Member preview
Psst...there’s an audio version of this story. Upgrade to listen.

Have I Been Here Before?

The science behind memory and déjà vu

Photo: Pixabay

How does your brain tell the difference between somewhere you have been and somewhere you haven’t? As we wander around our world, we encounter many familiar places along familiar routes — houses and shops, bars and cafes. If we wander off somewhere new, we instinctively, instantly, know it is new. And we form new memories: These new places become familiar.

By clever tricks of wiring in the hippocampus, our brains store this new place as a new memory set apart from all previous memories. When these tricks go wrong, they give you the creeping sensation — the déjà vu — that, impossibly, you’ve been here before.

When you walk into work, how do you know you’re at work? Or at home, or the shops, or your office? That constant feeling of familiarity is a really odd, metacognitive sensation. Without thinking, you know that you know where you are every minute of every day. Imagine if we couldn’t feel familiarity. Everywhere, everywhen, would be a new experience. It would be overwhelming, disorienting, every minute of every day like being a newborn baby who happens to walk, talk, and wipe its own bum.

Some seriously heavyweight calculations are happening in your brain to make such an everyday sensation. Deciding if someplace is familiar or not needs a comparison between the set of things we can see around us and the set of memories we have of places. If the sensory information coming in from our eyes matches a stored pattern of sensory information, then we are somewhere familiar. That stored pattern is a memory. This comparison uses a neat trick: We don’t need a complete set of sensory information, nor do we need to laboriously compare each separate memory to the incoming sensory information, like a rubbish magician checking each and every card in the deck to find out which one was yours. No, our brains have solved these two problems in one go.

The solution is pattern completion. The incoming sensory information is passed to a set of neurons in the CA3 area of the hippocampus. The neurons there are interconnected and can excite each other. These connections are weak, except when the connections are between pairs of neurons that are part of the same memory. That is, memory is stored here as the pattern of which neurons are active.

Now imagine that you glimpse somewhere familiar (even in an unfamiliar context — your office, but with lots of smiling people; your couch, but it’s clean). The sensory information, incomplete as it is, activates a few neurons in the existing pattern for your office (or your couch). These neurons in turn excite others in the pattern, as they have strong connections to them. These others in turn excite the original ones back, in a cascading, swirling feedback of activity, and the whole pattern of neurons lights up. So our brain finds the familiar memory—and with only partially complete information.

But if the sensory inputs go to CA3 neurons that are not strongly linked, then they will not activate each other or recruit others — there will be no pattern completion, because there is no pattern. So we are someplace new. And we want to remember this place.

Then we need the even cooler trick of pattern separation.

Pattern separation is vital. We don’t want the new memory to overwrite an old memory. We need to make sure the new information is stored as a pattern as separately from old patterns as possible. So we need to make sure that the set of neurons whose activity encodes the new memory is different from the sets of neurons encoding all existing memories. Even when — especially when — a new memory is similar to an existing one, our brains need to make sure they are stored far enough apart that recalling one does not automatically recall the other instead.

Stop reading, and look up. Where are you? You’ve likely been in many places similar to the one you’re in now. (Unless you’re in the belly of a whale. In which case, put your damn phone down and go explore.) Similar offices, bedrooms, or lounges. Similar trains, planes, or automobiles. For your brain to know exactly where you are—that you are in this specific place and not one of the many places very much like it—your brain needs to access the specific pattern of neurons that hold this specific memory, without any interference or corruption from memories of similar places. And this means storing any new memory of a place that is similar to old memories must be done by different neurons. The new memory must be separate.

I mean, how many churches have you been inside? Old churches have a bauplan: a classic cross shape, a long central aisle, vaulted ceiling, altar at one end, and every nook and cranny is festooned with candles and statues. So many churches, all so similar. If your brain could not separate the memories of the inside of each new church you’d been in, then each time you stepped into a church, all the memories of all the churches you’d ever been in would turn on, merging into one melange of churchness. But you clearly know when you’re in Notre Dame and when you’re in St Paul’s. Without the brain’s ability to separate these very similar but not identical collections of “stuff in a church” into separate memories, we’d be hopelessly confused.

Pattern separation seems to depend on the dentate gyrus, one of the less glamorous bits of the hippocampus. No Nobel prizes yet for this bit. The dentate gyrus receives the same input, the same sensory information, as CA3. But it is has weird features that combine to make its party trick of pattern separation.

For one, the dentate gyrus has many more neurons than it has inputs, so it deliberately spreads out those inputs, making sure only a few neurons get to sample the incoming sensory information. This means only a few dentate gyrus neurons should be active in any given place.

For another, the main neurons in the dentate gyrus do not connect to each other, so they cannot activate each other and interfere with each others’ activity. More, these main neurons send input to the CA3 region that performs pattern completion. And although the dentate gyrus has many more neurons than the CA3, one dentate gyrus neuron connects to just a handful of CA3 cells. So, again, not only is the input deliberately spread out, the output is as well.

But each connection neurons make in the dentate gyrus to those in the CA3 is very strong. So strong that the connection is called a “detonator” synapse: It’s strong enough to cause an explosion of activity in the CA3 neuron each and every time that connection is active.

Now let’s put that all together. You’re in a new place, and pattern completion has not happened — your CA3 is not bursting with the activity of a stored memory pattern. Then the sensory information about the new place will be spread out across the activity of just a tiny set of dentate gyrus neurons. And because that set is so tiny, in all probability this set of dentate gyrus neurons will never have been active together before. It is a unique, separated pattern. Yet because the connections to CA3 are so strong, this tiny, unique set of neurons in the dentate gyrus will activate a unique set of CA3 neurons at the same time.

And then, the coup de grâce: Because these CA3 neurons become active together, the connections between them will get stronger. And what do strong connections between CA3 neurons mean? A memory pattern! So, the next time you visit this place, and that same sensory input arrives, the pattern can be completed. A memory is born. Separately.

As with most ideas in neuroscience, this is an evolving, churning theory. It originated from some astute theorists noting these odd properties of dentate gyrus neurons and saying, “Hey, doesn’t that look exactly like what we’d need to make sure our memories are stored separately?” In fact, it’s one of the few neuroscience theories that have genuinely progressed from thinking about how to solve a problem faced by the brain to a program of experiments and testing.

Three studies testing pattern separation ideas just came out together in the same issue of Neuron. These are the first studies in which researchers recorded from a large set of dentate gyrus neurons at the same time and could say for certain which neurons they were looking at. For a change, the studies all agreed on most things. They all showed that dentate gyrus neurons were very sparsely active, exactly as predicted by the theory of pattern separation: If the dentate gyrus is really separating patterns of incoming information, then only a few, different dentate gyrus neurons should turn on in each new environment. Better, all three studies showed that each neuron represented only a single point in the environment, again meaning that the activity of just a few neurons would clearly, cleanly represent a distinct set of places. Remarkably, after 40 years, the pattern separation theory is alive and kicking.

The theory also gives us clues as to what would happen if pattern separation failed. Imagine if that tiny, random set of dentate gyrus neurons happened to be not random enough and accidentally activated a set of CA3 neurons that were connected together—that were a previous memory. Then, although your eyes would be screaming at you, “Look at all this newness!” your CA3 would be telling you, “This is familiar.”

Déjà vu, we call it. When your eyes are telling you one thing yet your sensations are telling you another: “I’ve been here before.” That’s so creepy that it can lead people to rationalize the sensation of familiarity as the memory of a past life. But no. Déjà vu is a failure to separate, a failure of the brain to make sure a new memory isn’t interfering with an old one.

It’s remarkable this doesn’t happen more often. Given that we live an average of 70 years and experience around 2,207,520,000 seconds in our lifetimes. More than 2 billion seconds. We make memories throughout this time, often down to the scale of a few seconds. It’s likely that we store millions of new memories using our hippocampus, and yet a mere handful at most get accidentally turned on in the wrong place.

Give thanks for pattern separation. When you remember your wedding and not every wedding you’ve ever been to; when you remember your first day on your exciting new job and not every office where you’ve ever worked. Happy memories, kept separate.