Educational research has shown the importance of adopting a multimodal approach to pedagogy by combining, integrating, and organizing diverse semiotic resources for learning. As aesthetic content and forms are significant aspects of multimodality, arts integration is crucial to achieve multimodal knowledge practices. In this paper, we develop a conceptual framework for analysing, discussing, and designing arts-integrated processes of teaching and learning. The framework integrates a social semiotic theory of multimodality and the Legitimation Code Theory of Semantics as a methodology through which teachers may avoid segmental learning practices in favour of an arts-integrated cumulative pedagogy. In this way, multimodal approaches to text and context relate to a social realist view on knowledge building, where boundaries between knowledge practices are weakened and meaning is condensed through processes of cumulative learning.