Segalowitz (2012) has identified three types of speaking fluency: cognitive, utterance, and perceived fluency. Of particular interest is utterance fluency – the features of utterances that can be acoustically measured. Utterance fluency can be further divided into three facets: breakdown fluency, speed fluency, and repair fluency (Skehan, 2003). We carried out a pre-post analysis using app-based shadowing practice as a treatment to see if participants could improve their speaking performance. Based on our analysis of speech samples using the PRAAT software, we have characterized the nature of the interaction between speed fluency and breakdown fluency within intermediate English learners in Japan; and show how app-based shadowing practice can aid in improving speaking articulation and auditory perception. Learners with a high breakdown fluency tend to have a high speech density. This combination of speech characteristics, namely, high speaking density and large number of pauses, creates the perception of a highly disfluent speaker and can cause raters on standardized speaking tests to give disproportionately low scores. We will elaborate on the relationships between these speaking fluency metrics and their implications, and briefly describe the use of PRAAT in the analysis.