New AI-Generated Visual Anagrams Transform Brain Research

Researchers at Johns Hopkins University have developed innovative artificial intelligence-generated images known as “visual anagrams,” which are helping to advance understanding of how the human brain processes visual information. These images can appear to be one object but resemble something different when rotated, opening new avenues for psychological research.

The study, funded by the National Science Foundation Graduate Research Fellowship Program, addresses a crucial need for standardized stimuli in visual perception studies. Lead researcher Tal Boger emphasized the significance of these images, stating, “These images are really important because we can use them to study all sorts of effects that scientists previously thought were nearly impossible to study in isolation—everything from size to animacy to emotion.”

Understanding Visual Anagrams

Visual anagrams, described as orthogonal transformations, are crafted using a new AI tool. The team produced various images, including one that depicts both a bear and a butterfly, another that can be seen as an elephant or a rabbit, and a third that resembles both a duck and a horse. This duality allows researchers to explore complex perceptions regarding size and object recognition.

Initial experiments focused on how participants interpret the real-world size of objects, a longstanding puzzle in perception science. According to Boger and his colleagues, understanding whether individuals respond to an object’s size or its other visual properties—such as shape, color, or texture—has proven challenging. The brain employs a mechanism called size constancy, which enables stable perception of an object’s size despite changes in its retinal image size.

Insights from Experimental Findings

The findings from the visual anagram research indicate that classic real-world size effects are still present, even when participants interact with rotated versions of the same image. For instance, previous research revealed that individuals prefer images of larger animals, such as bears, to be depicted in a size that reflects their real-world dimensions compared to smaller animals like butterflies. In line with this, the current study found that when participants adjusted the bear image to its ideal size, they made it larger than when adjusting the butterfly image, despite both being the same image in different orientations.

The research team aims to utilize visual anagrams to further investigate how people respond to animate and inanimate objects. This is particularly relevant given that different areas of the brain process these categories distinctly. Future applications may include creating anagrams where one orientation resembles a truck, while another looks like a dog, offering a broad scope for psychological and neurological experiments.

The study is set to be published in an upcoming issue of Current Biology, marking a significant step forward in visual perception research. The potential applications of visual anagrams are extensive, and the research team anticipates that this innovative approach will be adopted by other researchers for various purposes in the field.

Dr. Tim Sandle, Editor-at-Large for science news at Digital Journal, highlights the interdisciplinary implications of this research, noting its relevance in psychology, neuroscience, and beyond.