Brady R.T. Roberts¹, Wilma A. Bainbridge¹; ¹University of Chicago

Recent work has begun to evaluate the memorability of everyday visual symbols (e.g., !@#$%) across individuals as a new way to understand how abstract concepts are processed in memory. Symbols are highly memorable, especially relative to words, but it remains unclear what drives this memorability. Across two experiments in the current study, we first identified which visual attributes predict memory for 80 conventional symbols, then we manipulated those features in the generation of novel symbols that were designed to be memorable or forgettable. To provide a metric of each symbol’s memorability, 248 participants completed a continuous recognition memory test. This task revealed that certain symbols were reliably more memorable than others across participants. Next, participants sorted symbols spatially based on their visual similarities. A principal components analysis on the Euclidean distance between each pair of symbols revealed three visual dimensions participants used to sort symbols: thin to thick, straight to curvy, and good to poor vertical symmetry. All three of these principal components were then validated using computer vision analyses, and each uniquely predicted memory for symbols in a multiple regression. Next, generative artificial intelligence was used to create a set of novel symbols while accentuating or downplaying these predictive visual features to create memorable and forgettable symbol variants for a list of abstract words, respectively. Memory was tested across 329 participants, revealing that performance (as measured by both recognition and cued recall) was substantially improved for symbols that were designed to be memorable. Put simply, not only were symbols that were designed to be memorable easier to recognize on a memory test, but their links to associated abstract words were also improved. This work demonstrates for the first time that certain visual features drive memory for symbols and offers clear evidence that memory can be intentionally engineered.

Acknowledgements: This research was supported by a Natural Sciences and Engineering Research Council (NSERC) of Canada Postdoctoral Fellowship to BRTR and National Eye Institute Grant R01-EY034432 to WAB.