Architectural design representations comprise physical (models, drawings) and digital media (3D models). Multimodal combinations of physical and digital representations are commonplace in professional contexts such as design reviews or building sites, and 'making sense' of these assemblies requires complex socio-cognitive processes and professional expertise. In that context, this paper follows a 'sense-making' approach, and theorises the impact of multimodal representations in architectural design tutorials. Hybrid physical/digital representations have been built using Augmented Reality (AR) and utilised in dialogues between students' groups and design tutors. Design tutorials have been documented through video, workshops, observational notes and interviews to generate a grounded theory through iterative and reflexive coding. Following a multimodal approach, the paper frames AR as an enabler and mediator of design communication, evidenced through multimodal choreographies of physical/digital media, speech and the individual and collective performances of participants' bodies and actions in space. The resulting theory is composed of 7 concepts outlining the impact of AR in multimodal architectural communication, including the major constructs 'AR-mediated Interaction' and 'Augmented Pedagogies'. The paper outlines this conceptual taxonomy and provides fieldwork evidence supporting a methodological shift from technology-focused to sense-making observations of technology in design activity.