Thelen and colleagues recently proposed a dynamic field theory (DFT) to capture the general processes that give rise to infants' performance in the Piagetian A-not-B task. According to this theory, the same general processes should operate in noncanonical A-not-B-type tasks with children older than 12 months. Three predictions of the DFT were tested by examining 3-year-olds' location memory errors in a task with a homogeneous task space. Children pointed to remembered locations after delays of 0 s to 10 s. The spatial layout of the possible targets and the frequency with which children moved to each target was varied. As predicted by the DFT, children's responses showed a continuous spatial drift during delays toward a longer term memory of previously moved-to locations. Furthermore, these delay-dependent effects were reduced when children moved to an "A" location on successive trials, and were magnified on the first trial to a nearby "B" location. Thus, the DFT generalized to capture the performance of 3-year-old children in a new task. In contrast to predictions of the DFT, however, 3-year-olds' responses were also biased toward the midline of the task space - an effect predicted by the category adjustment (CA) model. These data suggest that young children's spatial memory responses are affected by delay- and experience-dependent processes as well as the geometric structure of the task space. Consequently, two current models of spatial memory - the DFT and the CA model - provide incomplete accounts of children's location memory abilities.