Does Making Predictions Impede the Formation of Memories?

When your brain is presented with a scenario, it makes a decision: Does it file this moment away as a unique event, or slot the information into an existing pattern?

The lyrics to “We Didn’t Start the Fire,” Henry VIII’s wives, all the winners of RuPaul’s Drag Race, the anatomy of a kidney—I remember all these things just fine. But ask me what I did two days ago, or to retell an important story from high school, and I often blank.

I have a bad memory for events. Facts, figures, and concepts crystallize in my mind with no problem, but when I try to conjure up a moment from the past, I can feel my brain scrambling to assemble an amalgam of trends and generalizations, cobbling them together into some plausible reality. The details are never vivid; the timelines are always wonky. This was a huge source of tension between me and my parents when I was a teenager—they could recall in excruciating detail all my trespasses, while I was often left with no way to verify or refute the record. I developed workarounds: I wrote everything down (giving myself permission to forget), and I pegged circumstances to as many facts as I could so no memory was an island. I used every “present” moment to set myself up for the best future odds at all times, trying to anticipate future events and plan out likely scenarios, plot alternate decisions. If I couldn’t rely on my memory, at least I could trust that my past self had done everything “right.”

Memory is an unruly machine, embedded in a Russian nesting doll of systems and circuits that is the brain. When I got to university, I learned in an introductory neuroscience course what I’d always suspected: that remembering events versus facts are two different (but related) phenomena. There are many branches and subcategories of memory. There’s nondeclarative memory, the kind that leads to habit formation or muscle memory. Then there’s declarative memory, which exists in two flavors: episodic and semantic. Episodic memory is the kind that allows for retrospective mental time travel; its fraternal twin is semantic memory, the kind we use for facts and ideas.

“I have great semantic memory but terrible episodic memory,” I told my friends. And I believed it.

In late 2020, I came across new research suggesting that if you are trying to predict what is going to happen next in any given moment, even subconsciously, you impede your brain’s ability to encode your current moment in memory. Meaning: Sometimes, looking forward can prevent you from being able to later look back. In the study I read about, neuroscientists showed participants a series of photographs, each of which depicted a particular setting (beach, mountain, field, forest, etc.). Unbeknownst to participants, some image categories always followed others—beach images preceded mountain images, making beaches “predictive” and mountains “predictable.” Researchers wanted to see if the participants would pick up on those patterns consciously or subconsciously, whether their brains showed evidence of predictive behavior, and how all this might affect their ability to recall.

When scientists later put participants through fMRI, they detected distinct brain patterns representing each image category. They looked for those patterns while showing participants the same images again—but this time, they mixed in new photo examples from the same categories. With each image, scientists asked participants whether it was one they had seen before. Participants’ success rates with images for other categories were much higher than for the predictive ones (like the beach pictures).

But the most remarkable finding was how predictions showed up in the brain. Detecting patterns means compromising, shifting gears away from memory formation, for the sake of prediction. When images from “predictive” categories were on display, the brain showed neural patterns for the “predictable” categories—for example, when a beach was on-screen, the participant’s memory centers fired signals as if they were seeing a mountain. The stronger this effect, the worse that participant did in the memory test. And it wasn’t simply that these people couldn’t recall the information—they hadn’t encoded or stored memories for the beaches in the first place.

Reading that paper, I felt like Dorothy stepping out of her house and into Oz for the first time, life in technicolor. Might it be harder for me to revisit the past because I’m always living in the future? Perhaps anticipation, trying to predict what will happen, has always been a diversion for me, where the “now” gets swallowed up by my guessing at whatever comes next.

Memory is an unruly machine, embedded in a Russian nesting doll of systems and circuits that is the brain.

To be clear, I’m not looking to explain away my memory problems with this one phenomenon. Attributing the entirety of my experience to my proclivity for prediction and anticipation would be an overdramatic extrapolation, and there are certainly other variables at work. Still, Brynn Sherman, the lead author of the prediction research, says that it’s not unfair to say these effects manifest in everyday life to some extent. When your brain is presented with a scenario, it makes a decision: Does it file this moment away as a unique event, or should it slot the composite information into an existing pattern or trend? Our memory centers are constantly toggling between those two functions, toeing that line.

From an evolutionary perspective, Sherman says it’s a way to minimize redundancy. If you have a handful of memories going to a certain bush for nonpoisonous berries, you can probably bet that it’s a safe spot for the next time you want a snack. There’s no need to clutter storage space with every discrete visit from that point on—you have all the info you need. In a modern scenario, your brain may not need to remember where you parked because you always go to the same spot. You might forget what you ate yesterday, but don’t you always spring for the same lunch salad? Memory is learning, and a successful prediction means you have a firm grasp on a lesson.

Intuitively, this research makes a lot of sense to me. I’m big on efficiency, and at times I can’t help but jump to conclusions, tease apart situations, attempt to decipher patterns. I’m always comparing something to something else, or bracing myself for a worst-case scenario. Scientists don’t yet know how the brain decides to switch between those gears—between filing a memory as something unique, or placing it within a larger trend or prediction—but I must have one of those brains that is quick to pull the trigger on the latter. In a way, it’s almost an act of hubris on my part, a way of telling myself that I have all the necessary information and that nothing else is required to keep me alive.

Cognitive neuroscientist Endel Tulving is credited with first proposing the distinction between the two branches of declarative memory. He compared episodic memory to a kind of map that connects items of semantic memory in proverbial space: If semantic memories are plots of knowledge parceled out on the landscape of your mind, episodic memories are the scenery that fills the spaces between—the grassy knolls, the glens and creeks: context.

With my subpar episode recall, I sometimes feel unmoored from context. It’s like I’ve lived my life collecting viewfinder snapshots, mining scenes for their parts, squirreling facts into the folds of my brain. In high school, while preparing for AP exams, I remember learning that visualizing yourself in a setting and imagining realistic but positive outcomes for that scenario helps improve your performance. I really took that to heart. I preemptively visualize almost everything, even now, from job interviews to dates to social gatherings. Why wouldn’t I? I want to optimize my experience, perhaps save myself from potential embarrassment.

This kind of prediction likely affects memory too, though not in quite the same way. Sherman says that if she had to theorize, each elaborate hypothetical scenario probably works as a schema, a kind of memory scaffolding. By imagining detailed play-by-plays, the mind builds a framework of what it expects to happen. When the real show begins, the brain then has to reconcile that framework with real-world events as they come. But our neural circuits are imperfect, so it can be hard in retrospect to tease apart the actual versus predicted realities.

Sherman hopes to further study the relationship between episodic and semantic memories and whether individuals can intentionally affect how their brains store that information. But the brain is constantly making trade-offs, and swaying that decision in one direction likely means compromising the other. That said, Sherman doubts that anyone can really alter their brain’s default habits and tendencies, although you can probably consciously exert some slight influence on a case-by-case basis.

I’m learning that you can’t always live in the future—or, more accurately, that I no longer want to. I want to be able to remember not just fragments, but whole chronologies. I don’t want present me to take away future me’s ability to remember; it does us both a disservice.

I can consciously try to stay in the moment when I want to remember, but I will probably always be a pattern-seeking person, stuck with my brain and all its low-def glory. As far as my memories go, I’ve been spending a lot of time lately trying to think back and unearth those buried or missing gems from my childhood and teen years. But if predictions do impede memory encoding, maybe I just never formed some of those memories to begin with. You cannot excavate what you never buried.