<![CDATA[The relationship between the linguistic features that candidates produce during oral proficiency interviews and the band scores they are subsequently awarded remains a central, yet still under-specified, concern in language testing research. Drawing on Conversation Analysis (CA), this article examines the interactional organisation of the IELTS Speaking Test (IST) and re-examines how candidate talk at different score levels differs in its topical, sequential, and lexico-grammatical properties. Sixteen extracts drawn from publicly available IELTS test recordings were transcribed using standard CA conventions and analysed turn-by-turn for topical coherence, engagement with the topic, repair practice, identity construction, colloquial delivery, lexical choice, hesitation phenomena, and syntactic complexity. The analysis shows that higher-scoring candidates display topic elaboration beyond minimal response, deploy repair and clarification requests as resources for maintaining fluency, construct favourable interactional identities, and produce syntactically complex, low-dysfluency talk, whereas lower-scoring candidates orient minimally to topic development, exhibit extended hesitation and self-repair trouble, and produce comparatively simple morphosyntax. These patterns are discussed in relation to the official IELTS band descriptors and to the wider construct of interactional competence, and are triangulated against recent quantitative and discourse-analytic studies of the Speaking Test. The article argues that candidate talk is not merely a symptom of an underlying score but an interactionally accomplished performance that examiners and raters orient to, with implications for rater training, task design, and the validity of oral assessment more broadly. Methodological limitations associated with a small, opportunistically sampled corpus of publicly available recordings are acknowledged, and directions for larger-scale, ethically sourced CA-based validation research are proposed.]]>
