Memory Modeling – Your mind in rewrites
“Time and memory are true artists; they remold reality nearer to the heart’s desire.”
—JOHN DEWEY, RECONSTRUCTION IN PHILOSOPHY
Are You Sure You Saw “X”?
As a consulting analyst for the US Environmental Protection Agency, I spent quite a bit of time conducting research to find out if a national campaign to raise awareness about childhood lead poisoning was reaching the eyes and ears of the right people: Parents of young children in cities with older housing. One of the ways the lead-awareness message was advertised was in premovie reels at movie theaters in multiple cities. I conducted on-site research at theaters in Baltimore that included two types of in-person surveys: a premovie survey to determine peoples’ knowledge of the lead-poisoning issue, and a postmovie survey to determine if people had been influenced by the lead-awareness message. The first survey established a benchmark of awareness; the second gauged how much the advertising had elevated awareness. At the same time, the postmovie survey determined to what extent people recalled the lead-awareness advertisement.
After the first couple on-site studies, it became clear to me that something odd was going on. I knew going in that under the best circumstances, we would see at most a 30-40 percent recall rate for the advertisement (realistically closer to 25 percent). Yet we were getting 60-80 percent results at the movie theaters—double the expected rate, and virtually unprecedented for this sort of public outreach. So, in the next theater study I decided we should begin asking a couple new questions in the postmovie survey to determine if people could recall details about the advertisement and not just a message. Doing this revealed that relatively few people could recall anything specific about the advertisement, although they claimed that they did recall seeing the leadpoisoning-awareness message.
What was really happening? After some digging, I reached a few conclusions—chief among them that people were unknowingly picking up on shards of information in the theater lobby, such as briefly seeing the EPA logo at a table we were using, and some had heard the words “lead poisoning” as they walked by other people being interviewed, or overheard people that had completed the survey talking about it on their way out of the theater. The reality was that no more than 30 percent were actually recalling the premovie advertisement; others had assembled enough fragmented information that they were sure they must have seen it. The fragments comingled in their memory, and when they were asked a question that served as a sort of mental glue for the fragments, they were sure they had seen the ad.
This example illustrates a simple truth that we are typically reticent to admit: Our memories are wrong at least as often as they are right. At best, they are incomplete, though we might swear otherwise. This affects countless aspects of our lives, and in many cases our memories—true or false—affect others’ lives. But before exploring the fallibility of memory in greater detail, let’s take a few minutes for an overview of what we know about how memory works.
A Brief Primer on Memory
Memory can be separated into two main divisions: explicit and implicit. Explicit memory (also called declarative memory) principally holds words, numbers, and events. Or, to use the parlance of neurobiology, it is memory that’s semantic and episodic. When we are trying to remember what happened on the camping trip we took with our in-laws in late 2004, for example, explicit memory is engaged.
Implicit memory (also called nondeclarative memory) is where socalled muscle memory is found—the motor skills that, once learned, are always available to us. How is it that you never have to remember how to clip your nails or brush your teeth? Because implicit memory has you covered.
Those are the two main divisions of memory, but there are also two temporal (time-based) memory categories. The first is working memory, or what we usually call “short-term memory”—the category of memory that includes anything we are actively thinking about right now. In earlier models of memory, working and short-term memory were considered two different categories, with short-term thought of as a passive holding zone for information with closer links to attention and awareness. But in current memory parlance, these categories are synonymous.
The second category is long-term memory—the place where everything “remembered” resides. Long-term memory is engaged in the short term (e.g., word memorization) and in the long term (e.g., childhood memories). One way to think about short-term (working) memory versus long-term memory is that anything actively in your awareness right now involves short-term memory. In most views, short-term memory has extremely limited capacity—about seven items at a time. Long-term memory includes anything that is not currently active but could be recalled and made active. This could be anything from information you crammed for an exam yesterday to the address of the house you lived in when you were eight years old.
Perhaps the most exciting neuroscience discovery of the last several decades is that our brains are not static hunks of tissue but flexible and adaptive organs that change throughout our lives. The term used to describe this new understanding is brain plasticity. The flexibility of your brain is essential to memory and indispensable to learning. Specifically, the “plastic” parts of our brain are synapses—the connection points that allow neurons to transmit signals between each other. Hypermagnification of synapses in an adult human brain shows various sizes and shapes, some shaped like mushrooms, others shaped more like small hills, and others shaped like broad-based mountains. The incredible part is that in your brain and mine, synapses are morphing from one shape to the next depending on the need—how fluid the connection between neurons needs to be, for example—and this continues happening throughout our lives.
Another discovery related to brain plasticity is that the amount of activity between neurons corresponds directly to how strong their connection will continue to be. Cognitive scientists commonly use the phrase, “Neurons that fire together wire together.” What this means in terms of memory is that the more intense the activity is between neurons constituting your memory of any given event, the more robust the memory will be. That is one reason why emotionally charged memories frequently percolate to consciousness in vivid detail. I can still remember almost exactly where I was standing outside my high school in Florida just after the space shuttle Challenger exploded. I was looking up at the sky and could see the entire shape of the explosion outlined in smoke. All “where were you when” moments have a degree of this same intensity, which makes them easier to recall than anything else that was happening at the time. In fact, as I write this book, history transpired: The president announced that Osama bin Laden had been killed. I feel as though I now have bookend memories of the horrible events of 9/11. I can recall in detail what I was doing when I first learned of the terrorist attacks in 2001, and now I will have a memory of where I was when the mastermind of that attack was finally dispatched almost ten years later. These are powerful, emotional events accompanied by intense neural activity. The imprint, though still imperfect, remains with us for a lifetime.
That is not true, however, for most memories. Even for the sharper memories born from strong emotions (often called flashbulb memories), time erodes the infrastructure, leaving cracks and gaps. Instead of remembering specific, perfectly accurate details, what constitutes memory over time are general impressions of events with spotty details—and the older we get, the spottier they become.
Remember Once, Forget Twice
To figure out why that happens, among other perturbations, cognitive science has tackled memory more aggressively than perhaps any other topic. As a result, we have a growing wealth of research to draw upon to better understand the quixotic art of remembering. What we now know is that our brains happily reconstruct memories, though we are frequently fooled into thinking that the reconstructions are seamlessly recorded recollections.
What I want to convince you of in this post is that our memories are anything but concrete and can be altered with relative ease. That’s the bad news. The good news is that imperfect memory is an evolutionary adaptation that serves our species well much of the time. Loss of memory, and creation of new memory, is central to a relatively efficient system of information processing that never sleeps. The selective movement of information into long-term memory is an adaptive marvel that allows our brains to store crucial pieces of information that we will rely on in the future, and shed information not worth holding onto. The process is not neat and tidy, and memory selectivity often works against us (think about how many memories you would love to forget). But when you view the process through the lens of species survival, it makes unassailable sense. In those terms, it is crucial to remember where the best sources of food are located; where the best hunting grounds are located; which areas to avoid lest you become something else’s dinner; and how to return safely back home. For our ancestors, reliance on memory of particular details was a matter of life and death.
The problem for us moderns is that memory, incredible adaptation though it is, faces relentless challenges in societies driven by information. We simply have too much to remember at any given time, and the vast majority of our brains are not equipped to handle the deluge. Our expectations for what we should be able to recall are hardly in line with what our brains are capable of processing—which, by the way, is an enormous amount.
We have also adopted inaccurate metaphors for memory that lead us to incorrect conclusions. The “bookshelf” metaphor, for example, which suggests that when we need to recall a memory we simply find it categorized on a mental shelf, ready for consumption. Or the computer metaphor, which suggests that our brains store files on a cerebral hard drive that we can access as we would files on our laptop. These and similar metaphors are wrong for roughly the same reason: Memory does not reside in any one place in our brains, but rather is distributed across multiple brain regions. But because we more easily connect with a metaphor like those above rather than the messy truth of distributed and reconstructed memory, the misunderstandings persist.
Next we will review a few experiments that put a finer point on just how changeable memory can be.
A Photo Is Worth a Thousand Ways to Change Your Memory
Most of us realize that memory is fallible because of the little things that happen all the time. We forget things like car keys, passwords, whether we turned off the oven, and so on. But how many of us would admit that our memory is susceptible to change from the outside? That’s different from simply forgetting—something we all do on our own—because someone else changing our memory requires “getting in our heads,” so to speak, right?
The truth is that this sort of outside-in influence does not take very much effort to accomplish—just a few images and a little time. A study conducted by Linda Henkel in the Department of Psychology at Fair-field University tested whether showing people photos of completed actions—such as a broken pencil or an opened envelope—could influence them to believe they’d done something they had not, particularly if they were shown the photos multiple times.2 Participants were presented with a series of objects on a table, and for each object were asked to either perform an action or imagine performing an action (e.g., “crack the walnut”). One week later, the same participants were brought back and randomly presented with a series of photos on a computer screen, each of a completed action (e.g., a cracked walnut), either one, two, or three times. Other participants were not shown any photos.
One week later, they were brought back to complete a memory test in which they were presented with action phrases (e.g., “I cracked a walnut”) and asked to answer whether they had performed the action, imagined performing it, or neither, and rate their confidence level for each answer on a scale of one to four. The results showed that the more times people were exposed to a photo of a completed action, the more often they thought they’d completed the action, even though they had really only imagined doing it. Those shown a photo of a completed action once were twice as likely to mistakenly think they’d completed the action than those not shown a photo at all. People shown a photo three times were almost three times as likely to think they had completed the action as those not shown a photo.
Two factors in this study speak to the malleability of memory. The first is duration of time. The experiment was carried out with a week between each session, enough time for the specific objects and actions to become a little cloudy in memory, but not enough time to be forgotten. This lines up well with real-world situations, such as someone providing eye-witness testimony, in which several days if not weeks might elapse between recollections of events.
The second factor is repeat exposure to images. The study showed that even just one exposure to a photo of a completed action strongly influenced incorrect memory, while multiple exposures significantly increased the errors.
If the Video Says So, Then I Must Be Guilty
So if static images can be used to manipulate our memories, what about video? After all, in a world dominated by endlessly pliable electronic media, you can never be 100 percent sure that what you’re seeing on screen is what really happened. Two memory studies conducted by researchers Kimberley Wade, Sarah Green, and Robert Nash at the University of Warick (UK) illustrate that point nicely.
In the first study, researchers wanted to know if they could convince people that they had committed an act they never did. To accomplish this, they created a computerized multiple-choice gambling task for participants to complete, which entailed increasing the winnings from a sum of money as much as possible by answering questions. The money was withdrawn from an online bank based on cues given to participants by the computer program: When they answered questions correctly, they were told to withdraw money from the bank; when they answered incorrectly, they were instructed to deposit money back into the bank. Subjects were videotaped while they completed the task.
Afterward, participants were asked to sit and discuss the task with a researcher. During the discussion, the researcher said he had identified “a problem” during the task, and then accused the participant of stealing money from the bank. Some of the participants were told that video evidence showed them taking the money (but they weren’t actually shown the video), while others were shown video “proving” that they took the money. What the participants didn’t know is that the video had been edited to make it appear as if they did something they had not. Participants were then asked to sign a confession stating that they did in fact take money from the bank when they should have deposited it back. Participants were given two chances to sign the confession, and by the end of the day, all of them did. In fact, 87 percent signed on the first request, and the remaining 13 percent signed on the second. Interestingly, even participants merely told that the video showed them taking the money eventually complied with the confession.
The next study used the same principle, but this time to see if people would accuse someone else of doing something they had not. Again a gambling task was used, but instead of one person completing it, two people placed side by side completed it—sitting not even a foot apart, with monitors in full view of each other. Subjects were videotaped as before, and the video was doctored as before to show one of the two participants taking money.
Afterward, the “innocent” participant was asked to discuss the task with a researcher, and told that video proof had been obtained showing that the other participant stole money. In order to pursue action against that person, the researcher said, the innocent participant would have to sign a witness statement corroborating the video evidence. Some of the participants were, as before, only told that the video existed, while others were shown the edited video (and there was also a control group neither told about nor shown video).
The results: When first asked to sign the witness statement against the other person, nearly 40 percent of the participants who watched the video complied. Another 10 percent signed when asked a second time. Only 10 percent of those who were only told about the video agreed to sign, and about 5 percent of the control group signed the statement. These results point to the alarming power of video to shape and distort memory—not only about others, but about ourselves as well. In the first study it wasn’t only watching a video that made a difference; merely being told that a video existed made nearly as big an impact. And it is worth noting that in the second study, some of the people who signed the witness statement became so convinced that the other person was guilty that they went on to insert even more details of suspicious behavior, as if they knew the other person was doing something wrong all along.
Trusting Your Way into False Memories
The examples we just discussed address what can happen to our memory when visual information is manipulated. Let’s now remove the visual element and focus instead on how the integrity of information we receive affects memory—or, more precisely, the integrity of the information provider. If, for example, you follow a news commentator closely, reading everything he or she writes in whatever venue it appears, you may unknowingly be in a trust trap. Studies have shown that once we invest trust in a particular source of knowledge, we’re less likely to scrutinize information from that source in the future. A study conducted by Elizabeth Loftus at the University of California, Irvine, and her team took this investigation a step further, showing that the trust trap can also result in the creation of false memories—and not only in the short term.
Researchers crafted an experimental design in which they exposed two groups of participants to a series of images followed by narration about the images. The first group (referred to as the “treat-trick” group) received mostly accurate narration about the images. The comparison group received mostly misinformation. Both groups then completed tests of recall to determine how much accurate versus inaccurate information they remembered.
One month later, the participants were brought back to undergo the same experiment, except this time the treat-trick group was given misinformation during the narration (i.e., the “trick”), as was the comparison group. Both groups again completed tests of recall.
Here’s what happened: In the first session, the treat-trick group had a significantly higher rate of true memory versus the comparison group (which we would expect since only the comparison group was given misinformation during this session)—at a rate of about 82 percent for the treat-trick group and 57 percent for the comparison group.
But in the second session, in which both groups were given misinformation one month later, the treat-trick group had significantly lower true memory recall than the comparison group: 47 percent versus 58 percent.
The most likely reason for this effect is that the treat-trick group fell into a trust trap. Because information provided by the narrative source in the first session was accurate (and test scores were high as a result), participants believed the source to be credible and trustworthy. The comparison group, on the other hand, had no reason to invest trust in the original source and exhibited recall at roughly the same level for both sessions.
What’s most interesting is the time frame of this effect. Researchers conducted the sessions a month apart, allowing ample time for a trust effect to wear off. But it didn’t. Once again, we see that the real-world implications of this research are important. Eyewitness testimony can be changed when a witness listens to an information source she has previously trusted as credible (media, interrogators, or other people), and this study suggests that the window of opportunity for this effect is large. Any follow-up information received by an eyewitness from any number of sources can significantly alter his or her memory.
False Beliefs: Spawn of False Memories
If there’s anything that cognitive psychology studies have made clear over the years, it’s that humans can be exceptionally gullible. With a little push, we’re prone to developing false beliefs not only about others but also about ourselves with equal prowess—and the results can be, well, hard to believe. And at the core of many of these false beliefs live false memories.
For example, a study in 2001 asked participants to rate the plausibility of having witnessed demonic possession as children and their confidence that they had actually experienced one. Later, the same participants were given articles describing how commonly children witness demonic possessions, along with interviews with adults who claimed to have witnessed possessions as children. After reading the articles, participants not only rated demonic possession as more plausible than they’d previously said, but also became more confident that they themselves had actually experienced demonic possession as children.
Another (less dramatic) study asked participants to rate the plausibility that they’d received barium enemas as children. As with the other study, participants were later presented with “credible” information about the prevalence of barium enemas among children, along with step-by-step procedures for administering an enema. And again, the participants rated the plausibility of having received a barium enema as children significantly higher than they had before.
A study conducted by researcher Stefanie Sharman at the University of South Wales sought to determine the effect of prevalence information (information that establishes how commonly an event happens, making it seem more likely and therefore more self-relevant) on the development of false beliefs. Participants were asked to rate the plausibility of ten events from 1 (“not at all plausible”) to 8 (“extremely plausible”), and how confident they were that they’d experienced each event from 1 (“definitely did not happen”) to 8 (“definitely did happen”). The events included a range of the highly plausible (“I got lost in a shopping mall as a child”) to the highly implausible (“I was abducted by a UFO”).
Two weeks later, participants were brought back and given information on four of the events they’d previously rated, all in the low-to-moderate plausibility range (no UFOs). The information included newspaper articles, third-person descriptions, and data from previous study subjects—all of which were designed to establish higher prevalence of the events. The results showed that high-prevalence information from all sources affected the development of false beliefs. In particular, participants given high-prevalence information in false newspaper articles became more confident that they had actually experienced the events, testifying to the power of the printed word on memory.
How a Memory-Reinforcing Checklist Can Postpone Your Bucket List
A study published in the New England Journal of Medicine about the results of using surgery-safety checklists at major urban hospitals around the world underscores how dangerously fallible memory can be. Around the globe, 234 million operations are performed yearly. It’s difficult to get a handle on the death-rate percentage from postsurgery complications, but I’ve seen estimates anywhere from 1.5 to 5 percent within the first thirty days after surgery. If we take the low end, that’s more than 3.5 million postsurgery deaths a year. Of that number, a significant percentage—perhaps as high as 50 percent—is attributable to infections and complications that could have been prevented if safety procedures had been followed.
This study included eight hospitals in eight cities, including Toronto, New Delhi, Manila, London, and Seattle—a socioeconomic and cultural cross-section of hospitals that participated in the World Health Organization’s Safe Surgery Saves Lives program. As a benchmark, data was collected on 3,733 consecutively enrolled patients sixteen years of age or older who were undergoing noncardiac surgery. After the surgical-safety checklist was introduced, researchers collected data on 3,955 consecutively enrolled patients with the same criteria. Both the death rate and overall complications rate were analyzed for the first thirty days after the operation. The results: The rate of death at the hospitals was 1.5 percent before the checklist was introduced and declined to 0.8 percent afterward. The rate of complications was 11 percent before the checklist, and declined to 7 percent after. In short, using the checklist cut the death and complications rates nearly in half.
This basic tool simply made people remember to follow the safety procedures every time, resulting in fewer complications and fewer patient deaths.
The takeaway here probably has a few prongs. First, we shouldn’t discount the possibility that we’re just as susceptible to developing false beliefs as anyone else walking around on this planet. The brain is a superb miracle of errors and no one, except the brainless, is exempt. On the other hand, knowing that to be true is also the best preventive against chasing the make-believe rabbit down its hole.
Last Word: Total Future Recall
If you stop and think about it, the ability to construct future scenarios in our minds is really quite remarkable. As far as we know, all other species, including our closest primate relatives, react to events as they occur. They can learn from these events and apply that learning in the future (think of chimpanzees learning how to catch ants by poking a stick into the mound and doing the same thing for every new ant mound they find), but they do not assemble complex pieces of information into a coherent whole of the future. How exactly do we accomplish this?
As with many issues in cognitive science, it is difficult to say for certain, but the latest thinking is that we engage in something called episodic future thinking, which means that we simulate the future by using elements from the past. Recent brain imagining studies show that some of the brain regions that are activated when recalling a personal memory—the posterior cingulate gyrus, parahippocampal gyrus, and left occipital lobe—are also active when thinking about a future event.
Our future simulations are obviously not carbon-copy replicas of the past—but we draw on experience to generate the simulations in the same way that a sketch artist uses the pieces of information provided to her to compose an image. Sometimes we get close, other times we are far off; the farther removed the future scenario is from actual experience, the less likely it is to be in the proverbial ballpark.
Perhaps this ability, among other abstract-thinking abilities, confers an adaptive advantage. For the happy brain to make anywhere-near-accurate predictions about our environment, it helps to have access to as much past information as possible to construct multiple scenarios of what could happen. It could be that what we lack in instinctual gumption, like that of other primates, we make up for in imagination. The ability is far from perfect, but it is the most powerful organically based prediction tool yet evolved.