Echoes of Memory

Image credit: Ionut Stefan

Memory is fundamental to our sense of self. It’s what allows us to learn, to adapt, to imagine. But how the brain actually stores memories remains one of neuroscience’s most compelling open questions. In Berlin, dozens of research groups are working together under the umbrella of the SFB1315, a collaborative research center funded by the German Science Foundation focused on understanding the mechanisms of memory consolidation. From flies to birds to rodents to humans, and from molecular pathways to theoretical models, the SFB approaches the problem across species and scales.

As it enters its third funding phase, the center is also launching a new outreach initiative to make this research accessible through a series of interactive podcasts featuring some of the most prominent memory researchers in the world. We sat down with Prof. Matthew Larkum, spokesperson of the SFB1315 and professor of Neuronal Plasticity at Humboldt University, to talk about the science, the collaboration, his own journey into the field, and why engaging the public is a vital part of the project.

“I was always interested in the brain”

To understand the science, it helps to understand the scientist. So we began with a simple question: how did it all start?

“Because I was always interested in the brain, I went to the professor of neurobiology and asked if I could come to his lab after I finished my undergraduate studies and he said ‘Yes, but don’t do biology, do something with mathematics and hard science because you can pick up the biology later.’ He himself was an engineer. So I started at Sydney University in Australia, doing Computer Science.”

“Then I did an Honors in Physiology. And just as I was starting my PhD with this professor, my wife wanted to go and study the violin in Switzerland. So I transferred my PhD to a Swiss laboratory. And that’s how we ended up in Europe. We were initially going for two years and we’ve never gone back.”

It wasn’t the Alps that convinced him to stay though. During his time in Switzerland, Prof. Larkum began recording from networks of neurons, a step that ultimately led him to a postdoc in Germany at the Max Planck Institute in Heidelberg with patch clamp pioneer Bert Sakmann. There, he targeted the dendrites of cortical pyramidal neurons and became interested in how individual neurons transform input into output. It was, as he puts it, a computational question hiding in plain sight.

“The computation of single neurons is extremely, extremely complex because it’s not as in most models of the brain or in artificial intelligence. There you have a very simplified model of a neuron. It’s basically got no dendrites, it just counts the number of inputs and if it reaches a threshold, it gives an output. But it turns out that the dendrites themselves have their own kinds of action potentials and they’re very active and nonlinear.”

“Basically, any input that comes to the top has almost no influence on the bottom. And you would think, ‘Well, why have this architecture then?’”

The interest in memory followed from the realization that dendritic properties had a special influence on memory. Following his postdoc, Prof. Larkum started his own lab in Switzerland, and was eventually offered the Neuronal Plasticity professorship in Berlin. Throughout his career, he continued to investigate the influence of complex neurons on cognition. This led to his now central theory: that memory isn’t just stored in connections between neurons, but is shaped by how information flows within individual neurons themselves. This dendritic integration theory suggests there might be a generative aspect to memory, closer to prediction than to storage.

Cortical pyramidal neurons all share a similar shape: “They all look a bit like an oak tree with a base, with roots, a long, long trunk, and a canopy of branches at the top.” The canopy, as Prof. Larkum calls it, sits at the very top of the cortex, right below the skull, and it’s unusually devoid of cell bodies. And that’s not the only strange thing. “It’s unusual to see a great density of long-range projecting wires in the cortex because most of the connections are nearby. So there’s this strange thing going on that the canopies of all of these neurons are receiving a whole lot of long-range input. And that always struck me way back in Heidelberg in my first postdoc. I was struck by how surprising this is because it means all of this long-range input hits the part of the cell that’s furthest away from the output. In the context of a very long tree, from an electrical perspective, this is huge. Basically, any input that comes to the top has almost no influence on the bottom. And you would think, ‘Well, why have this architecture then?’”

As it turns out, this peculiar architecture may be the key to something powerful. The tree-like structure allows single neurons to separately process two streams of information: external sensory input at the base; and internal signals, predictions or memories, at the top, in the canopy. These internal inputs, many of which come from memory-related regions like the hippocampus, appear to play a crucial role in learning: when they’re blocked, learning itself is disrupted. At the same time, because both streams enter the same neuron, they can also be integrated in a complex, nonlinear way, enabling the neuron to compare incoming sensory data with stored internal models. In Larkum’s view, this dendritic integration is central not only to memory, but to how we learn, adapt, and maintain a stable sense of the world.

“We devised an experiment that was only possible because of the collaborative kind of nature of Berlin”

But much of this work wouldn’t have been possible without the collaborative spirit of Berlin’s research landscape. The city has a long and rich history of memory research, stretching back to the days of Hermann Ebbinghaus and beyond. In that sense, the SFB1315 is both a natural continuation of this tradition and a leap forward. “When the time came to design a collaborative research center, it was, from my point of view, the obvious topic for Berlin to carry on this work on memory research.”

Even before SFB1315 was established, researchers across institutions like the Humboldt University, Charité Medical University, and the TU Berlin were already investigating memory consolidation from different angles. The SFB1315 brought these efforts together, building on an existing foundation of infrastructure, expertise, and shared curiosity, and uniting basic research and theory, across species and methods, under one collaborative framework.

“In computational terms [there’s] a generative aspect to memory consolidation. […] And this fits very much with the architecture of the cortex and these trees that have two regions and the fact that so much really important information that’s internal is going to the top of the tree.”

The project, launched in 2018, is now nearing the end of its second funding period and preparing for its third and final phase, twelve years being the maximum duration for a collaborative research center. Reflecting back on the SFB’s accomplishments, Prof. Larkum says: “The most exciting thing about the seven years so far is that I think we’ve been able to have a big influence on the understanding of the term ‘systems memory consolidation’.”

For decades, the field was shaped by the famous case of Henry Molaison or H.M., a patient whose hippocampus was surgically removed, leaving him unable to form new memories. This led to an entire paradigm focused on understanding memory consolidation as a transfer of information between the hippocampus and the cortex. “In psychology, but also in neuroscience, people are looking for which information is going where.”

“Memory becomes nothing less than the expectations you have on the basis of prior experiences.”

But the SFB has contributed to a shift in that narrative and to a more nuanced view of memory consolidation. “What’s being understood now, not just by our consortium, but around the world, is that there’s a big influence of prior information on the consolidation of [new] information, and that it’s actually a refinement of information. So there’s not so much a transfer of information as a specialized, criterion-dependent stabilization of the very regions that were responsible for the ‘thoughts’ in the first place.”

“I think it’s a topic that the public immediately identifies with on many levels.”

This evolving view of memory, as refinement rather than transfer, has implications far beyond the lab. It reshapes how we think about learning, development, and even identity. But it’s also a story that remains largely unknown to the broader public. And when it comes to public understanding, there’s a clear disconnect. “On the one hand, there’s a natural interest by the public”, Larkum says. Then again, “when you watch a Hollywood movie you get the feeling that one day we will be programming people by changing their memories”. But we’re nowhere near that.

Bridging this gap between scientific work and mainstream knowledge is part of what motivated the SFB’s new outreach initiative, an ambitious public-facing project planned for the final funding period. At its heart is a ten-part podcast series that will bring together leading figures in memory research, many of whom have been part of the SFB’s journey over the past seven years. These conversations will go beyond the lab, addressing memory not just as a biological process, but as something intrinsically human: essential to who we are, vulnerable to disease, and increasingly relevant in an age of digital and artificial memory.

The format is as thoughtful as the concept. Each episode will be recorded as a live event in Berlin, with the public invited not just to listen but to participate by asking questions, engaging with the speakers, and shaping the conversation. The episodes will be made freely available online, ensuring lasting access beyond the live events themselves.

“You could imagine having a public event in a memorial part of the Tiergarten park, talking about the way memories are kept over centuries.”

There’s something quietly poetic about the whole idea. The podcast is envisioned as a kind of homecoming: a return of people, ideas, and insights to Berlin, a city with a rich and complex memory of its own. Some episodes may even be recorded in historically significant locations, grounding the science in the fabric of the city.

Of course, turning this vision into reality won’t be without its challenges. “There’s a natural interest by the public,” according to Prof. Larkum, “but it’s hard to get the information out. Scientists are not very good at this.” That’s why the podcast won’t follow the usual format of researchers talking to each other. While the scientific content will be carefully developed behind the scenes by members of the SFB, the public-facing conversations will be led by a professional podcaster, someone who’s skilled at navigating complex ideas and making them accessible.

There’s also a broader challenge: ensuring that these ideas reach beyond the usual academic or media bubbles. “If you try to manage this such that it only goes to one section of society, then it doesn’t have the influence it ought to have.” It’s a struggle not just of communication, but of participation, and a potential reason, he notes, why the German Research Foundation places such emphasis on outreach. “Everyone sees that a society that understands the relevance of top-level investigation and knowledge is a society with an advantage.”

Still, the goal is clear. “We want this to be a legacy project,” Prof. Larkum emphasizes. “A project that started with the collaborative research center, but continues after its end and really defines Berlin as a place where we started dreaming about what is possible on the topic of memory research.”

---

This article is part of the scientific outreach efforts sponsored by the SFB1315. You can find out more about the project at www.sfb1315.de.

What did you think about this post? Let us know in the comments below. And if you’d like to support our work, feel free to share it with your friends, buy us a coffee here, or even both.

Subscribe to our RSS feed here.

You might also like: