Category Archives: memory

On Finding Yourself

The hippocampus and associated structures such as the entorhinal cortex, have long been known to play an extremely important role in navigation and memory formation (as previously discussed in this forum: 1, 2). For example, the hippocampus is enlarged in London taxicab drivers, who presumably employ it heavily for navigating around the city, and individuals with damage to this area are unable to form new memories at all, though they can recall past experiences with no loss of fidelity.

The entorhinal cortex feeds into the hippocampus, and it seems to be far more specialized for navigation purposes. There are cells in this area that seem to encode the direction that an animal’s head is pointing. There is another varietal, referred to as the grid-cell, whose response-properties are illustrated below.

from reference 2

Grid-cells have the intriguing property of responding vigorously in a regular array of spatial locations. If an experimenter puts an animal in a small confined space, the regularity of these responses are evident after a brief period of exploration by the animal. On the left, you see the black trace of a rat’s position as it wanders around this enclosure, with red traces representing locations where the response of a single neuron under consideration was strongest. In the middle, you can see a rasterized representation of this information, and on the right, a “cross-correlation” of the middle plot, showing the regularity of the responses.

Much of the pioneering work on the hippocampus and entorhinal cortex has come from the lab of Edvard and May-Britt Moser, a married pair of neuroscientists working at the Kavli Institute for Systems Neuroscience at the Norwegian University of Science and Technology. They have recently discovered another, infrequently occurring, class of cell in the entorhinal cortex, the border cell. Predicted from theoretical considerations in the year 2000 by Neil Burgess, these cells respond preferentially when the animal is placed near a border, of a certain orientation, of an environment (these can be walls, or drop-offs; see image, below).

from reference 1

Panel A shows that the response is independent of the size of the space, and that if the space is expanded, so too does the area over which the neuron responds (right). Panel B shows that if a new separation is inserted into the space, the cell begins to respond strongly to the boundary. Panel C shows that the response properties persist even after the walls are removed, leaving only a drop-off). Panel D shows that the orientation specific quality of the responses are independent of the shape of the room. If a landmark, in the form of a marker on one of the walls, is employed, the responses rotate along with the landmark (see below). Which is impressive, but also expected and necessary for these cells to function efficiently as part of a navigational system.

It is still a mystery how these cells produce these responses. It must be the case that a computational transformation of a variety of sensory and motor information must contribute to the computation of border location. These cells represent some of the few that have such clearly defined properties. That is to say, in many other parts of the brain, single neurons contribute only a small part of the over-all response to a stimulus, and it is thus surprising to find single cells devoted to the entire border of a space. Another impressive and enigmatic feature of the responses of these cells is the fact that they must rapidly re-compute and respond to the borders of a novel place. This sort of short-time-scale neuronal plasticity is of great interest to neuroscientists everywhere.

Understanding the properties of these cells, and their role in navigation and memory formation will be a great and rewarding challenge, one that I’m sure the Mosers are up to.

1. Solstad T, Boccara CN, Kropff E, Moser MB, Moser EI. Representation of geometric borders in the entorhinal cortex. Science 322: 1865-1868, 2008.
2. Hafting T, Fyhn M, Molden S, Moser MB, Moser EI. Microstructure of a spatial map in the entorhinal cortex. Nature 436: 801-806, 2005.

On Memory and Experience

from reference 1

As reported in the New York Times, a new study has demonstrated an aspect of memory that has long been hypothesized. That being: the same neurons that fire during an experience fire during the memory of that experience. The research, published in the journal Science, relies on recordings from the brains of epileptic patients undergoing surgery to remove the parts of their brain which cause excesses of neuronal activity, essentially the only way to record the activity of neurons in awake human beings1.

The authors of the study took an approach where they recorded the activity of single hippocampal brain cells while subjects were watching a variety of video clips. Unsurprisingly, certain cells responded best to certain clips. Then, after a brief interim during which the experimenters distracted the patients, they asked the subjects to recall the video clips. Not only did the activity of the neurons during recollection correlate with activity during first viewing, the experimenters were able to predict which video clip the subjects were remembering based on the recorded activity! Interestingly, however, the hippocampus (the area of the brain being recorded from in this study) is not required for the recall of long term memories. Thus, in some ways this work further deepens the mystery of how short term versus long term memories are encoded in the brain and the involvement of hippocampus in these processes; a subject previously touched on in this forum.

Reading about this research reminded me of my favorite definition of memory, as the ability to:

“repeat a mental or physical act after some time despite a changing context…. We stress repetition after some time in this definition because it is the ability to re-create an act separated by a certain duration from the original signal set that is characteristic of memory. And in mentioning a changing context, we pay heed to a key property of memory in the brain: that it is, in some sense, a form of constructive recategorization during ongoing experience, rather than a precise replication of a previous sequence of events.

…the key conclusion is that whatever its form, memory itself is a [property of a system]. It cannot be equated exclusively with circuitry, with synaptic changes, with biochemistry, with value constraints, or with behavioral dynamics. Instead, it is the dynamic result of the interactions of all these factors acting together, serving to select an output that repeats a performance or an act.

The overall characteristics of a particular performance may be similar to previous performance, but the ensembles of neurons underlying any two similar performances at different times can be and usually are different. This property ensures that one can repeat the same act, despite remarkable changes in background and context, with ongoing experience.”2


1. Gelbard-Sagiv H, Mukamel R, Harel M, Malach R, Fried I (2008) Internally Generated Reactivation of Single Neurons in Human Hippocampus During Free Recall. Science 10.1126/science.1164685
2. Edelman GM, Tononi G (2000) A Universe of Consciousness: How Matter Becomes Imagination, Basic Books, New York

On Visual Working Memory

Suppose you’re preparing dinner and you realize that you’d like to add some more tomatoes, onions, and mushrooms to the salad that you’re making. You have no trouble storing images of these items in your head and locating them in the fridge. Suppose instead that you’re eating dinner and you decide that you absolutely must have another bottle of red wine for your guests, but you really want the cote du rhone, and not the boring merlot. Again, you can quite easily go and grab the bottle based simply on its label, but in this latter situation, the amount of visual information: the detail that must be stored for comparison upon reaching your liquor cabinet is far greater.

In both cases, you’re employing some form of working visual memory. This form of memory is thought to be highly plastic, short-term storage. It is possible however, that it might have different characteristics. The forms that working memory might take are (1) a set of fixed “slots,” each having a discreet capacity or (2) a set of dynamic slots which can be more tailored to the specific use.

The fixed-slots concept is similar to the pictures taken by current digital cameras. The images are always the same number of (mega)pixels, no more, no less. While the dynamic-slots concept is more akin to being able to vary the number of pixels in an image depending on demand. For instance, if I was taking a picture of a pure red wall, I wouldn’t need any more than 1 pixel because every part of the image would be identical. However, if you were taking a picture of a Jackson Pollack painting, you might want to combine several digital camera images to get a really accurate rendering of the details held therein, a more flexible allocation of working-memory resources.

The scenarios I’ve highlighted above don’t really serve to answer the question of which form of working memory our brains employ because in both cases, we can imagine either form working just fine (as long as we accept the notion that the complexity of the wine bottle label is within the capacity limits of the fixed-slots). However, a recent paper published in Nature claims to have employed a savvy enough experimental approach to disentangle the two possibilities.

The research, published by Steven J. Luck (who has done quite a bit of excellent work in the field of Visual Neuroscience in general) and a colleague, Weiwei Zhang, is sadly quite brief, having clearly been edited for length by the editors at Nature. In fact, this curtailed version is quite difficult to follow, given the subtlety of the approach that the authors used, but their essential point comes through.

The approach they take is a common one. They measure human performance on a variety of working-memory tasks and attempt to fit these data to models with different assumptions. In this paper, they compare how well the data can be fit to a fixed-slots model versus a dynamic-slots model. Although they conclude that the dynamic-slots model simply doesn’t explain the data as well, and they thus discard the notion of flexible resource allocation, one of the final sentences betrays what they must admit: “This model does not completely eliminate the concept of resources, because the slots themselves are a type of resource.” In other words, it is possible to allocate multiple slots to the same item object-to-be-held-in-memory. However, it does appear that the slots themselves are of a fixed size.

This result places limits on the possible anatomical underpinnings of working-memory, and makes predictions about how one might expect human beings to perform in other, working-memory tasks. It will be interesting to see if the conclusions that these authors reached will be borne out in future work.

On Hippocampal Memory

The Hippocampus

Lydia Kibiuk, copyright © 2002 Lydia Kibiuk

The hippocampus is the area of the mammalian brain most closely associated with memory, particularly spatial memory, meaning internal maps. It is common for individuals with hippocampal lesions (and surrounding related regions) to have amnesia, as in the famous case of H.M., and more recently E.P. Also, it has been shown that London cab-drivers, who presumably require large internal maps for navigation, have enlarged hippocampi1. However, it is has become clear that the hippocampus isn’t required for all forms of memory. A set of looming questions, then, is: where memories are stored, how are they formed, and what role various structures in the brain play in these activities.

I attended a seminar (3/26/08) given by Larry Squire, who has been studying the role of the hippocampus in memory formation, retrieval and storage for quite some time. He summarized results comparing normal patients to those with hippocampal lesions. In general, it seems as though this structure is required for the formation of certain types of new memories only, not for recall or storage, since the lesioned patients had no trouble remembering thigns from their past (including how to navigate their childhood neighborhoods). However, the truly fascinating result he presented was from an experiment designed to reveal what role the hippocampus plays in building new memories.

Performance Graphs

(from reference 2)

The figure above summarizes the data gathered from normal and hippocampal-lesion patients during a memory task. The task was quite straightforward, subjects were presented with 8 pairs of objects, one pair at a time, in which one of the pair was always “correct.” In a given trial, the subject was presented with a pair, and chose one by reaching out and picking it up. On the underside of each object was a sticker that either read “correct,” or “incorrect.” As represented in the left panel of the figure above, normal humans were able to reach perfect performance in this task by repeated presentation of these pairs3. In addition, these subjects were easily able to cope with a variation on the task. All 16 objects were presented at once, and the subject was asked to separate out the correct from the incorrect items (indicated by the grey bar to the right of the trace).

The hippocampal-lesion patients, however, showed dramatically different results. They required 12 times as much training, didn’t get up to the same level of performance as the normal subjects, and were unable to perform the task-variant. This is to say nothing of the fact that they didn’t recall having ever attempting the task before when queried on the 2nd through 36th sessions.

It is fascinating that these patients were able to learn this task at all, and it is clear that they are using some completely different strategy from the normal subjects (one which relies heavily on the paired-object context, as revealed by their confusion at being presented with all the objects at once). There are several questions that are raised by this research. First, if there is an alternative pathway for learning such tasks, how does the brain choose to use the hippocampal path to record a particular type of memory? One might suggest that the brain uses the hippocampus for all memories unless it isn’t useable as in these patients, but we know that many types of motor learning, learning to play the piano for instance, do not require the hippocampus. A further set of interrelated questions are where and how the memory is being stored, and how these differ from those patients with intact hippocampi.

This type of research shows us definitively that we have the capacity for different types of memory, and that we have a ways to go in understanding how and where it operates.

Notes & References:

1. Maguire EA, Frackowiak RS, Frith CD. (1997) Recalling routes around london: activation of the right hippocampus in taxi drivers. J Neurosci; 17(18):7103-10.
2. Bayley PJ, Frascino JC, Squire LR. (2005) Robust habit learning in the absence of awareness and independent of the medial temporal lobe. Nature; 436(7050):550-3.
3. A “session” in this experiment consisted of 40 trials, 5 presentations of each of the 8 pairs of objects.

On Believing Yourself

I have always been quite troubled by the fact that I can remember things that never happened. If I am confident that a childhood friend’s name was Paul when it was actually Roger, how am I to be certain that I correctly remember how to perform the act of addition, or my distaste for the texture of most mushrooms?

Perhaps even more troubling is the fact that studies devoted to exploring the interplay between confidence and memory have found that, in general, the memories we’re most confident in are most likely to be authentic1 (see figure, below).

The paradox is fairly clear: how can we be confident in a false memory, if confidence correlates with accuracy?

The authors of a recent study suggest, and go a ways towards demonstrating, that two distinct mechanisms are at work, one at work when we express confidence in veridical memories, and one for when we express confidence in false recollections2.

Specifically, these authors use fMRI, and a categorized word recall task, to demonstrate that distinct brain areas are active when we’re sure of veracious retrospection and another when we’re confident in specious anamnesis. The researchers speculate that the latter is due to the familiarity of certain events based on the anatomy of the active sites revealed by the scan (see figure, below).

As a final note, the two areas identified in this study are quite far apart in brain terms, once again pointing to the notion that memory is physiologically and anatomically diffuse. So when you can’t remember your first pet’s name, don’t get too worried, your brain is a big place to search.


1. Lindsay DS, Read JD, Sharma K (1998) Accuracy and confidence in person identification: the relationship is strong when witnessing conditions vary widely. Psychol Sci 9:215–218.
2. Kim H, & Cabeza R (2007) Trusting Our Memories: Dissociating the Neural Correlates of Confidence in Veridical versus Illusory Memories. J. Neurosci 27(45):12190-12197

On Emotion and Memory

from reference 1

Why is it that experiences imbued with emotion crystalize into easily recollected memories? Our memory is quite limited, so we need a system for deciding what to remember and what to forget. Emotions may thus act as a filter, marking certain experiences as being of particular importance. In this way, we have templates of states we felt were positive or negative, examples of the consequences of our behaviors, with significantly happy or sad outcomes featuring as the most poignant reminders.

None of this gestalt psychological explanation is informative as to the neurophysiological mechanisms underlying this phenomenon. However, some recent research does address what mechanisms may be at work on a molecular level. Joseph Ledoux and Robert Malinow have been working on memory for a quite a while, and they are the two most senior authors on a paper published in Cell concerning AMPA receptors, emotion, and memory (ref. 1). AMPA receptors are one of the major glutaminergic receptors in the brain. Glutamate is the neurotransmitter they recognize, and it is the major excitatory neurotransmitter in the brain. So if one neuron wants to send a signal to turn on another, it will almost invariably release glutamate at it’s axon terminal, and that glutamte will likely be recognized by an AMPA receptor on the post-synaptic cell (the target of the excitation). The major finding of this paper is that norepinephrine (more commonly known as adrenaline) facilitates the incorporation of AMPA receptors into the membranes of cells, during periods of high activity.

It is commonly known that adrenaline is released during times of emotional distress and happiness, these researchers have found that one specific effect of the adrenaline is to increase the number of receptors being incorporated into a synapse, again during periods of high activity. Let’s imagine a scenario where this might apply. An animal is being chased by a predator. His motor planning and execution areas are blasting away action potentials, they’re highly activated. He makes a decision about some route to take during his escape, activating a specific subset of pathways. It is these connections that will be strengthened by the application of adrenaline. Because more receptors are being integrated into the synapses in these circuits, they will be more likely to be activated the next time he is in the same situation. In this sense, he has formed a memory of the experience which is modulated by the amount of adrenaline, and by extension the intensity of the emotion experienced.

This is essentially what these researchers observed. While it is impossible to directly modulate the emotional state of the animal, they can apply norepinephrine during a learning task. What they found was that animals who received larger doses of applied norepinephrine were more likely to remember the task. The figure at the top of this piece illustrates the finding. The authors compared genetically altered (GA) mice – which lack the effects of adrenaline on AMPA receptor trafficking – to “normal” or wild-type (WT) mice. The graph on the left displays the responses of the WT mice, with the GA mice on the right. The key is that two data points are significantly different (marked with an asterisk) on the left, but not on the right.

While this work doesn’t do much to help understand the biosychological basis of Proust, it does illuminate one more minuscule thread in the web of conscious experience.


1. Hailan Hu, Eleonore Rea, Kogo Takamiya, Myoung-Goo Kang, Joseph Ledoux, Richard L. Huganir and Roberto Malinow, (2007) Emotion Enhances Learning via Norepinephrine Regulation of AMPA-Receptor Trafficking, Cell 131,1, pp 160-173 [doi:10.1016/j.cell.2007.09.017]