Mental Leaps Cued by Memory’s Ripples

The same mental processes that organize memories may also coordinate how we make decisions.

Ashley Mackenzie for Quanta Magazine

Over the past few decades, researchers have worked to uncover the details of how the brain organizes memories. Much remains a mystery, but scientists have identified a key event: the formation of an intense brain wave called a “sharp-wave ripple” (SWR). This process is the brain’s version of an instant replay — a sped-up version of the neural activity that occurred during a recent experience. These ripples are a strikingly synchronous neural symphony, the product of tens of thousands of cells firing over just 100 milliseconds. Any more activity than that could trigger a seizure.

Now researchers have begun to realize that SWRs may be involved in much more than memory formation. Recently, a slew of high-profile rodent studies have suggested that the brain uses SWRs to anticipate future events. A recent experiment, for example, finds that SWRs connect to activity in the prefrontal cortex, a region at the front of the brain that is involved in planning for the future.

Studies such as this one have begun to illuminate the complex relationship between memory and the decision-making process. Until a few years ago, most studies on SWRs focused only on their role in creating and consolidating memories, said Loren Frank, a neuroscientist at the University of California, San Francisco. “None of them really dealt with this issue of: How does the animal actually pull [the memory] back up again? How does it actually use this to figure out what to do?”

The new results are also prompting a broad shift in our understanding of the hippocampus, a C-shaped nub of brain tissue behind each ear. Since the late 1950s, when the region was famously tied to memory loss in the patient known as H.M., researchers have focused on its role in creating and storing memories. But newer studies have shown that the hippocampus is active when people imagine performing a task in the future. Similarly, people who have damaged hippocampi cannot imagine new experiences. The hippocampus doesn’t just allow instant replays — a kind of mental time travel into the past — it also helps us mentally leap forward.

In fact, complex planning may be the true benefit of the hippocampus. “That’s the point of having a memory, right?” Frank said. “To go back to the experiences you’ve had, extract general principles from them, and then use those principles to figure out what to do next.”

A Symphony of Memory

The New York University neuroscientist György (Yuri) Buzsáki still remembers the day, sometime in 1982, when he first heard the distinctive sound of the brain’s instant replay. He was a postdoctoral fellow at the University of Western Ontario in Canada at the time, a young researcher trying to learn how to record the soft pings of single neurons.

As he inserted an electrode through layer after layer of a rat’s brain tissue, an attached speaker emitted intermittent pops of neuronal noise. Suddenly, as the wire hit the bottom of the hippocampus, the speaker came alive with a crashing chhhhh sound. It disappeared within a fraction of a second, but Buzsáki was dumbstruck. “I didn’t know what to make of it,” he recalled. It was as if he had been listening to a group of orchestra musicians idly tuning their instruments, and the next moment they abruptly united in the thrilling harmonies of Beethoven’s 5th Symphony. “It was the most synchronous pattern I had ever seen,” he said.

In 1989, Buzsáki published a seminal paper proposing that the purpose of SWR brain waves in the hippocampus is to help organize and consolidate memories. He outlined a two-stage model of memory. In the first stage, called active learning, cells in different parts of the cortex — the outer layers of the brain — broadcast signals representing sensory information such as the screech of a seagull or the smell of the surf.

In Buzsáki’s second stage of memory, the consolidation stage, the hippocampus receives those disparate firing patterns and synthesizes them into a single SWR. The SWR encodes the memory, storing it in some other part of the brain for future retrieval. Just as an entry in a book’s index lists all of the pages that mention a particular word, an SWR seems to provide a distinct code that reminds the rest of the brain of a specific past event. “We call it the index code,” said Bruce McNaughton, a neurobiologist at the University of California, Irvine, and the University of Lethbridge in Canada. With their unique codes, SWRs allow us to store a coherent memory assembled from a diverse array of sensory inputs.

In 1994, McNaughton and his colleague Matthew Wilson, now a neuroscientist at the Massachusetts Institute of Technology, published a paper in Science providing the first direct evidence for the memory consolidation theory. Scientists already knew that SWRs only crop up when animals are sleeping, under anesthesia, or in a period of immobility. They also knew that neurons fire in a certain pattern when animals run through a simple maze. McNaughton and Wilson inserted large arrays of electrodes into the brains of rats and monitored these patterns. They found that the specific neural patterns that rats produce while running through a maze will later repeat as SWRs. This replaying of experiences during sleep, the researchers hypothesized, was somehow part of how the brain transferred memories from the hippocampus into long-term storage in the cortex.

Olena Shmahalo/Quanta Magazine, source: Nature Neuroscience

Over the next two decades, researchers established that SWRs are a crucial component of memory consolidation, and not only during sleep. In 2006 Wilson and his postdoctoral researcher David Foster reported in Nature that the sequence of brain activity that fired when a rat was following a straight track also replayed in a precise order during rest periods immediately afterwards. Intriguingly, the researchers found that the pattern fired in reverse — the rat reviewed its most recent experiences first, perhaps because these memories were the most useful to the animal in its current state.

The Decision Tree

Over the past few years a number of research groups have investigated how animals use these instant replays to guide behavior. Loren Frank’s team has dug into this question by training rats to run on metal tracks shaped like an E rotated 90 degrees. The animals are rewarded when they follow a certain pattern: from the middle arm to the left arm, then back to the middle, then to the right arm and back to the middle, and so on. To do this successfully, they must learn two rules. The first one is easy: If I’m in the outer arm, go to the middle. The second rule takes more thought: If I’m in the middle arm, go to the outer arm I didn’t just come from.

When rats pause in the middle arm, their hippocampi sputter SWRs, and these codes represent the paths the animal has just taken in the maze. It’s as if the animal is stopping to consider: Where did I just come from? Where should I go next?

In a study published in Science in 2012, Frank and his colleagues showed that these SWRs turned out to be essential to learning the task. When the researchers wiped out SWRs with electrical stimulation, the rats could no longer learn the alternating maze pattern. They learned the first step — go out and come back — but not the second. Frank theorizes that the SWRs integrate a rat’s immediate past information — what it just did — with what it needs to do next.

Some neuroscientists, however, aren’t ready to say that SWRs are intimately involved in planning, decision making, and other complex behaviors that rely on the coordinated activity of many brain regions. Buzsáki, surprisingly, is one of the skeptics, believing that SWRs are limited to memory consolidation. “I love sharp waves, and I try to promote these [studies] as much as possible,” he said. “But I’m a little cautious.”

He points out that SWRs last for only a short period of time, around 100 milliseconds. “That’s way too short to arouse the entire brain,” he said, citing studies showing that the act of making a conscious decision requires at least 500 milliseconds. What’s more, SWRs often occur when decisions are not being made. “If they were related to decision making, then you’d expect them to occur when you’re fully alert, not when half asleep,” he said.

Long-Distance Connection

If SWRs really are involved in planning and decision making, then they must somehow communicate with brain regions outside of the hippocampus. “The first question is, does the rest of the brain care?” Frank asked. According to his latest study, it does. “The rest of the brain does care, and it cares quite a lot.”

In this study, Frank’s team used the three-arm track again, but this time the researchers recorded the response from neurons in a rat’s prefrontal cortex, the region associated with planning for the future, at the precise time that the animal’s hippocampus was sending out SWRs. Frank found that SWR codes are strongly synchronized with firing patterns in the prefrontal cortex.

During an SWR, about one-third of the neurons in the prefrontal cortex change their firing activity, he found. This study, which Frank presented in June at the Areadne conference in Santorini, Greece, is the first to find such a tight, coordinated connection between SWRs and the prefrontal cortex while an animal is awake, Frank said. “It looks like there is a very strong mode of communication between the hippocampus and this prefrontal brain area, which is quite a long ways away.”

Buzsáki is excited about the evidence of interactions between the hippocampus and the rest of the brain. He cites a 2012 Nature study in which researchers scanned the brains of monkeys under anesthesia and showed that SWRs seem to activate most of the cortex while at the same time silencing non-cortical areas.

“What they observed is quite remarkable,” Buzsáki said. “The hippocampus can reach out to wide areas of the cortex.”

When he first put forth his theory that SWRs are involved in memory consolidation, Buzsáki thought that these interactions were somewhat one-directional, with the cortex sending messages to the hippocampus during learning and the hippocampus sending messages to the cortex during memory consolidation. But now he suspects that communication between the cortex and the hippocampus is multilayered. “Things are trickier now, and more beautiful,” he said.

This article was reprinted on

View Reader Comments (1)

Leave a Comment

Reader CommentsLeave a Comment

  • The paper DOI: 10.1126/science.1236425 ´Deep cortical layers are activated directly by the talamus´ describes, that a new stimulus is processed with two identical signals in different layers (lamina) of the cortex.
    To process an identical signal parallel in different areas might be a hint of a comparison procedure, which additional might be necessary for decisions and plans for the future.

Comments are closed.