The purpose of this study was to examine early mathematical skill differences among preschool children in US Head Start classrooms. Latent class analysis based on six early mathematical subtest scores (i.e. counting aloud, measurement, counting objects, numbers and shapes, pattern recognition, and grouping) from a sample of 279 Head Start children yielded evidence for a high-achieving class, a typical-achieving class, and a low-achieving class, relative to other children attending Head Start. Average skill profiles of the three latent classes were in general parallel to one another, reflecting that most of the differences across latent classes were in level rather than type of skills. Changes in subtest scores over a 3-month interval indicated that the skill levels of the low-achieving class at time 2 were still below those of the typically achieving class at time 1. These findings provide evidence for skill variability among children enrolled in Head Start and a group of children who appear unlikely to demonstrate the skill level of their peers without additional instruction or intervention.