How and when infants and young children begin to develop emotion categories is not yet well understood. Research has largely treated the learning problem as one of identifying perceptual similarities among exemplars (typically posed, stereotyped facial configurations). However, recent meta-analyses and reviews converge to suggest that emotion categories are abstract, involving high-dimensional and situationally variable instances. In this paper we consult research on the development of abstract object categorization to guide hypotheses about how infants might learn abstract emotion categories because the two domains present infants with similar learning challenges. In particular, we consider how a developmental cascades framework offers opportunities to understand how and when young children develop emotion categories.