A groundbreaking new study from NYU Langone Health has illuminated the complex ways in which the brain manages to store multiple memories without blending or erasing vital pieces of past information. This discovery centers on an intriguing subset of neurons within the hippocampus, an area known for its role in memory formation. Researchers found that approximately 25% of these hippocampal CA1 neurons act as hubs that facilitate the seamless transmission of information from one region of the brain to another, effectively functioning like a biological switchboard managing countless memory signals.

For decades, neuroscientists have grappled with the paradox of how the brain maintains a delicate balance between adaptability and stability—retaining established memories while accommodating new information. This study provides fresh insights into this dilemma by exploring the neural interplay along pathways between the hippocampus and the neocortex. Specifically, the focus was on the CA3 and CA1 regions of the hippocampus and their communication with the retrosplenial cortex, a crucial site involved in navigation and spatial memory recall.

The CA3 region is known to send rapid and fluid streams of information, and, remarkably, the research demonstrated that most of these incoming signals converge on a small cohort of CA1 neurons. These same neurons then process and relay information to the retrosplenial cortex, but in a distinctly different firing pattern, which creates an independent outgoing communication channel. This dual functionality allows the neurons to multiplex incoming and outgoing signals without blending them, preserving the clarity of each memory trace.

This complex system can be likened to an advanced electronic switchboard that directs multiple phone calls without their lines crossing, ensuring that new experiences are integrated into the brain’s map without disrupting existing knowledge. The retrosplenial cortex benefits from this arrangement by maintaining a stable representation of the environment—essential for spatial navigation—while the hippocampal regions continue adapting and learning from the ongoing stream of experiences.

Dr. Joaquín Gonzalez, a postdoctoral fellow and co-lead author of the study, emphasized the significance of this firing pattern adjustment: “Instead of recruiting new neurons for every novel experience, the brain modifies the firing patterns of a stable cellular core, thereby organiz-ing information effectively and safeguarding previously encoded memories.” This mechanism highlights the brain’s remarkable ability to adapt dynamically while retaining long-term memory integrity.

Interestingly, the study also uncovered that these pivotal CA1 neurons are not confined to processing information during active waking hours—they remain engaged during sleep, participating in sharp-wave ripple events that are critical for memory consolidation. This nocturnal activity is believed to involve the replay and reinforcement of memory traces, further stabilizing learning while the brain rests.

The persistence of activity in these core neurons during sleep suggests a continuous information relay between the hippocampus and cortex, facilitating the integration of memories into long-term storage. By employing the same neural architecture for both daytime encoding and nighttime replay, the brain ensures that its memory network remains both flexible and coherent.

Dr. Mihály Vöröslakos, another postdoctoral researcher on the team, highlighted the methodological breakthrough that made this discovery possible: “Our ability to simultaneously record hundreds of individual neurons across multiple connected brain regions in freely moving mice was instrumental. This approach revealed the nuanced patterns of communication that traditional recording methods could not detect.”

Moreover, the study’s findings carry potential implications beyond basic neuroscience. The analogy between neural switchboards and artificial intelligence systems underlines a key challenge in AI—catastrophic forgetting—where machines lose previously learned information upon training on new tasks. By understanding how the mammalian brain protects old memories while learning new ones, scientists hope to inspire the development of next-generation AI technologies that can continuously learn without forgetting.

Dr. György Buzsáki, co-senior author and a renowned neuroscience expert, suggested that this research might shed light on neurodegenerative conditions such as Alzheimer’s disease, where memory circuits deteriorate. “Our discovery of a ‘memory switchboard’ within the hippocampus could provide vital clues about the early mechanisms of memory failure in such diseases,” Dr. Buzsáki remarked.

The experiment involved training six mice to traverse a linear track rewarded at both ends with water. As the animals moved, high-density electrode arrays captured the simultaneous neural activity across hippocampal and cortical regions, while behavioral tracking allowed researchers to correlate precise brain signals with physical navigation and exploration.

Further analysis during sleep revealed that while the original patterns of activity were replayed, they mutat-ed dynamically within and between the hippocampus and neocortex, underscoring a sophisticated neural choreography that supports memory consolidation and flexibility concurrently.

Despite the advances, the authors caution that extrapolation to human brain function requires further research. The controlled environment of the study and differences between species mean that confirming the presence of similar switchboard mechanisms in humans remains an open question.

As they look to the future, the research team plans to explore whether comparable subspace communication channels exist in other areas of the brain responsible for diverse types of memory processing. Such investigations could lead to a more comprehensive neural map of memory architecture, with profound impact for both neuroscience and artificial intelligence.

This research was supported by several grants from the National Institutes of Health, highlighting the critical role of federal funding in fostering cutting-edge brain science. The collaborative effort included leading neuroscientists and scholars from NYU Langone Health and NYU Grossman School of Medicine.

By unlocking new dimensions of how individual neurons coordinate complex memory signals, this study offers unprecedented insights into one of biology’s most enduring mysteries—how the brain manages to be both ever-changing and enduring, preserving the richness of past experience while embracing the potential of new learning.

Subject of Research: Animals
Article Title: Subspace communication in the hippocampal–retrosplenial axis
News Publication Date: 13-May-2026
Web References: http://dx.doi.org/10.1038/s41586-026-10481-z
References: Nature, May 13, 2026; DOI: 10.1038/s41586-026-10481-z

Keywords

Memory, Long term memory, Memory formation, Memory processes, Spatial memory, Sleep, Hippocampal neurons, CA1 cells, CA3 cells, Hippocampus, Hippocampal circuits, Artificial intelligence

The idea that memories might not correspond to real events but could actually be illusions created by chance from cosmic static has been discussed in physics for more than a century. Recently, three physicists examined the logic behind this idea and found that arguments on both sides may be fundamentally circular.

A recent study published in the journal Entropy by Santa Fe Institute Professor David Wolpert, physicist Carlo Rovelli, and Jordan Scharnhorst revisits the Boltzmann brain hypothesis. This thought experiment, based on statistical mechanics, suggests that random fluctuations in entropy could, in theory, create a fully formed brain with false memories and a sense of a coherent past.

Rather than trying to prove or disprove the Boltzmann brain hypothesis, the researchers focused on identifying a structural flaw in the way scientists have debated the issue.

Where the Logic Breaks Down

The Boltzmann brain paradox comes from the H theorem, developed by Austrian mathematician and physicist Ludwig Boltzmann. This idea is key to statistical mechanics and supports the second law of thermodynamics, which explains why disorder (or entropy) increases over time and why we perceive time as moving forward. However, the H theorem itself treats the past and the future identically in its equations.

This symmetry creates a problem. If entropy can decrease in the future just as easily as it increased in the past, then the patterns that form our memories could just as likely come from random fluctuations as from real events. In other words, our memories might not necessarily correspond to actual past events.

The usual response is that this scenario is extremely unlikely. The chance of a functioning brain forming from random thermal noise is so small that it would take much longer than the current age of the universe for it to happen. However, the new study shows that this argument depends on assumptions that may not even be justified.

A Never-Ending Circle

To clarify the debate, the researchers created a mathematical framework that models the universe’s entropy as a time-symmetric Markov process, which they call the “entropy conjecture.” In this framework, they identified a key issue: physics alone cannot determine which moment in time to use as a reference point. That choice must be assumed.

This assumption leads to circular reasoning. Arguments against the Boltzmann brain hypothesis, including those that appeal to the second law of thermodynamics, usually assume that our memories accurately record real events. Yet the main reason to trust our memories is that the second law suggests they should be reliable. In other words, the conclusion relies on the premise, and the premise relies on the conclusion.

Arguments in favor of the hypothesis show the same circularity. The study finds that the Boltzmann brain hypothesis and the standard “past hypothesis,” which assumes the universe began in a low-entropy state at the Big Bang, have the same structure. Each approach analyzes the problem from a different moment in time, changing only which moment it treats as fixed.

Reframing the Question

The researchers stress that their findings are meant to diagnose the problem, not to give a final answer. Their study does not decide whether the Boltzmann brain hypothesis is true or whether our memories are real, but it does show that current arguments do not properly answer the question.

The team formalized the entropy conjecture as a mathematical process and revealed a problem earlier studies overlooked: every argument in this debate depends on assumptions about which facts to treat as fixed, and physics alone cannot resolve the issue.

Fundamentally, any real resolution has to come from outside the math—whether from prior beliefs, or from Bayesian reasoning. That, the authors suggest, underscores why the debate has continued to go in circles for so long.

The recent study, “Disentangling Boltzmann Brains, the Time-Asymmetry of Memory, and the Second Law,” appeared in the journal Entropy. 

Austin Burgess is a writer and researcher with a background in sales, marketing, and data analytics. He holds an MBA, a Bachelor of Science in Business Administration, and a data analytics certification. His work focuses on breaking scientific developments, with an emphasis on emerging biology, cognitive neuroscience, and archaeological discoveries.