This study combined an event schema approach with top-down processing perspectives to investigate whether high-functioning children and adults with autism spectrum disorder (ASD) spontaneously attend to and remember context-relevant aspects of scenes. Participants read one story of story-pairs (e.g., burglary or tea party). They then inspected a scene (living room) of which some objects were relevant in that context, irrelevant (related to the non-emphasized event) or neutral (scene-schema related). During immediate and delayed recall, only the (TD) groups selectively recalled context-relevant objects, and significantly more context-relevant objects than the ASD groups. Gaze-tracking suggests that one factor in these memory differences may be diminished top-down effects of event schemas on initial attention (first ten fixations) to relevant items in ASD.