The immune responses that defend us against pathogens are driven by stochastic processes amongst populations of cells. Enormous progress in immunology over the last few decades has identified most of the components of this complex system, including the cell types and the molecules used for communication. But understanding how the dynamics of the system emerge from these components remains a major challenge. The presence of large numbers of cells following random migration, interaction, birth, and death rules suggests we take a statistical physics approach. The major hurdle to this approach is the difficulty of observing these processes in their native context at system scales. I will describe recent progress in developing a zebrafish and light-sheet microscopy system to address this gap, focused on the dynamics of adaptive immunity. As an example of this approach, I will discuss our recent observations of naive T cells in the live zebrafish, and describe how T cell motility forms a behavioral manifold that generates a broad range of length scales of exploration across the population. I will conclude by briefly discussing new opportunities opened by this system in physics and immunology.