This case study focuses on investigating and describing the speech disfluencies in three English dialogues with three proficient non-native speakers. The aim of the study is to qualitatively examine the distribution of speech disfluencies in the analyzed dialogues and discuss the findings relative to previous research on speech fluency and disfluency. The study participants are independent users of English, who share a similar proficiency level. The analysis is based on four measurable variables that, according to previous research, are related to perceived fluency and disfluency in dialogue (1. *rate of speech* – the number of words per minute of speech, 2. *breakdown fluency* – the number and length of pauses, 3. *repair fluency* – the number of false starts, corrections, and repetitions, 4. *correlation of speech disfluencies with the flow of the dialogue* – the number of disfluencies per one illocutionary act type). Most disfluencies in the data occur at the beginning of a dialogue act. Major factors that may cause disfluency include difficulties in structuring the phrase, remembering a rarely used word, describing complex entities, and hesitating. Confirming the previous findings, the analysis showed that more cognitively demanding tasks lead to higher numbers of disruptions. However, some results of the data analysis appear to contradict the previous experiments. As such, the paper found out that the fastest speaker produced the highest number of repair disfluencies, which intersects with the previously established correlation between language proficiency and rate of speech.