Children with Specific Language Impairment (SLI) lag behind peers with typical language (TL) in vocabulary. We ask what impact this lag has on shape bias (generalization based on shape in Naming contexts [“here’s a dax; find another dax”] but not Classification contexts [“look at this; find another like this”). Smith (2000) argues that shape bias depends on and drives vocabulary development; vocabulary is a basis for detecting covariation between objects and names, and name learning accelerates once the bias emerges. 51 three and four year-old children (16 SLI, 16 matched TL, 19 additional TL) participated in Naming and Classification tasks, a paired visual association (PVA) task, and an assessment battery. The SLI group was significantly worse at PVA and did not exhibit a shape bias. Individual differences revealed wide variation in both groups, and that shape choices in Naming were better predicted by PVA than standardized assessments.