Children develop adult-like image memorability by age 4

The latest Brain Bridge Lab study could have major implications for designing educational materials.

By Sarah Steimer

The Brain Bridge Lab has been able to predict what images adults are more or less likely to recall — but what about children? A new study published in Journal of Experimental Psychology: General found that children develop adult-like visual sensitivity to image memorability by the age of 4 years, a finding that could have major implications for designing educational materials.

The study, led by doctoral student Xiaohan Guo and co-authored by lab director Wilma Bainbridge, an Assistant Professor in the Department of Psychology, explored whether people are born with an innate sensitivity to certain images over others, or if it takes time to accumulate enough experience in our lives. Researchers have been able to predict memory performance in adults, but it’s been unclear when consistencies in what is remembered or forgotten come online. Plus, there’s an effect called infantile amnesia, which suggests an inability in humans to remember episodic experiences that occurred between birth and 3 years of age.

A final motivation of the study, Bainbridge says, is that if you can predict what children will remember and forget, it has enormous educational implications: It could help determine what will be harder or easier material to recall, for example. Or help in choosing what images could be used to help kids recall more complicated information.

For their study, the Brain Bridge researchers used data made publicly available online by the Hartley Lab at New York University. The Hartley Lab had asked 3-, 4-, and 5-year-olds to do a memory task: The children were shown storybooks that showed different animals’ favorite places to be. For example, a dog’s favorite place was the library, and a cat’s favorite place was a playground. After a 5-minute, 24-hour, or one-week delay, the experimenters asked the children if they remembered different animals’ favorite places. They would ask what a cat’s favorite place was, and then show the children four different photos to choose from.

The Brain Bridge researchers then tested adults’ memory of the same scene images using ResMem, a machine learning model for predicting the intrinsic memorability of an image. The goal was to determine how the ResMem predictions map onto instances the children remembered or forgot the images.

What they found was that they could not predict a 3-year-old’s memory, but they could predict a 4- or 5-year-old’s memory. They also checked the results between older and younger 4-year-olds and found that the neural network did a better job at predicting the older segment versus the younger.

“As early as 4 years, kids are remembering and forgetting the same things as grownups,” Bainbridge says. “This means that they don’t need an adult-like experience out in the world. Some of the images were things that they might never have seen, like an office.”

The team also found that the 3-year-olds’ memories were unpredictable, but not because they have poor recall and aren’t remembering things. Rather, the 3-year-olds tended to remember and forget the same images as each other, meaning there is something else that drives memory in these younger children, compared to the older kids and adults.

According to Bainbridge, those designing educational materials could feed images into ResMem (which is publicly available) and determine the chances that a child 4 years and older is likely to remember the picture. It is also possible that the memorability of one item can translate to other items that are associated with it. A storybook could be designed around a memorable image that could boost a child’s ability to learn concepts associated with that image.

“Pictures are an important part of our daily lives,” Bainbridge says. “The fact that you can use this to design education materials is really exciting.”

As for implications on the neuroscience side, Bainbridge says her lab is interested in how this research could be used to create brain maturation maps. If a child’s brain is more adult-like, what they recall would be more adult-like and predictable by the neural network. Such maps could be used to determine if a child is following a typical or atypical maturation timeline.

Most memorable:

1. grocery (store): .92
2. subway: .80
3. candy (store): .75
4. basketball (field): .72

Most forgettable:

1. playground: .25
2. mountain: .40
3. pool: .45
4. bedroom: .50