In formal assessments of second-language speakers' fluency, the criteria used frequently fail to differentiate between real disfluency and interactionally effective practices, while superficially disfluent, that help with communication and are used by first language speakers. In this study, we focused on pausing, silences and self-repairs in the speech samples in two data sets: the classroom talk of two L1 teachers of English, and mock IELTS interviews of i-Kiribati students. We found that these features and their frequency and distribution were often very similar in the two data sets suggesting that a superficial observance of disfluency markers in the case of L2 speakers hides practices that may in fact display a high level of competence. We highlight the caution that needs to be exercised when making fluency judgements.