Inspired by the belief that cognitive and pragmatic models of communication and discourse processing offer great potential for the study of Audiovisual Translation (AVT), this presentation will review such models and discuss their contribution to conceptualising the three inter-related sub-processes underlying all forms of AVT: the comprehension of the multimodal discourse by the translator; the translation of selected elements of this discourse; and the comprehension of the newly formed multimodal discourse—e.g. an audio described or subtitled film or performance—by the target audience. The focus will be on two models: Relevance Theory (Sperber & Wilson 1995, Wilson & Sperber 2004), which presents the most comprehensive pragmatic model of communication, and Mental Model Theory (Johnson-Laird 1983, 2006), which underlies cognitive models of discourse processing. The models will be used to discuss and question common perceptions of AVT as being ‘constrained’ and ‘partial’ translation (Bogucki 2004, Benecke 2014). A further point for discussion will then be whether and to what extent the use of cognitive and pragmatic communication models to conceptualise AVT can contribute to the empowerment of the audiovisual translator as a linguistically, culturally and socially responsible agent and creative decision maker.