Certain visual qualities make symbols more memorable or forgettable, suggesting we can strengthen these simple communication devices

New research out of the Brain Bridge Lab points to clear and interpretable design rules for creating better symbols.

By Sarah Steimer

Symbols can transcend culture and language, making them easily interpretable across the globe. Consider a triangle appearing on a media file: It’s a simple, standardized image known worldwide to mean “play.”

What’s less clear are what aspects of a symbol make it more or less memorable, which plays a crucial role in visual communication. But a new study published in PNAS helps clarify how we give definition to abstract concepts with symbols and how we later process them in our visual memory systems.

In his previous work at the University of Waterloo, Brady Roberts found that symbols are better remembered than their word counterparts — for example, “$” versus “dollar” — meaning they’re likely processed more visually than as a word.

Now working as a postdoctoral scholar in the Brain Bridge Lab, Roberts wanted to explore the intrinsic memorability of symbols, or how reliable a symbol is remembered across people. He and his coauthor, Wilma Bainbridge, associate professor in the Department of Psychology and lead investigator at the lab, also explored the visual features and semantic aspects of a stimulus that make them more memorable. 

“Symbols are very simple designs, but they often carry this deep, abstract meaning,” Roberts says. “We can look at the combination of those two things (conceptual and visual features) to see which aspects predict later memory.”

First, the team tested the memorability of known or conventional symbols — those that people already had an attached meaning to. During this process, they also measured the visual attributes of those symbols to determine which visual attributes predicted memory. For this part of the study, participants were asked to sort the symbols spatially on their screen, based on how similar they were perceived. For instance, a plus sign might be grouped with a Christian cross.

Using that data, the researchers were able to capture three primary metrics that participants were using to sort the symbols: thickness, curvature, and vertical symmetry. These three metrics also predicted memory performance: Symbols tended to be better remembered if they were thin, used straighter lines, and were more asymmetrical in the left-right mirror.

The team then gathered lists of abstract words using an online English word database, and then removed any words that already had related symbols. The remaining words were fed into a generative AI platform (the Dall-E 3 image generator) to create brand-new symbols. Using the identified memorable features, the researchers then created a memorable and a forgettable symbol for each abstract word. 

For the second half of the study, participants were shown a novel symbol and its associated abstract word. To bind meaning to the symbol, participants were asked to rate how well the symbol represented the word. They reviewed and rated a series of symbol-word combinations and — after a brief delay — were then given a memory test for those symbols and words.

“Critically, in the memory test, we actually captured two aspects of memory,” Roberts explains. “We captured recognition memory by asking participants to say whether or not they remember just the symbol from earlier, regardless of whether they can remember the word that was tied to it. And then immediately afterward, we said, Okay, now that you said whether or not you recognize the symbol, tell us what you think the word was.”

This captured cued-recall responses, allowing the researchers to explore changes in memorability for the symbol itself — which they had manipulated with AI —  and whether that extended its memory effects to the associated concept.

Roberts explains that the first experiment helped identify symbols’ specific visual properties that matter for later memory. The second experiment, then, showed that symbols can be manipulated to better help people both remember symbols and their meanings.

“The memorable symbols in experiment two were not only better recognized, but they also afforded more reliable access to their word associates, suggesting that visual design can support both learning of the design itself as well as comprehension about the design's meaning,” Roberts says.

The research points to clear and interpretable design rules, which Roberts notes go beyond the current black-box AI models that exist, making the metrics more interpretable to designers, engineers, and even policy makers (think: more easily understood traffic signs or digital menus).

“We're hopeful that this work will trigger a bit of an evolution in graphic design, whereby everyday things like symbols, icons, and logos are optimized for human cognition using these data-driven visual design methods,” Roberts says.

The research could serve as a roadmap for improved clarity in communication: Words can be muddy — i.e., “play” can mean start, or it can refer to a theatrical performance or a strategy in sports. Symbols, on the other hand — such as a triangle on a media file — are far cleaner mappings to an idea.