The implicit is an imprecise and heterogeneous notion that plays a role, not only in the global comprehension of utterances, but also in the interpretation of reduced phenomena like multimodal referring actions, which combine visual perception, language, and gesture. The identification of the intended referents relies on the correct identification of implicit information that is communicated with the multimodal utterance. The implicit can be linked either to the conjoined use of multiple modalities, to the cognitive status of referents and reference contexts, to the dialogue history, to inferences which occur during the interpretation process, and to interpretative effects for future utterances. For each of these types of the implicit, we focus on the determination of clues that allow the specification of reference domains (structured sub-contexts for reference resolution). We show that this notion of reference domain can integrate all these aspects with computational objectives, thus laying the foundations to a future algorithm for implicit identification.