copyright 2000 by George Johnson

originally published in Time, June 12, 2000. all rights reserved.

How Memories Are Made

by George Johnson

Scientists have long believed that constructing memories is like playing with neurological Tinkertoys. Exposed to a barrage of sensations from the outside world, we snap together brain cells to form new circuitry-patterns of electrical connections that stand for images, smells, touches and sounds.

The most unshakable part of this belief is that the neurons used to build these memory impressions are a depletable resource, like petroleum or gold. We are each bequeathed with a finite number of cellular building blocks, they believed, and the pile just gets smaller each year. that is certainly how it feels, as memories blur with middle age and it gets harder and harder to learn new things.

But like so many absolutes, this time-honored notion may have to be forgotten. In the last year, a series of puzzling experiments have forced scientists to rethink this and other cherished assumptions about how memory works, reminding them how much they have to learn about one of life's great mysteries: how the brain keeps a record of each individual's passage through life, allowing us to carry the past inside our heads.

"The number of things we know now that we didn't know 10 years ago is not very many," laments Charles Stevens, a memory researcher at the Salk Institute in La Jolla, Calif. "In fact in some ways we know less."

This is what we do know-or thought we knew. Memory traces, called "engrams," are forged deep inside the brain in an area called the hippocampus-named, because of its arching shape, after the Latin word for seahorse. From this temporary scratchpad, the patterns are transferred somehow (perhaps during sleep) to permanent storage sites throughout the cerebral cortex, the area behind the forehead often described as the center of intelligence and perception. There the information is thought to reside in the form of neurological scribbles, etched in clusters of connected cells.

It's been considered almost gospel that these patterns are constructed from the same supply of neurons that have been in place since birth. New memories, the story goes, don't require new neurons-just new ways of stringing the old ones together. Retrieving a memory is a matter of activating one of these circuits, coaxing the original stimulus back to life.

The picture appears eminently sensible. The trillion neurons in a single brain can be arranged in countless combinations, providing more than enough clusters to record even the richest life. If adult brains were cranking out new neurons as easily as skin and bone grow new cells, this would only serve to scramble memory's delicate filigree.

Studies with adult monkeys in the mid 1960s seemed to support the belief that the supply of neurons is fixed at birth. Hence the surprise when Elizabeth Gould and Charles Gross at Princeton University recently reported last year that the monkeys they studied seemed to be minting thousands of new neurons a day in the hippocampuses of their brains. Even more jarring, they found evidence that a steady stream of the fresh cells may be continually migrating to the cerebral cortex.

No one is quite sure what to make of the results. There were already hints that the spawning of brain cells, called neurogenesis, occurred in animals with more primitive nervous systems. For years, Fernando Nottebohm at Rockefeller University has been showing that canaries create a new batch of neurons every time they learn a song, sloughing them off when it's time to change tunes.

But it was widely assumed that in mammals and especially primates (including the subset Homo sapiens) this wholesale manufacture of new brain parts had been phased out long ago by evolution. With a greater need to store memories for the long haul, these creatures would need to ensure that the engrams weren't disrupted by interloping new cells.

Not everyone found the argument convincing. (Surely birds had important things to remember too.) When neurogenesis was found to occur in rats-and shortly after, even people-the rationalizations began to take the tone of special pleading: There was no evidence the new brain cells had anything to do with memory, or that they did anything at all.

That may still turn out to be the case with the neurons found by the Princeton lab. The mechanism Gould and her colleagues uncovered in macaque monkeys could be nothing more than a useless evolutionary leftover, a kind of neurological appendix. But if, as many suspect, the new neurons turn out to be actively involved with inscribing memories, the old paradigm is in for at least a minor tune-up and maybe a complete overhaul.

It is telling that the spawning ground for the neurons is the hippocampus, which is indisputably crucial to memory. Patients with hippocampal injuries lose their ability to acquire new facts, though they can still recall impressions laid down in the years before the damage occurred. Maybe, Gould speculates, the newly generated hippocampal neurons are especially agile at forming connections with one another. As in the canaries, the new cells would readily join hands to encode a new memory, then be flushed from the system when they were no longer needed-when the temporary impression has been transferred elsewhere for safekeeping.

That explanation would fit pretty well with the old theories. More puzzling, though, is another of the study's findings-the steady migration of new neurons from the hippocampus to the cerebral cortex. Could these neurons somehow be involved in ferrying information into long-term storage?

Perhaps, Gould and her colleagues ventured in a recent paper, this purported transport mechanism provides a means of "time stamping" long-term memories, helping us keep track of when we learned what. Older memories would be somehow associated with older neurons. The idea is presented as pure speculation; no one is even guessing how this might work. But if memories are indeed flowing through the brain in rivulets of new neurons, then all the old ideas will have to be reconsidered.

The brain is so complex and neuroscientific experiments so difficult to interpret that this whole picture could change in a year. Whatever happens with neurogenesis, the fundamental notion that engrams are made by stringing together neurons-whether new ones or old ones or a combination of the two-is likely to survive in some form.

In the meantime other laboratories are trying to refine their understanding of just how neurons forge these connections. Here too, many long established assumptions don't seem so solid anymore.

For the past 20 years, neuroscientists have been piecing together a story in which the key to linking neurons is a kind of molecular switch called an NMDA receptor. (The letters stand for the polysyllabic name of a chemical used to identify these molecules in experiments.) The mechanism is thought to work like this: If one neuron repeatedly sends signals to another neuron, its NMDA receptors respond by unleashing a cascade of chemical reactions that alter both cells, strengthening the bond between them. Just how these reactions work remains a matter of almost religious debate. But the result, the theory goes, is that more of the chemical messengers called neurotransmitters flow across the junctions connecting the neurons. In some cases, entirely new connections may be formed.

It's been known for years that mice whose NMDA receptors have been chemically blocked have trouble learning their way around mazes. In the most dramatic demonstration of the power of the idea, Joseph Tsien, another Princeton researcher, recently developed a genetically engineered breed of "smart mice" with souped-up NMDA receptors and showed that they had enhanced powers of memory.

But just as the pieces were starting to fall together, Tsien's lab did another experiment that complicated matters. Mice were bred with no NMDA receptors in a region of the hippocampus known to be especially crucial to memory. As expected, these mice showed seriously diminished memory power. But if they were then exposed to a stimulating environment full of toys and exercise wheels, they got their memory back. When the scientists examined the mice's hippocampal tissue with an electron microscope, they found that they had formed new neural connections, without the aid of the seemingly crucial NMDA memory switches. "That was really surprising," Tsien says.

There are a couple of plausible explanations. Neurons in the hippocampus might be making new connections using some entirely different means that has escaped researchers' attention. Or the connections normally forged in the hippocampus might be formed instead in the cortex, where the mice's NMDA receptors remained intact. Brains are amazingly resilient and it is common for functions lost in one area to be taken over by another. In any case, the neat lines of the old picture have been fuzzed up again.

Zeroing-in on the mechanism of imprinting engrams, and determining whether or not neurogenesis is involved, will be just the first steps in a long progression toward understanding how we remember. If memories are indeed stored as configurations of connected cells, then what do these patterns look like? How many neurons does it take to represent the image of your pet cat, and how is that pattern connected to the patterns that represent the abstract categories of cats, pets, mammals, and living things?

When you read a book, how are the neurons stitched together to record the memorable passages? How are they filed so you know the memory came from a book and not from your own experience? And while you are scanning the pages, how do you call up the patterns that represent the definitions of the words and their sounds, and the rules for unpacking meaning from a sentence?

Half a century ago, the neuroscientist Karl Lashley wrote a paper called "In Search of the Engram," describing his frustrating attempt to find the cluster of neurons in which a rat stored its memory of a maze. After training the animal to negotiate the labyrinth, he snipped away at the brain bit by bit. While the animal became increasingly sluggish and confused, Lashley was never able to find a single locale where the memory was inscribed.

"I sometimes feel," Lashley ruefully concluded, "that the necessary conclusion is that learning just is not possible."

Fifty years later, memory researchers find themselves with the same mix of confusion and awe.