There is currently a lack of systematic study examining the integration of verbal-visual elements in science textbooks. In particular, few research has investigated how different types of visual representations (e.g. photograph, diagram, table) vary across the major written genres of science (e.g. information report, explanation). Based on the perspective of social semiotics, this paper aims to understand the co-dependency between visual representations and written genres by analysing a corpus of multimodal texts from a major Australian textbook series. Quantitative and qualitative analyses reveal a number of interesting patterns in terms of how several types of representations tend to appear with certain genres, as well as possible reasons for this co-dependency. In particular, the analysis accounts for why and how: (a) photographs and tables are used to support information report, (b) diagrams and photographs for explanation, and (c) diagrams and tables for experimental account. This study provides empirical evidence to support the importance of analysing multimodal genre for science teaching and learning as well as science education research.