Scientific Understanding of Consciousness
Science 5 October 2012: Vol. 338 no. 6103 pp. 30-31
How Are Memories Retrieved?
So much of memory is a puzzle. How can the experiences of a lifetime—the sights and sounds, people and places, successes and failures—be recorded in the soft tissue of the brain? How can those memories persist for decades even as the neurons that encode them undergo constant molecular remodeling? And how can we (more often than not) recall a particular bit of information almost instantaneously, and with little prompting?
This last question may be the most mysterious of all. “Retrieval is such a rich phenomenon,” says Michael Hasselmo, a neuroscientist at Boston University. “You get a reminder from somebody that's maybe just a word and you somehow turn it into a rich reflection of events that you're moving through with a perspective and a location and a sense of time passing.” Our memories are part of what makes each of us unique, and they give us a sense of self-identity and continuity as we move through life. “Without our memories, we're just zombies,” says György Buzsáki, a neuroscientist at New York University in New York City.
The neuroscience of memory is a complex and contentious area, but most researchers agree on a broad-brush account that goes something like this, at least for episodic memories, or memories of events. These memories are initially encoded and stored mostly in the hippocampus, deep inside the temporal lobe of the brain. For long-term storage, memories are filed away to other areas, including the neocortex, the thin sheet of tissue on the surface of the brain. A memory of any given event, the thinking goes, is represented by a sparse and scattered network of synaptically connected memory traces through dendritic trees of neurons, such that the sights, sounds, and emotions associated with the experience may each be ingrained in different brain locations. To recall a memory, the brain must reactivate just the right network of pathways through the dendritic trees of a widespread and sparse pattern of linkages through many millions of neurons of the neocortex. Many details of this process are not known (or are disputed). Ongoing research work suggests a fluid role of memory, one in which retrievals plays a crucial role in reshaping memory over time.
So what should researchers look for if they hope to learn how the brain recalls the past? One clue comes from functional magnetic resonance imaging (fMRI) studies of the human brain suggesting that remembering reactivates some of the same neural circuitry as the original experience. Recalling a face, for example, activates a part of the fusiform gyrus thought to specialize in face recognition. Recalling a place evokes a different pattern of brain activity that includes the parahippocampal gyrus, an area that lights up when people view images of landscapes and other scenes.
“We have a pretty good idea that the brain uses the same machinery for remembering that it does for experiencing things,” says Loren Frank, a neuroscientist at the University of California, San Francisco. When it comes to episodic memories, Frank says, what's stored in the brain are little snippets of the experience that can be compiled into a kind of composite representation. The neural signature of memory retrieval, Frank argues, should look much like the neural signature of the actual experience played in fast-forward.
Do the hippocampal firing sequences of rats in a maze represent the rat reminding itself what it needs to do next? Buzsáki thinks so. “Plans are based on memories,” he says. Buzsáki speculates that such sequences also play a broader role in recalling episodic memories. “The hippocampus is like a librarian,” he says: Its job is to record new experiences, help file them away to the neocortex, and later retrieve them on demand. In the case of episodic memories, the firing sequence of hippocampal cells might serve the same purpose as the identifying bar code on the spine of a book, indicating which subset of neocortical neurons represents a given memory.
Recent studies support the concept that memory retrieval involves reactivating small but specific sets of neurons. In one, researchers used genetic engineering methods to tag neurons in the hippocampus that were activated as a mouse learned to associate a flash of light with an impending shock. When the researchers then reactivated those same neurons with a pulse of laser light delivered by an optical fiber, the mice froze in fearful anticipation even though they hadn't seen a flash of light.
Bridging the gap between such rodent studies and the human brain isn't easy. In rare cases, neuroscientists have taken advantage of monitoring electrodes placed in or on the brains of epilepsy patients awaiting surgery. Such studies are done only when they don't interfere with medical care. In a 2008 study, researchers asked patients to watch several short video clips from TV shows and movies and then recall as many of them as possible a short time later. Individual neurons in the hippocampus seemed to develop a preference for a specific clip, firing strongly a second or so before a patient named a clip, say, from The Simpsons, but not when he or she recalled any of the other clips. The findings provide some of the best evidence that reactivation of specific hippocampal neurons is involved in the conscious experience of memory retrieval.
Memory researchers have also begun to use methods borrowed from machine learning to analyze patterns of brain activity from fMRI scans on a finer scale than conventional methods allow. “We can actually now identify individual memory traces by the pattern of activity in areas of the brain such as the hippocampus,” says Eleanor Maguire, a cognitive neuroscientist at University College London. “We can predict in an experiment the memory that someone is recalling.” (But only when the choices are limited: The technology is nowhere close to being able to read out any fleeting memory that crosses someone's mind.)
Maguire says that work by her group and others is beginning to challenge the dogma that the hippocampus is not involved in recalling older episodic memories. The overarching role of the hippocampus, she hypothesizes, is to pull together the various aspects of a memory residing in different regions of neocortex and bind them into a coherent scene in the mind's eye. Recent studies suggest that the hippocampus even does this when people ponder imaginary scenarios. By stitching together pieces of the past, Maguire says, the hippocampus enables us not only to vividly remember but also to envision possible futures.
Another recent challenge to traditional views of memory retrieval comes from research suggesting that memories can be incrementally strengthened, weakened, or otherwise altered each time they're recalled. “That work has changed the way people think about the persistence of memory,” says Yadin Dudai of the Weizmann Institute of Science in Rehovot, Israel. Such findings point to a more malleable memory system in which retrieval presents an opportunity to update old memories in light of new experience. Our repository of memories may be less like a library and more like Wikipedia, where each entry is open to editing anytime it's pulled up.
This type of plasticity may be crucial for fitting new memories into the existing network of old memories, says Howard Eichenbaum, a neuroscientist at Boston University. “Everything you learn has to fit in with what you already know,” Eichenbaum says. How the brain accomplishes that never-ending task is a puzzle scientists have only begun to explore.
[end of paraphrase]
Return to — Declarative Memory As Reconstruction