Measuring coherence is a standard practice in oral proficiency (OP) tests, but poses problems for inter-rater reliability and test validity. The model of cohesion developed in SFL (Halliday & Hasan, 1976) has been advanced in a way that helps to explain coherence, but attempts to apply SFL to rating scales are lacking. This study uses the SFL methods of clause complex analysis, thematic progression, and lexical cohesion to analyse responses in Part Two of the IELTS Speaking exam and investigates how these impact the coherence criterion in the IELTS rubric. The study discusses coherence as defined in the Fluency and coherence scale in the IELTS scoring descriptors and attempts to uncover whether SFL can be effectively applied in OP testing of non-native English speakers. Preliminary findings suggest that coherence in the IELTS scales needs to be defined more clearly to avoid rating by holistic scoring or overlapping with other descriptors. While IELTS test samples have been provided for analysis of responses in a language proficiency interview, the analysis is part of a larger project aiming to develop an alternate SFL-based rating scale for what is referred to as coherence, for use in a range of standardised oral proficiency contexts.