The present article investigates spatial- and social-cognitive processes in toddlers' mapping of concepts to real-world events. In 2 studies we explore how event segmentation might lay the groundwork for extracting actions from the event stream and conceptually mapping novel verbs to these actions. In Study 1, toddlers demonstrated the ability to segment a novel multiaction event by selecting a single action for behavioral reenactment when prompted. In Study 2, a single action embedded in the event sequence was specified as the referent for a novel verb through cues to support intentional inference. As in Study 1, toddlers spontaneously segmented the sequence. In addition, they mapped a novel label to the embedded action specified by the novel label and intentional cues. These data are consistent with current thinking that suggests a convergence of spatial and social cognition in verb learning. In the present research, toddlers built upon event segmentation skills and used intentional inference when mapping verbs to actions embedded in the motion stream. (Contains 3 tables and 2 figures.)