Figure 2. .

Common experiment structure. Each experiment had 3 phases. (Top) Initial Structure Learning: Participants viewed a series of scene images that had a defined statistical structure between the scene categories. The category-level structure remained constant throughout the experiment (e.g., amusement park followed by aquarium, followed by art gallery) but the specific scene exemplars changed from trial to trial (e.g., different amusement parks, different aquariums). (Middle) Simultaneous Prediction and Encoding: Participants viewed a novel exemplar of a previously seen category and were asked to use that image to predict upcoming scene categories. Crucially, viewing the cue image places potentially opposing demands on the memory system: participants can incidentally encode the novel cue exemplar and/or retrieve upcoming scene categories. (Bottom) Surprise Memory Test: Participants were shown images from the “Simultaneous Prediction and Encoding” Task, as well as category exemplars that they had not seen before. Participants were asked to report if each image was “new” or “old” with a 6-point confidence scale. This “Surprise Memory Test” was the same for all 3 experiments, but “Initial Structure Learning” and “Simultaneous Prediction and Encoding” had experiment-specific features.