Summary: For years, the “London Taxi Driver” study suggested that elite sailors have larger brain regions. However, new research suggests that this may not be the case for most people.
Using Deep Learning and Convolutional Neural Networks, researchers analyzed MRI scans of healthy adults to see if AI could detect subtle structural patterns that predict navigational ability. Surprisingly, even the most sophisticated AI tools have found no measurable connection between the shape or size of brain regions—including the hippocampus—and how well a person finds their way.
Important Facts
- AI vs. Tradition: While simple size measurements have suggested a link, deep learning models capable of detecting complex structural patterns have failed to detect a “navigation signal” in healthy adults.
- Hippocampus and Thalamus: The study directly compared the hippocampus (the common “GPS” of the brain) and the thalamus (control region). Navigation performance was indistinguishable between the two based on structure alone.
- Healthy Adults: Data were collected from 90 participants (average age 23) who learned routes in a virtual environment.
- Disease Prediction vs. Conduct: The researchers noted that while AI excels at predicting disease states (such as Alzheimer's), it struggles to map everyday behavioral tasks such as spatial navigation.
- To redefine research: The findings challenge the notion that “more real estate” in the brain equates to better cognitive function, suggesting that work again communication may be more important than the macroscopic structure.
Source: T Arlington
Steven Weisberg, a researcher at the University of Texas at Arlington, found that advanced artificial intelligence tools could not reveal a clear connection between the structure of the brain and the ability to navigate in healthy young people-challenging long-term ideas about how the brain helps us find our way.
For decades, many in the scientific community believed that people with superior navigation skills—such as learning quickly and remembering complex routes—may have larger or differently shaped brain regions than others.
For example, a famous study of London taxi drivers suggested that intensive navigation training may lead to more “homes” in certain parts of the brain.
In a new study, Dr. Weisberg and his team, including University of Florida Ph.D. candidate, Ashish Sahoo, tested that hypothesis using new analytical techniques, including deep convolutional neural networks and other machine learning models that can detect subtle patterns in brain scans beyond simple size measurements.
Despite these advanced methods, researchers have found no measurable connection between brain structure and navigational performance in healthy young adults.
Understanding navigation is important given its real-world impacts on everyday life, including independence, memory, and dementia risk.
“With the quality of the data we have from MRI scans and this population of healthy adults, there doesn't seem to be a signal available using these advanced metrics,” said Weisberg, who conducted research at the University of Florida before joining UT Arlington last fall as part of the RISE 100 initiative.
The study, published in a peer-reviewed journal Neuropsychologyanalyzed data from 90 participants with an average age of 23.1 years. Participants learned two routes using a virtual environment.
The results showed little difference in navigation performance when two brain regions were compared: the thalamus, which served as a control area, and the hippocampus, a region typically associated with navigation and memory.
While the findings point to the limits of what AI can currently reveal about everyday cognitive abilities, the technology remains a powerful research tool. Weisberg said more robust models could find the difference in future studies.
“Our study should be one data point in a large area of what AI can tell us about how brain structure and function map to behavior,” Weisberg said.
“Machine learning and AI have been very successful in predicting disease states. What we're interested in is that these models have utility in behavioral applications—things like cognitive training or education.”
Future research will focus on larger samples and older adults, Weisberg said.
“Our ability to navigate enables basically everything we do. Studying how the brain supports navigation helps us understand what's needed when it's going well and what's missing when it's not.”
Important Questions Answered:
A: That's not the case. Intensive, multi-year training (such as that of London taxi drivers) can still create structural changes. However, for the average healthy person, being a “good” or “bad” sailor is not written in the physical form of your brain. It's most likely your neurons fire together, not how much space they occupy.
A: Physically, on a macroscopic level, yes. This study suggests that “superior” navigation skills in young adults do not require a large hippocampus. The difference may be “under the hood” in very small fibers or chemical signals that AI-processed MRI scans have yet to detect.
A: Navigation is often the first skill to decline in Alzheimer's. If we cannot find a systematic “basis” for good navigation in healthy people, it suggests that early diagnosis should be the focus of conduct changes and to work connection than just looking at the size of the hippocampus in a slice.
Editor's Notes:
- This article was edited by a Neuroscience News editor.
- The journal paper is fully revised.
- More content has been added by our staff.
About this AI and neuroscience news
Author: Drew Davidson
Source: T Arlington
Contact person: Drew Davison – UT Arlington
Image: Image posted in Neuroscience News
Actual research: Open access.
“Deep learning methods for mapping individual differences in macroscopic neural structure through spatial navigation behavior variability” by Ashish K. Sahoo, Hajymyrat Geldimuradov, Kaleb E. Smith, Aaron Zygala, Yiming Cui, Mahsa Lotfollahi, Kuang Gong, Alina Zare, and Steven M. Weisberg. Neuropsychology
DOI:10.1016/j.neuropsychology.2025.109352
Abstract
Deep learning methods for mapping individual differences in macroscopic neural structure through spatial navigation behavior variation
Understanding the connections between human brain structures and individual differences in behavior is an ongoing endeavor, challenged by the complexity of the brain.
Previous methods, limited to measurements of simple neural structure such as brain volume or cortical thickness, have given way to more advanced modeling methods.
Empirical evidence using these simple metrics sometimes shows that hippocampal structure is related to individual variables of spatial navigation ability, especially in older people or navigation professionals (such as London taxi drivers). Yet more powerful studies, previously enrolled in normal adults, have not revealed an association between hippocampal volume and navigational ability.
Here, we follow a data-driven approach to develop and compare deep learning methods (graph convolution neural networks, GCNN; 3DCNN) to analyze how complex features of brain structure predict spatial navigation ability in young people.
To that end, we trained GCNNs and 3DCNNs on a T1 MRI dataset (N = 90) to predict navigational ability as measured by a virtual reality spatial memory test in which participants created a map as accurate as they would in a virtual environment.
For all methods, we found a weak prediction value on the out-of-sample test data, despite the good fit to the training data.
These results may indicate the need for much larger data sets, including extensive behavioral measures (as this study was limited to one measurement) to improve predictions but may also support the idea that hippocampal structural characteristics may not be the main factor associated with navigational ability in healthy young adults.
