Memory encompasses everything from thoughts of childhood friends to a mental list of what we need to pick up at the grocery store. It is essential for our sense of self, and allows us to learn from our previous experiences. In general, a memory is a piece of information stored in your brain, but the quality of this information and the length of storage time vary greatly. How memories are formed, and what causes us to forget, have long been topics of great interest in the field of neuroscience.

The brain is the most complex human organ. It is made up of millions of cells called neurons that are interconnected in a vast network. Cells in certain regions of the brain perform specialized functions. For instance, one particular area of the brain is important for vision and another for movement. Functions of many brain areas have been worked out through extensive study of people who have suffered brain damage in addition to studies with model organisms, such as mice.

It is believed that long-term memories are stored in different areas across the brain, depending on the contents of the information. A single memory can even be partitioned to multiple brain regions. For example, the visual trace of a memory is stored in the area of the brain involved in sight perception. If part of this visualization region is damaged, our memories of what we see may be affected as well. For example, if the color processing part of the brain is damaged, a person recalls previous experiences in black and white [1]. This is a logical way to store information, as it allows the brain to gain quick access to past information when it needs to be integrated with incoming sensory perceptions. Navigation is a clear example of this. Certain visual cues in the environment (e.g. a blue post box, an old gnarled tree) may trigger a memory of past times you have travelled the same route, helping you decide where to make the next turn.

Decoding how the brain stores memories is a tricky business. When neurons are activated, their electrical charge changes briefly. If strong enough, this change in electrical charge can trigger the neuron to release chemicals that signal to connected neurons. Information can be encoded in this network of neurons in multiple ways. The particular signal received by the brain depends on which neuron is activated, when it is activated, the duration of its activation, and how often it is activated. While imaging techniques such as magnetic resonance imaging (MRI) have allowed us to look into a working brain while people perform particular tasks, the resolution is low. It has helped us determine which brain regions are important for a given task, but it does not tell us the specific neuronal activity pattern needed to store information in the associated region of the brain.

A recent paper in the journal Nature sheds light on how neurons encode a memory, and how the brain uses this information to make decisions [2,3]. The researchers in this study, led by Dr. David Tank, aimed to better understand brain function by developing new and exciting techniques that now allow imaging of individual brain cells in live mice as they perform behavioral tasks.

To achieve this, they inserted new genes into mice, causing the mice to produce proteins that enabled the researchers to directly visualize neuronal activity. When a neuron is activated, pores on the surface of the neuron open up and allow electrically charged calcium to flow into the cell. This calcium influx triggers the release of chemicals from the cell, which in turn allows the neuron to communicate with its neighbors. To track the location, timing, and duration of the activation of individual cells, scientists inserted a special gene into the mice that encodes a protein that causes cells to light up in the presence of calcium.

However, in order to clearly image brain cells, the mouse’s head must be still. This is a problem if you want to look at neuronal activity while the mouse is running around. Cleverly, Dr. Tank sidestepped this issue by developing a virtual reality system that allows the mice to run through a maze while their heads are held in place. The mice are trained to run on a suspended ball, the rotation of which controls the scene presented on a screen in front of them. The mouse is thus able to navigate through the virtual environment, and can be taught to run through a maze in order to gain a reward, all while researchers look at which cells are being activated during this task [4].

Tank’s team devised a maze that tests both a mouse’s memory as well as its ability to use its memories to make decisions. In the maze, mice run through a straight corridor that contains visual cues. Different cues alert the mice to turn either left or right at an upcoming fork for a reward. The mice then pass through a ‘delay’ period, a straight corridor that lacks any visual information. After this corridor, mice must remember the previous cues in order to make the correct navigation choice that will lead to a reward.

Tank and his colleagues found that as mice ran through this maze, a certain sequence of neuronal activity was triggered and differed depending on whether the cues signaled a left or right decision. Each individual neuron involved in these sequences was only active for a short period of time, but their combined activity formed a specific and distinct temporal sequence that began after receiving either a left-turn signal or a right turn one. This activity pattern was similar between trials near the beginning of the task, but as the mouse ran, the “right” and “left” firing sequence became increasingly distinct, until the signals were easily distinguishable by the time the mouse decided to turn. Visual cues encountered during the first part of the task trigger a specific pattern of neuronal activation, allowing the mouse to choose the correct path later. Interestingly, when the mouse made a wrong turn, the neuronal pattern began correctly, but at some point during the trial switched to the neuronal pattern of the opposite turn. Researchers could actually see the mouse change its mind as the neuronal firing pattern shifted. This shift was most likely to occur during the delay period, but could occur at any time during the task, even when the mouse was still running through the visual cue area. While specific neurons preferred either left or right, these neurons were intermingled together within the same area of the brain, indicating that although large regions may be responsible for certain types of tasks, within those regions the specific neurons required for different memories are mixed together.

This paper provides new insight into previous findings from studies on human memory. For example, researchers have found that when people are having difficulty remembering a specific word, their memory may be triggered by a word that shares common features (e.g., someone might recall the word “fluorescence” after someone else mentions the word “floor”). While no one knows whether words are stored in a similar manner as navigational memories, you can imagine that the ‘floor’ sound might trigger a neuronal activity sequence shared by the word fluorescence [5]. This may also help explain age related dementia and confusion – if the connections between neurons are not as strong, it may lead to more frequent switching between activity sequences.   While Tank’s research provides us valuable insight into how the brain processes and stores information, more research is necessary to determine how this occurs with different tasks and in disease states.

Rebecca Reh is a Ph.D. candidate in the Program in Neuroscience at Harvard Medical School.

Additional Resource

Video of mouse moving through virtual reality maze (from Tank study)

References

[1] Squire, Larry and Wixted, John. The Cognitive Neuroscience of Memory since H.M. Annual Review of Neuroscience, v. 34: 259-288 (July 2011).

[2] Harvey, Christopher et al. Choice-specific sequences in parietal cortex during a virtual-navigation decision task. Nature, v. 484: 62-68 (5 April 2012).

[3] Zandonella, Catherine. Princeton scientists identify neural activity sequences that help form memory, decision-making. News at Princeton. <http://www.princeton.edu/main/news/archive/S33/17/36M20/index.xml?section=topstories> (28 July 2012).

[4] Keim, Brandon. Scientists Scan the Brains of Mice Playing Quake. Wired Science. <http://www.wired.com/wiredscience/2009/10/mouse-virtual-reality/> (14 October 2009).

[5] NPR: The Bryant Park Project. Scientists Changing Theories About Memory. <http://www.npr.org/templates/story/story.php?storyId=91142967> (4 June 2008).