New assessment types that include multimodal and digital elements are increasingly being used to assess university students' 'soft skills' such as communication, as well as their science content knowledge. However, very little is known about how or how well such products assess communicative elements, particularly when these elements are so intricately linked with disciplinary knowledge. This paper presents a novel way of looking at these new digital assessments in science. Using semantic density, a concept from the framework of Legitimation Code Theory (LCT) that conceptualises complexity, we consider how to characterise learners' communication of complex science in the digital products. Results show that successful products 'negotiate' complexity in distinct ways and that language and image work together in the products to build meaning. This approach is a 'first step' in characterising discipline-based communication skills through the development of a preliminary conceptual framework that will inform pedagogies and assessment practices surrounding student-generated digital products, in an effort to improve outcomes for science students.