Interaction sequences can be used as an abstraction of an interactive system. We can use such models to consider or verify properties of a system for testing purposes. However, interaction sequences have the potential to become unfeasibly long, leading to models which are intractable. We propose a method of reducing the state space of such sequences using the self-containment property. This allows us to hide (and subsequently expand) some of the model describing parts of the system not currently under consideration. Interaction sequences and their models can therefore be used to control the state space size of the models we create as an abstraction of an interactive system.