There is growing interest in the construct of "transduction", first introduced by (Kress, Cope and Kalantzis (eds), Multiliteracies: Literacy learning and the design of social futures pp.153-161, Routledge, 2000), p. 159) to name how meanings in one mode are remade in another. Science educators now broadly agree that students need to learn how to interpret, make, co-ordinate and integrate meanings in multimodal representations of scientific claims. The question of how exactly this transductive capability is enabled, constrained or theorised is receiving renewed attention. In this paper, following a pragmatist socio-cultural perspective, we propose that transduction in science entails creative reasoning enabled by both cognitive and semiotic resources. We claim that that this process entails students achieving both complementarity as well as coherence or consistency of meanings across modes, but is not reducible to mono-modal "translation" of a singular meaning. Because different modes do different work in meaning-making in science, we claim instead that transduction is the process of seeking and recognising both reiteration and complementarity of linguistic and non-linguistic meanings that constitute the nature of concepts and processes in science. While this "content" can be re-represented in verbal shorthand, science meaning-making and discourse are not reducible to this mode. We draw on relevant literature and two indicative examples of student representation construction to support our case. We focus on how students attend to correspondence and coherence requirements to transduce meanings across modes. In our micro-ethnographic analyses of students' collaborative reasoning on the topic of astronomy, we interpret data from multiple video and audio capture. Students use 3D models, 2D representations, gesture and talk as semiotic resources to reason about scientific concepts.