Yael Schems Maimon¹ (), Roy Luria¹; ¹Tel Aviv University

It is well established that semantic information in long-term memory (LTM) is organized as networks of associations between concepts. Yet many memory studies rely on repetition-based paradigms to investigate learning in LTM. We previously showed that Wide Learning, a rich and varied exposure to texts, images, and videos of objects from an unfamiliar category, creates a semantic network in LTM through three processes: binding category members to one another, linking the new network to relevant pre-existing knowledge, and separating it from irrelevant concepts. The present study directly contrasted Wide Learning with repetition-based learning. Whereas repetition learners repeatedly studied the same facts and the same visual exemplars of the objects, Wide Learning participants were exposed to numerous details, diverse exemplars, and rich contextual information. The structure of the semantic network and memory performance were measured three times in both groups: before learning, about two days post-learning, and about six weeks post-learning. In both groups, learning produced a semantic network with the same three processes of within-category binding, linking to prior knowledge, and separation from unrelated concepts. Critically, despite these similarities in network structure, and despite the advantage repetition learners had in repeatedly studying the exact content and exact visual exemplars that appeared in the test, Wide Learning participants consistently outperformed repetition learners across all tasks. They showed better object naming, higher knowledge scores, and improved visual working-memory performance, both in the immediate test and after six weeks. Furthermore, in two tasks designed to assess the use of semantic networks in natural behavior, a natural language production task and a visuospatial object-organization task, Wide Learning participants demonstrated better category separation at both post-learning time points. Overall, our findings indicate that repetition learning, widely used in exam preparation and in many experimental paradigms, produced weaker memory performance than context-rich, association-based learning.