The Effects of Prior Knowledge on Episodic Memory Depend on Retrieval Goals and Orientation

Prior knowledge can enhance memory for both schema-congruent and schema-incongruent events. According to the SLIMM framework, memory performance follows a U-shaped pattern, with both highly expected and highly surprising events being better remembered than neutral ones. We tested this prediction across six experiments in which participants viewed scenes containing congruent or incongruent objects, rating each as expected or surprising using a four-level scale. Memory was assessed through three tests: (1) item memory (old/new) recognition categorized by recollection and familiarity; (2) "congruency memory," where participants recalled whether objects had been expected or surprising; and (3) forced-choice associative memory, where participants selected the scene associated with an object. Results showed that schema effects on memory varied by test type and retrieval orientation. Item memory consistently favored expected over surprising objects, primarily via recollection - contrary to SLIMM predictions. We propose that semantic association generation during testing enhanced recollection of expected objects. After controlling for semantic retrieval, associative memory revealed a benefit for surprising events, attributed to greater distinctiveness. Congruency memory was the only test producing a reliable U-shaped pattern. We suggest participants tagged events as "expected" or "surprising" during study and could later recall the tag, even without recalling the associated scene. However, this U-shaped pattern depended on retrieval orientation: when participants focused on surprise, the U-shape emerged; when focusing on expectedness, only surprising events showed advantage. These findings reinforce the complex factors modulating prior knowledge effects on episodic memory.