BY WILLIAM SKAGGS
For many years scientists believed that you were born with all the neurons you would ever get. The evidence for this dogma seemed strong: neuroanatomists in the early 20th century had identified immature neurons under the microscope but only in the brains of mammalian embryos and fetuses, never after birth.
We now know that the truth is not quite so simple. By radioactively labeling DNA, researchers gradually began to find exceptions to the rule against new neurons in the adult brain. Today scientists have identified two small regions where neurogenesis, or the birth of new neurons, continues throughout life: the olfactory bulb and the hippocampus. The former area is part of the brain’s odor-discrimination system, so neurons there likely participate in this process. But the hippocampus has a much broader function. It gives us memory.
The discovery of nascent neurons in the adult human hippocampus, first reported in 1998, came as a surprise to many in the field. Although sprouting new brain cells may sound useful, the costs are potentially high. After all, space within the skull is finite, and newcomers could disrupt the delicate neural networks that store knowledge.
Neuroscientists now suspect that neurons born in the hippocampus help the brain create and sift through the millions of memories we form over the course of a lifetime. If this is true, neurogenesis might solve a puzzle that has perplexed memory researchers for more than 60 years: how our brain keeps separate memories of similar events. These discoveries may ultimately reveal not only how we recall the episodes of our lives but also how we can preserve our brain’s powerful record-keeping faculties despite the inevitable decline of aging.
Making Memories
In 1949 Canadian psychologist Donald O. Hebb proposed a theory of memory that would come to dominate the field. Hebb suggested that each neuron in the cerebral cortex, the brain’s large outer layer essential to thought and intelligence, encodes some feature of the world and becomes active whenever that feature is present. He also noted that every brain cell is connected to many others by links called synapses. His idea was that we encode memories by creating alliances between groups of neurons. When two connected neurons are active at the same time, the synapses holding them together grow stronger. In other words, “cells that fire together wire together.” To understand how this works, imagine how a memory might map onto a set of interconnected neurons.
Say you take a trip every summer. On one occasion, you pack your backpack for a journey to the mountains, including your favorite book. The features of the event—backpack, mountains, book—will each map onto separate neurons in the cortex. As you unpack the book at your alpine campsite one evening, those neurons fire together, bolstering the connections between the three elements and thereby storing the memory. In reality, the brain uses far more than three neurons and their connections to encode memories, but the principle is the same. If any one of the neurons in the stored memory were to become active later, an electrical impulse would propagate to the other cells in that network. As a result, the neurons representing all three features would fire, encoding the full memory.
This process, called pattern completion, is the way that memories are retrieved, according to Hebb’s theory. It explains how merely glimpsing the backpack after the trip can conjure up mountain vistas in your mind’s eye. Yet this explanation of memory has a problem: What happens when features from different memories overlap? Suppose, for example, that on a second summer vacation you pack the same backpack, but this time with a newspaper for a trip to the beach. For this memory to be stored, neurons relevant to the backpack, beach and newspaper will need to connect. When you recall the episode, the attempt to perform pattern completion will activate the backpack neuron and send a burst of activity through both sets of connections. The memories of the two trips would become conflated. This phenomenon is known as interference. It is an inevitable consequence of Hebb’s hypothesis and is not easy to fix. Neuroscientists have spent decades devising ways around the interference problem.
One simple solution is to minimize the number of shared features in the memories to be stored. The most straightforward way of doing that is to use features that are very specific. For example, instead of just storing the memory of a book in your backpack, you mentally classify it as a copy of James Joyce’s Ulysses and the newspaper from the beach trip as the New York Times. Yet this work-around has its drawbacks. The brain learns about the world by detecting patterns: consistent relations among sets of features.
For example, you may come to appreciate sunblock after sustaining multiple burns on beach trips in which you neglected to pack sunblock. If these features are categorized so specifically that they rarely recur, the memories to which they belong will offer no basis for learning. The importance of sunscreen, for instance, applies to every sunny day—regardless of the beach you visit or brand of lotion. These constraints seem to pit memory and learning against each other. Optimizing the brain for memory requires minimizing overlap, whereas learning depends on easy access to common elements so we can make associations.
Neurogenesis to the Rescue
Forty years after Hebb proposed his theory, three neuroscientists came up with an alternative approach. James L. McClelland and Randall C. O’Reilly, then at Carnegie Mellon University, and Bruce L. Mc-Naughton, then at the University of Arizona, were pondering the two brain regions involved in memory— the cerebral cortex and the hippocampus—when it dawned on them that the brain might resolve the conflict between learning and memory by separating the two processes.
They suggested that to prevent the interference problem, the cerebral cortex would help us forge connections and the hippocampus would focus on filing away distinct memories. They dubbed this hypothesis “complementary learning systems.” Their basic idea hinges on adding another set of neurons to the memory network formed in a Hebbstyle trip to the beach. Each of these additional cells tags a small set of memories.
Let’s say you embark on yet another trip with your trusty backpack. Instead of linking the features of all backpack vacations to one another, the brain allocates a single memory neuron for the latest adventure, and the trip’s salient features all link to it. That single memory cell would reside in your hippocampus, whereas the feature-related cells would dwell in the cerebral cortex. Moreover, the cells involved in memory in the hippocampus fire only in discrete groups because they inhibit or compete with one another. In consequence, only one memory can be active at given time.
When McClelland and his colleagues advanced their theory, evidence for fledgling neurons was still weak, but changed. In 2006 neuroscientists Fred H. Gage of the Salk Institute for Biological Studies, Gerd Kempermann, then at the Max Delbrück Center for Molecular Medicine in Berlin, and others recognized the potential importance of new neurons in the hippocampus. In two separate papers, they proposed that neurogenesis might be the brain’s way of continually expanding its stores of memory. For one thing, they reasoned, the new cells can more easily connect to other neurons than older cells.
A second clue is that young neurons have a more uncertain fate than older neurons do. Many of these new cells die in their youth, but their probability of survival improves when an individual is forced to learn unfamiliar tasks—a prime opportunity to form new memories. In fact, as Rutgers University neuroscientist Tracey J. Shors observed in 1999, the rate of neurogenesis can increase during learning exercises. Thus emerged a radical new idea in the science of memory. When the brain needs to create mental records,it might just grow more neurons.
