In this paper we present initial information concerning a new direct observation system--the Observational System for Recording Physical Activity in Children-Preschool Version. The system will allow researchers to record young children's physical activity levels while also coding the topography of their physical activity, as well as detailed indoor and outdoor social and nonsocial contextual information. With respect to inter-observer agreement (IOA), the kappa and category-by-category agreement mean of those obtained for the three illustrative preschools were generally above 0.80. Hence, our IOA data indicated that trained observers in the three preschools frequently agreed on the eight observational categories and accompanying codes. The results for preschoolers' level of physical activity indicated they spent the majority of observational intervals in sedentary activity (i.e., more than 80% intervals) and were observed in moderate to vigorous physical activity much less frequently (i.e., 5% or fewer intervals). For the 15 indoor and 12 outdoor activity contexts, variability across both the activity contexts and the three preschools were evident. Nevertheless, three classroom contexts--transition, snacks, and naptime--accounted for the greatest proportion of coded activity contexts for the children. In the three preschools, 4 of 17 physical activity types--sit and squat, lie down, stand, and walk--accounted for the topography of much of children's physical activity behavior. Systematic observation of more representative preschool samples might better inform our present understanding of young children's physical activity in community preschool programs. (Contains 2 tables and 3 figures.)