Infants represent objects that are endpoints in motion events and show a preference for encoding the endpoint (the duck waddles into a bowl) over the starting point (the duck waddles out of a box). This asymmetry continues to appear in nonlinguistic cognition and language throughout development. This study tests whether this asymmetry also shows up in 16-month-old infants’ visual preferences for motion events, and if so, for which types of events. Infants looked longer at events depicting an “agentive” figure (e.g., duck) moving into an object (endpoint) than out of an object (starting point), and this asymmetry persisted even when the starting point object was larger and more colorful than the endpoint object and when it caused the motion of the figure. However, an asymmetry was not found when motion into/out of the endpoint/starting point involved was performed by a “nonagentive” (e.g., leaf) figure. These findings suggest that an endpoint/starting point asymmetry in infant cognition (1) extends to infants’ visual preferences of motion events, (2) shows up most strongly for events that involve an “agentive” figure, and (3) is largely unaffected by the physical saliency of the starting point object. How a visual endpoint preference may support the acquisition of spatial language is considered.