Abstract
Understanding the interplay between different cognitive factors in long-term memory is essential to understanding the nature of mental representations of our visual environments. In this study, across three online experiments, we investigated how task priorities and expectations arising from semantic knowledge based on schemas may concur to determine long-term memory for objects in real-world visual scenes. In Experiment 1, participants intentionally memorised a sequence of scenes, 35% of which contained a semantically inconsistent object. In Experiment 2, participants searched for target letters within the same scenes, and this was followed by a surprise memory test. Experiment 3 required intentional memorisation but presented 50% of the scenes with an inconsistent object. Presentation times (10s) and intervals between scenes (2.5s) were the same across experiments. Memory for two objects in each tested scene was examined in a four alternative-forced-choice paradigm, where participants also rated their confidence in each response on a Likert scale. One object, tested as first, was either (50% of the trials) semantically consistent or inconsistent with the scene, and the other object was always consistent. The results showed higher accuracy and confidence in intentional (Experiments 1 and 3) than incidental (Experiment 2) memorisation. Accuracy was higher for consistent than inconsistent objects in all experiments, suggesting that schemas guided retrieval, which was also supported by a bias to select consistent foils in wrong responses. Equating the number of consistent and inconsistent scenes in Experiment 3 did not considerably reduce the consistency advantage, showing that semantic knowledge played a greater role in the effect than any modulations of reliance on schemas due to the short-term task context. Confidence was higher for correct responses for inconsistent than consistent objects, especially in intentional memorisation, suggesting that, if correctly encoded, a memory trace that violates schema expectations has higher distinctiveness and precision.