Neuroscientists have known for decades that neurons in the hippocampus of the brain called place cells fire, or “spike”, as a rat moves through its environment, with the activity of individual place cells corresponding to specific locations in space. More recently, scientists from UC Berkeley and collaborators found that theta brain waves— which reflect the coordinated activity of large groups of neurons — can also indicate a rat’s location, at least when the waves are strong.
But strong theta waves are relatively rare. Scientists more frequently observe weaker, more intermittent theta waves. A new study from the lab of Helen Wills Neuroscience Institute (HWNI) Adjunct Professor Friedrich Sommer has now shown that these more common, weak theta waves also encode location, opening the door to new avenues of research about how the brain carries information.
The study was published on February 13, 2026 in Nature Communications, and was led by Gautam Agarwal, a Berkeley Neuroscience PhD alum and former research scientist in the Sommer lab, which is located in HWNI’s Redwood Center for Theoretical Neuroscience. Agarwal is now an assistant professor of neuroscience in the Department of Natural Sciences at Pitzer and Scripps Colleges.
In the Q&A below, Sommer and Agarwal explain their study and how it could impact the fields of both neuroscience and artificial intelligence.
Q: Why did you do this study?
A: Brains generate electrical rhythms ("brain waves"), and their role in brain function is poorly understood. Our earlier work (Agarwal et al., 2014) showed that particularly strong, low-frequency rhythms known as theta rhythms carry precise information about the location of a rat as it runs through its environment. This was exciting because it showed that this seemingly non-specific brain signal could be as informative about behavior as more precise measurements involving single neurons. However, such strong, coherent rhythms are relatively rare across species and brain states. More commonly, brain waves such as theta appear in weaker, intermittent bouts. Here we asked whether this more typical, weaker form of theta activity also carries behavioral information.
Q: How did you do the study and what were your major findings?
A: In collaboration with Seiji Akera, Brian Lustig, Eva Pastalkova, and Albert Lee, we studied hippocampal activity in rats during periods when they were stationary and theta rhythms were weak. We found that our earlier [location] decoding method performed much worse in this state. Using computational models, we showed that this drop can be explained by shared variability ("noise") in neural activity. Based on this insight, we developed a simple neural network that can detect subtle, behavior-encoding oscillations (which we call "place thetas", or pThetas) in the data. This approach allowed us to accurately decode the animal’s location both during movement and while stationary.
Q: How have your findings advanced our understanding of how the brain works?
A: We had previously expected that when theta rhythms become weak or irregular, neurons are unable to carry information through coordinated oscillation. Our results challenge this view. We show that even weak theta contains a structured, information-carrying component related to location. This component appears to reflect coordinated activity among place-encoding neurons, and is distinct from the stronger, more prominent theta signal that is more closely linked to other cell types.
Q: What impact do you think this study might have on the field of neuroscience?
A: Our findings suggest that weak and intermittent brain rhythms, as are common across many brain regions and species, can carry meaningful information when analyzed appropriately. This offers new ways of studying brain activity in conditions where strong, regular rhythms are absent.
Q: Could this study have implications for artificial intelligence as well?
A: Our results show that detailed information about behavior can be extracted from brain rhythms alone, suggesting that rhythms and spikes work together in neural computation. This could inspire more efficient neuromorphic AI systems that leverage interactions between rhythmic activity and discrete events, indeed some of which are being implemented in models such as recurrent neural networks (RNNs).
Q: What are your next steps in this line of research?
A: A key next step is to better understand how rhythmic activity and spiking interact in the brain, and how this interaction supports behavior across different conditions. We are approaching this by applying our tool to other modalities, such as human EEG recordings.
Funding for this research was provided by NIH grant #1R01EB026955 (Friedrich Sommer), the Howard Hughes Medical Institute (Albert Lee), and startup funds from the Department of Natural Sciences at Pitzer and Scripps Colleges (Gautam Agarwal).
Additional Information
-
Deciphering hippocampal place codes in weak theta rhythms, Agarwal et al. 2026, Nature Communications
