Following the methodological approach of multimodal conversation analysis, our study focuses on the emergence of a jointly constructed reference during a card game interaction: the participants need to find a common solution in order to continue their activity. The data collected in situ allow us to apprehend the different verbal and non-verbal resources mobilised by the players in order to show that in social interaction reference construction is a joint achievement that involves various types of resources which are temporally finely tuned (among others joint visual attention on the object gesturally put in focus). We also show that once a referent-function association is established and grounded, it can be "activated" later on by using an iconic gesture.