, and analyzed them. The results indicate that uh/um and self-monitoring influence perceived dysfluency in most cases while other prosodic features do not; that the filled pause in non-native speech is a salient feature of perceived dysfluency; and how a dysfluency is perceived. The study also finds the native speakers’ perception bias.