Exploring the Neural Mechanisms of Memory: from Brain Imaging to Single-Unit Recordings in the Hippocampus
Paula Albeda
Dr. Catherine Tallman is a Postdoctoral Scholar and Lecturer at the University of California, San Diego. Using neuroimaging and electrophysiological techniques, she researches the neural mechanisms of long-term memory consolidation in humans.
Her research has been supported by prestigious grants such as the $200,000 Postdoctoral Scholar Support Grant from the Neurtex Brain Research Institute and a Pilot Grant from the Altman Clinical and Translational Research Institute. However, before her doctorate education in Experimental Psychology at the University of California, San Diego in 2023, Dr. Tallman worked at Duke University’s Brain Imaging & Analysis Center as an undergraduate research fellow and later as a clinical research coordinator.
Moreover, her work has been published in major neuroscience journals such as Frontiers in Human Neuroscience and Cognitive Neuroscience, addressing topics like episodic memory, functional connectivity, and memory consolidation. Dedicated to her innovative work, she continues to contribute to the understanding of human memory and its implications. In addition to her research, Dr. Tallman is committed to mentoring the next generations of scientists.
Research & Scientific Contributions
1. Your research explores the fundamental mechanisms of declarative long-term memory. What initially drew you to this area of cognitive neuroscience?
Declarative memories are types of long-term memories that we can express consciously. They include facts and events we know and can share with others. For example, we all know 2 + 2 = 4 and we can communicate that—it’s a simple fact we can easily express. These memories can also be events in our lives tied to a specific time and place, like remembering your 5th birthday party or the day you adopted your family dog.
Memory is so integral to our everyday lives, and I think the ability to express memories is a big part of what makes us "human." They are our past experiences that shape how we think and act now in the present and in the future. I'm fascinated by how our brains manage this monumental feat. I'm interested in fundamentally understanding how our brains “capture” memories, make them concrete and lasting, and remember them even long after we first learned them.
2. Your graduate work focused on the functional brain networks involved in memory consolidation using fMRI. How has this background influenced your current research on single-unit recordings in the hippocampus?
My PhD dissertation focused on how brain regions communicate to support memory consolidation, which is how memories are stabilized over time. When we first “capture” or encode a memory it’s very, very fragile; the brain needs a way to make it last for long periods of time. I used functional MRI (fMRI) to examine how various parts of the brain work together to consolidate different types of episodic memories over time. We know that declarative long-term memories initially rely on the hippocampus for consolidation, and over time, its connections with the cortex and connections within the cortex change to stabilize these memories. Different theories suggest that the hippocampus and cortex interact differently depending on the type of memory. I was particularly interested in examining how these connections evolve for the different types of declarative memories and across different time scales, across months and years or even decades.
While fMRI provides a ‘big picture’ measure of brain activity by capturing the firing of millions of neurons in different brain areas, it does not isolate the activity of a single neuron. In my postdoctoral research, I’m looking at how unique episodic memories—memories with context—are formed by single neurons in the hippocampus. My main interest is in how the activity of individual neurons in the hippocampus represents a unique memory.
My background in memory consolidation influences the way I think about brain activity versus brain connectivity. Measuring brain activity in one brain region or neuron is not the same thing as measuring the connectivity of that region or neuron. Right now, I’m concentrating on the activity of individual neurons rather than connectivity, but I recognize that the signals from individual neurons in the hippocampus during memory encoding eventually lead to changes in connectivity between the hippocampus and cortex as memory consolidation occurs over time.
3. Much of your work examines sparse coding in the hippocampus. What are the key questions you aim to answer about how episodic memories are represented at the neuronal level?
As both a graduate student and now as a postdoc, my academic advisors taught me to conduct cognitive neuroscience research that was grounded in established theories and models. In my postdoc, I’m investigating whether a particular neurocomputational model accurately reflects how our hippocampus “represents” a single episodic memory. A sparse coding scheme in the hippocampus predicts that episodic memories have distinct patterns that help keep them separate from each other, allowing us to stabilize and remember them instead of having overlapping patterns that interfere with one another.
My current work builds on past research which used an unintuitive statistical method to investigate if there was evidence of sparse coding in patterns of single neurons in the hippocampus. My first interest was to replicate this finding in an independent dataset collected by a different research lab. My second interest was whether this pattern for an individual episodic memory is only present for memories that are remembered, rather than those that are forgotten. Additionally, I explored a theory called neuronal allocation, which suggests that neurons with high intrinsic excitability at learning tend to encode a specific memory. Our hypothesis is that excitable neurons in the hippocampus are the ones assigned to codes for specific items such as pictures or words. If a neuron has decreasing firing or consistently high or low firing rates at learning, it may not participate in sparse coding.
4. Some of your recent studies investigate memory reinstatement and the role of hippocampal activity in recall. What do your findings suggest about how memories are reactivated over time?
We know that memory is constructive; when you remember something, your brain tries to recreate the neural activity patterns from when you originally learned it. While there are various ways to examine this reconstruction, we tried to examine if there was general evidence for this reinstatement at the level of single neurons. In the single-unit recording data, we found statistical evidence of sparse coding specifically for pictures linked to excitable neurons at learning and furthermore, the evidence was only for pictures that were actually remembered. An interpretation of this statistical evidence could be that there are two factors associated with a sparse coding scheme: 1) sparse codes in the hippocampus contribute to remembering and 2) neurons that were particularly excitable during learning are allocated to a particular memory. The general findings from our analyses suggest there could be reactivation of the sparse code at retrieval for particular items that were allocated to excitable neurons at learning.
5. You’ve worked with both fMRI and intracranial recordings in epilepsy patients. How do these methods complement each other in understanding memory processes?
fMRI is a fantastic tool for examining large-scale brain networks, allowing us to see how millions of neurons and individual brain regions fire together. It measures brain activity indirectly by tracking changes in blood oxygenation: when a brain area is used, more blood flows to that area, leading to changes in oxygen levels in the blood. However, this signal is slow; you typically detect changes about six seconds after the brain becomes active, making it a delayed response. One advantage of fMRI though is that it’s relatively easy to acquire images. This means we can include a wide range of individuals in studies.
In contrast, single-unit recording is very rare because it is difficult to collect. This method often involves patients with epilepsy who consent to have electrodes placed in their brains during surgery to remove diseased tissue, typically in the hippocampus. These patients might stay in the hospital for days or weeks with electrodes directly monitoring their brain activity. Some graciously volunteer to participate in memory experiments during this time.
The major advantage of single-unit recording is that it provides a direct and instantaneous measure of neuron activity, even at the level of a single neuron. This contrasts sharply with the indirect and slow fMRI approach. This shows how different brain recording techniques can differ in different temporal and spatial resolutions and by measuring different biological signals. Researchers can select the appropriate recording technique that is tailored to their research question.
6. Memory consolidation involves interactions between distributed brain networks. How do you see your research contributing to a broader understanding of systems-level consolidation theories?
The theory of systems-level consolidation originated from observations of patients who experienced neuropsychological deficits after losing hippocampal function. These patients exhibited temporarily-graded retrograde amnesia, meaning they couldn’t remember recent declarative memories but could remember older memories. This suggests that while the hippocampus is crucial for recent memories, it may not be necessary for distant ones.
Patients with hippocampal damage are rare, thankfully! With the invention of fMRI, we can now test this theory in non-patient populations. fMRI allows us to examine not only the activity of the hippocampus over time but functional connections within the cortex. This enables a broader investigation of systems consolidation across larger groups since it’s relatively easy to scan someone and these types of experiments could be run on anyone, not just those with hippocampal damage.
There are competing theories, such as multiple trace theory and transformation theory, which propose that the hippocampus is always required for both recent and remote declarative memories. Functional MRI is well-suited to examine these theories, giving us the opportunity to determine their validity and see how they may evolve from the insights gained from larger datasets of people without hippocampal damage
7. Your work has implications for understanding memory impairments in neurological conditions. What insights might your research offer for disorders such as Alzheimer’s disease?
Some of my research that focused on large-scale functional brain networks associated with memory consolidation was trying to identify the retrograde memory network. Retrograde memory is memory for the past. If we can characterize how the retrograde memory network changes over time in people with normal cognition, we can see how that network might differ in patients with Alzheimer’s Disease. The retrograde memory network is one focus of the research of my graduate advisor, Dr. Smith. In Alzheimer’s disease, damage first starts in the hippocampus and spreads. One of her lines of research was to see if impairments in the retrograde memory network, which could manifest as consolidation deficits, could help diagnose Alzheimer’s in its pre-clinical stages by developing a clinical test that taps into this network.
Other Questions
8. If you had unlimited resources to investigate a memory-related phenomenon, what would you focus on and why?
If I had unlimited resources, I'd run this crazy experiment that surely no one would want to participate in. I’d love to explore how connections change between the hippocampus and cortex over time to consolidate single memory. If we’re going with a vaguely realistic approach, first I’d have to hire a team of MR physicists to develop the world’s best MRI machine with super high temporal and spatial resolution. Then I would scan a child shortly after an event like their 5th birthday party. We'd capture their brain activity as they remember that memory over and over again. Maybe every six months, or something like that, until they turn 80. While it sounds a bit ridiculous it would provide fascinating insights into how the connections between the hippocampus and cortex evolve over time for a specific episodic memory. Then ideally you’d want to collect lots of participants, let’s say 1000 participants. If we’re dreaming big, I’d find someone to invent a non-invasive way to measure single neuron activity and then re-do it all.
9. Do you think advances in AI and computational modeling will help us decode and manipulate episodic memories in the near future?
I absolutely think that advances in AI and computational modeling will greatly enhance our understanding of episodic memory. There's already been great progress in this area. One that I’m more familiar with is a technique called multi-voxel pattern analysis which uses machine learning to classify patterns of fMRI activity. In simpler terms, this method uses machine learning to identify specific patterns of brain activity and associate those patterns with individual memories or categories of memories. For example, we could decode from fMRI brain activity whether someone is looking at a picture of a dog or a cat with the right data.
As far as episodic memory manipulation, virtual reality could be an interesting way to have precise control over the memory-formation process. In combination with brain recording techniques, like EEG, which measures brain activity more easily than fMRI, it would be cool to start to manipulate episodic memories—maybe by introducing types of information to unknowingly change memories. Who knows! It’s an exciting field with research questions and applications I can’t even fully envision yet.
Personal Journey & Perspective
10. What advice would you give to young scientists interested in studying the neural mechanisms of memory?
I have two main pieces of advice. The first is more technical: cognitive neuroscience involves some fascinating brain recording tools and software, but it can also be quite coding and statistics intensive. When I first started, I didn’t realize how important programming would be. I recommend getting familiar with languages such as Python and R as they'll really help you hit the ground running if you get into this research field.
The second piece of advice is a bit more abstract, some might even call it “woo-woo!” I encourage any cognitive neuroscientist to be curious about the world by seeking out diverse life experiences and meeting different people. The more you engage with life, the better equipped you'll be to ask broad questions about how things work, like memory. The fun part is figuring out how to translate those human experiences into something you can test in a lab.
So, to sum it up: stay curious and have fun!
About the Author
Paula Albeda ('26) is a senior at The University of California, San Diego concentrating in psychology.
Her research has been supported by prestigious grants such as the $200,000 Postdoctoral Scholar Support Grant from the Neurtex Brain Research Institute and a Pilot Grant from the Altman Clinical and Translational Research Institute. However, before her doctorate education in Experimental Psychology at the University of California, San Diego in 2023, Dr. Tallman worked at Duke University’s Brain Imaging & Analysis Center as an undergraduate research fellow and later as a clinical research coordinator.
Moreover, her work has been published in major neuroscience journals such as Frontiers in Human Neuroscience and Cognitive Neuroscience, addressing topics like episodic memory, functional connectivity, and memory consolidation. Dedicated to her innovative work, she continues to contribute to the understanding of human memory and its implications. In addition to her research, Dr. Tallman is committed to mentoring the next generations of scientists.
Research & Scientific Contributions
1. Your research explores the fundamental mechanisms of declarative long-term memory. What initially drew you to this area of cognitive neuroscience?
Declarative memories are types of long-term memories that we can express consciously. They include facts and events we know and can share with others. For example, we all know 2 + 2 = 4 and we can communicate that—it’s a simple fact we can easily express. These memories can also be events in our lives tied to a specific time and place, like remembering your 5th birthday party or the day you adopted your family dog.
Memory is so integral to our everyday lives, and I think the ability to express memories is a big part of what makes us "human." They are our past experiences that shape how we think and act now in the present and in the future. I'm fascinated by how our brains manage this monumental feat. I'm interested in fundamentally understanding how our brains “capture” memories, make them concrete and lasting, and remember them even long after we first learned them.
2. Your graduate work focused on the functional brain networks involved in memory consolidation using fMRI. How has this background influenced your current research on single-unit recordings in the hippocampus?
My PhD dissertation focused on how brain regions communicate to support memory consolidation, which is how memories are stabilized over time. When we first “capture” or encode a memory it’s very, very fragile; the brain needs a way to make it last for long periods of time. I used functional MRI (fMRI) to examine how various parts of the brain work together to consolidate different types of episodic memories over time. We know that declarative long-term memories initially rely on the hippocampus for consolidation, and over time, its connections with the cortex and connections within the cortex change to stabilize these memories. Different theories suggest that the hippocampus and cortex interact differently depending on the type of memory. I was particularly interested in examining how these connections evolve for the different types of declarative memories and across different time scales, across months and years or even decades.
While fMRI provides a ‘big picture’ measure of brain activity by capturing the firing of millions of neurons in different brain areas, it does not isolate the activity of a single neuron. In my postdoctoral research, I’m looking at how unique episodic memories—memories with context—are formed by single neurons in the hippocampus. My main interest is in how the activity of individual neurons in the hippocampus represents a unique memory.
My background in memory consolidation influences the way I think about brain activity versus brain connectivity. Measuring brain activity in one brain region or neuron is not the same thing as measuring the connectivity of that region or neuron. Right now, I’m concentrating on the activity of individual neurons rather than connectivity, but I recognize that the signals from individual neurons in the hippocampus during memory encoding eventually lead to changes in connectivity between the hippocampus and cortex as memory consolidation occurs over time.
3. Much of your work examines sparse coding in the hippocampus. What are the key questions you aim to answer about how episodic memories are represented at the neuronal level?
As both a graduate student and now as a postdoc, my academic advisors taught me to conduct cognitive neuroscience research that was grounded in established theories and models. In my postdoc, I’m investigating whether a particular neurocomputational model accurately reflects how our hippocampus “represents” a single episodic memory. A sparse coding scheme in the hippocampus predicts that episodic memories have distinct patterns that help keep them separate from each other, allowing us to stabilize and remember them instead of having overlapping patterns that interfere with one another.
My current work builds on past research which used an unintuitive statistical method to investigate if there was evidence of sparse coding in patterns of single neurons in the hippocampus. My first interest was to replicate this finding in an independent dataset collected by a different research lab. My second interest was whether this pattern for an individual episodic memory is only present for memories that are remembered, rather than those that are forgotten. Additionally, I explored a theory called neuronal allocation, which suggests that neurons with high intrinsic excitability at learning tend to encode a specific memory. Our hypothesis is that excitable neurons in the hippocampus are the ones assigned to codes for specific items such as pictures or words. If a neuron has decreasing firing or consistently high or low firing rates at learning, it may not participate in sparse coding.
4. Some of your recent studies investigate memory reinstatement and the role of hippocampal activity in recall. What do your findings suggest about how memories are reactivated over time?
We know that memory is constructive; when you remember something, your brain tries to recreate the neural activity patterns from when you originally learned it. While there are various ways to examine this reconstruction, we tried to examine if there was general evidence for this reinstatement at the level of single neurons. In the single-unit recording data, we found statistical evidence of sparse coding specifically for pictures linked to excitable neurons at learning and furthermore, the evidence was only for pictures that were actually remembered. An interpretation of this statistical evidence could be that there are two factors associated with a sparse coding scheme: 1) sparse codes in the hippocampus contribute to remembering and 2) neurons that were particularly excitable during learning are allocated to a particular memory. The general findings from our analyses suggest there could be reactivation of the sparse code at retrieval for particular items that were allocated to excitable neurons at learning.
5. You’ve worked with both fMRI and intracranial recordings in epilepsy patients. How do these methods complement each other in understanding memory processes?
fMRI is a fantastic tool for examining large-scale brain networks, allowing us to see how millions of neurons and individual brain regions fire together. It measures brain activity indirectly by tracking changes in blood oxygenation: when a brain area is used, more blood flows to that area, leading to changes in oxygen levels in the blood. However, this signal is slow; you typically detect changes about six seconds after the brain becomes active, making it a delayed response. One advantage of fMRI though is that it’s relatively easy to acquire images. This means we can include a wide range of individuals in studies.
In contrast, single-unit recording is very rare because it is difficult to collect. This method often involves patients with epilepsy who consent to have electrodes placed in their brains during surgery to remove diseased tissue, typically in the hippocampus. These patients might stay in the hospital for days or weeks with electrodes directly monitoring their brain activity. Some graciously volunteer to participate in memory experiments during this time.
The major advantage of single-unit recording is that it provides a direct and instantaneous measure of neuron activity, even at the level of a single neuron. This contrasts sharply with the indirect and slow fMRI approach. This shows how different brain recording techniques can differ in different temporal and spatial resolutions and by measuring different biological signals. Researchers can select the appropriate recording technique that is tailored to their research question.
6. Memory consolidation involves interactions between distributed brain networks. How do you see your research contributing to a broader understanding of systems-level consolidation theories?
The theory of systems-level consolidation originated from observations of patients who experienced neuropsychological deficits after losing hippocampal function. These patients exhibited temporarily-graded retrograde amnesia, meaning they couldn’t remember recent declarative memories but could remember older memories. This suggests that while the hippocampus is crucial for recent memories, it may not be necessary for distant ones.
Patients with hippocampal damage are rare, thankfully! With the invention of fMRI, we can now test this theory in non-patient populations. fMRI allows us to examine not only the activity of the hippocampus over time but functional connections within the cortex. This enables a broader investigation of systems consolidation across larger groups since it’s relatively easy to scan someone and these types of experiments could be run on anyone, not just those with hippocampal damage.
There are competing theories, such as multiple trace theory and transformation theory, which propose that the hippocampus is always required for both recent and remote declarative memories. Functional MRI is well-suited to examine these theories, giving us the opportunity to determine their validity and see how they may evolve from the insights gained from larger datasets of people without hippocampal damage
7. Your work has implications for understanding memory impairments in neurological conditions. What insights might your research offer for disorders such as Alzheimer’s disease?
Some of my research that focused on large-scale functional brain networks associated with memory consolidation was trying to identify the retrograde memory network. Retrograde memory is memory for the past. If we can characterize how the retrograde memory network changes over time in people with normal cognition, we can see how that network might differ in patients with Alzheimer’s Disease. The retrograde memory network is one focus of the research of my graduate advisor, Dr. Smith. In Alzheimer’s disease, damage first starts in the hippocampus and spreads. One of her lines of research was to see if impairments in the retrograde memory network, which could manifest as consolidation deficits, could help diagnose Alzheimer’s in its pre-clinical stages by developing a clinical test that taps into this network.
Other Questions
8. If you had unlimited resources to investigate a memory-related phenomenon, what would you focus on and why?
If I had unlimited resources, I'd run this crazy experiment that surely no one would want to participate in. I’d love to explore how connections change between the hippocampus and cortex over time to consolidate single memory. If we’re going with a vaguely realistic approach, first I’d have to hire a team of MR physicists to develop the world’s best MRI machine with super high temporal and spatial resolution. Then I would scan a child shortly after an event like their 5th birthday party. We'd capture their brain activity as they remember that memory over and over again. Maybe every six months, or something like that, until they turn 80. While it sounds a bit ridiculous it would provide fascinating insights into how the connections between the hippocampus and cortex evolve over time for a specific episodic memory. Then ideally you’d want to collect lots of participants, let’s say 1000 participants. If we’re dreaming big, I’d find someone to invent a non-invasive way to measure single neuron activity and then re-do it all.
9. Do you think advances in AI and computational modeling will help us decode and manipulate episodic memories in the near future?
I absolutely think that advances in AI and computational modeling will greatly enhance our understanding of episodic memory. There's already been great progress in this area. One that I’m more familiar with is a technique called multi-voxel pattern analysis which uses machine learning to classify patterns of fMRI activity. In simpler terms, this method uses machine learning to identify specific patterns of brain activity and associate those patterns with individual memories or categories of memories. For example, we could decode from fMRI brain activity whether someone is looking at a picture of a dog or a cat with the right data.
As far as episodic memory manipulation, virtual reality could be an interesting way to have precise control over the memory-formation process. In combination with brain recording techniques, like EEG, which measures brain activity more easily than fMRI, it would be cool to start to manipulate episodic memories—maybe by introducing types of information to unknowingly change memories. Who knows! It’s an exciting field with research questions and applications I can’t even fully envision yet.
Personal Journey & Perspective
10. What advice would you give to young scientists interested in studying the neural mechanisms of memory?
I have two main pieces of advice. The first is more technical: cognitive neuroscience involves some fascinating brain recording tools and software, but it can also be quite coding and statistics intensive. When I first started, I didn’t realize how important programming would be. I recommend getting familiar with languages such as Python and R as they'll really help you hit the ground running if you get into this research field.
The second piece of advice is a bit more abstract, some might even call it “woo-woo!” I encourage any cognitive neuroscientist to be curious about the world by seeking out diverse life experiences and meeting different people. The more you engage with life, the better equipped you'll be to ask broad questions about how things work, like memory. The fun part is figuring out how to translate those human experiences into something you can test in a lab.
So, to sum it up: stay curious and have fun!
About the Author
Paula Albeda ('26) is a senior at The University of California, San Diego concentrating in psychology.