Conversation analysis (CA) continues to accrue interest within clinical linguistics as a methodology that can enable elucidation of structural and sequential orderliness in interactions involving participants who produce ostensibly disordered communication behaviours. However, it can be challenging to apply CA to re-examine clinical phenomena that have initially been defined in terms of linguistics, as a logical starting point for analysis may be to focus primarily on the organisation of language ("talk") in such interactions. In this article, we argue that CA's methodological power can only be fully exploited in this research context when a multimodal analytic orientation is adopted, where due consideration is given to participants' co-ordinated use of multiple semiotic resources including, but not limited to, talk (e.g., gaze, embodied action, object use and so forth). To evidence this argument, a two-layered analysis of unusual question-answer sequences in a play episode involving a child with autism is presented. It is thereby demonstrated that only when the scope of enquiry is broadened to include gaze and other embodied action can an account be generated of orderliness within these sequences. This finding has important implications for CA's application as a research methodology within clinical linguistics.