The Memory Illusion

by Dr Julia Shaw

  Essentially, we often make educated guesses about the world as we perceive it, based on our past experiences. When you think about it, this is an essential skill for survival. Since prehistory, humans have often had to make decisions based on limited information. If we were to take the time to thoroughly check what those glimmers that look like lion eyes were, we probably wouldn’t survive to ponder such questions another day. And we certainly do not have enough time to check whether all objects in our environment that look three-dimensional actually are. So we take what we might think of as reasonable interpretive shortcuts. Perception is experienced as a coherent and fluid process only because our brain is constantly making educated guesses, filling in the gaps in information.

  Let’s take first impressions as an example. What happens when we first meet someone? Do we look at them objectively, taking apart each physical feature to identify whether they are friend or foe? No, of course we don’t. In 2013 I published some research relevant to this.6 I was working with three memory scientists at the University of British Columbia, Natasha Korva, Leanne ten Brinke and Stephen Porter, and we were interested in how bias in perception might affect legal proceedings. In particular we wanted to know what makes someone appear trustworthy to others. Trustworthiness, it turns out, is an elusive concept. When we meet someone we may feel we are making an informed decision about whether or not we can trust them. We may think we are making a bottom-up decision, looking at all available evidence and making a rational decision as a result. This is not so. Not at all.

  In our study we gave participants a picture of a ‘suspect’ along with vignette about a crime they had supposedly committed. Then we allowed the participants to be jurors, deciding whether the person was guilty or not. They based their decision only on the photo, the vignette and a set of evidence. We randomly matched the vignette and evidence with a photo; if participants were being impartial they should have based their decision only on the evidence, not on the photo. The photos had already been assessed in a pre-screen, which asked participants to rate on a scale how trustworthy they found each face.

  As we had expected, participants were far harsher on those faces that looked untrustworthy than on those who looked trustworthy. They needed fewer pieces of evidence to come to a guilty verdict for the untrustworthy faces and they were less likely to change their minds when presented with exonerating evidence. Depending on the face of the alleged perpetrator, we got radically different verdicts based on exactly the same evidence. Clearly our participants were relying on something other than the actual evidence, apparently making decisions about a person’s guilt or innocence based on their existing biases.

  Most of the time the automatic involvement of memory to make assumptions and educated guesses is beneficial; it makes us considerably faster at interpreting the various stimuli around us and such guesses are usually accurate. Let’s return to our first impressions of people as an example. Research on what are referred to as ‘thin slices’, meaning brief observations of others that last from a couple of seconds up to five minutes, shows that people are often quite good at inferring certain kinds of character traits.

  Nalini Ambady and colleagues at Stanford University have been conducting studies in this area since 19927 and have demonstrated that people are pretty good at correctly guessing someone’s sexual orientation, teaching performance, and even their ability to deceive others, based only on such thin slices. Unfortunately for our study on how eyewitnesses were influenced by trustworthiness, this ability translated into biased legal decision-making, a context where relying on intuitions and stereotypes can lead to the wrong people going to prison.

  Our memories of previous experiences influence our understanding not only of how we expect people to behave, but more generally of how the world works – gravity, dimensions, possibilities. In the same way that we can be fooled by first impressions of faces, we can be fooled by general perceptual illusions, such as those used by the magician I encountered at the science centre in San Diego. If our perceptions have been in some way misled and we do not realise it, the guessing processes our brain employs to help us understand the world can actually have precisely the opposite effect, and can taint our memories with inaccuracy.

  Aroused

  Are you aroused? On a scale from 1 to 10, how aroused are you? What do you think increases your arousal?

  This may sound like the beginning of a poorly cast, low-budget adult film but it also sounds like some pretty typical memory research. So typical that if you type ‘memory’ and ‘arousal’ into Google Scholar, a version of Google that searches only academic sources, you get over a quarter of a million hits.

  Before you get too excited, though, I need to get your mind out of the gutter. When a researcher says a participant is aroused, they mean that their heart rate, sweating, pupil dilation or other physiological indicators are relatively increased. And it turns out that our level of arousal plays a major part in our ability to encode, store and retrieve memories.

  In one of many similar experiments from the 1990s, Larry Cahill and James McGaugh8 from the University of California wanted to examine this question of how arousal influences memory. In one of their studies, published in 1995,9 they assigned participants to one of two experimental groups – they were either given a neutral story or an emotional story. Participants in each group viewed the same series of images, but with different narrations played on tape to accompany them. Both stories showed a mother taking her young son to see his father at work – but in the neutral version the narration explained that the father was a mechanic repairing a car, while in the emotional version the narration described him as a surgeon working on car crash victims.

  Two weeks later participants were tested on their memory of the story. The researchers found that participants who had experienced the emotional condition could recall an average of 18 details of the event, while those in the neutral condition could recall only 13. In later experiments with a slightly modified method, it was again found that participants in the emotional condition performed better.

  These results seem to make it clear that an increase in arousal is associated with an increase in memory performance. This makes sense if we think back to our most vivid memories, which tend to be of emotional occurrences. It is therefore tempting to jump to the conclusion that more arousal is always better for memory retention. But let’s think about that a bit more carefully. If I were to ask my undergraduate students on a test day whether they agreed with this, I am not sure that they would – being over-aroused or panicked can make us go blank and forget information that otherwise would come to us quite easily. The ‘Ah! But I knew that!’ post-hoc insight after we have forgotten something during an exam is all too familiar to many of us. Similarly, being in an under-aroused state, such as feeling drowsy and lethargic, is also not going to get you far in an exam.

  So we need a more nuanced understanding of memory in relation to arousal. The Yerkes–Dodson theory of performance can help us with this. Developed in 1908 by Robert Yerkes and John Dodson, the theory suggests that performance on any task will improve as arousal increases up to a certain optimum point. However, beyond that point further arousal will actually worsen performance. The suggestion is that at the extremes, at no arousal or incredibly high arousal, a person cannot perform a given task at all. On a graph, this can be represented as an upside-down U – performance initially tracking up in line with increased arousal, then levelling off and decreasing as further arousal becomes detrimental – which is why the theory is called the inverted-U hypothesis.

  The inverted-U hypothesis

  In a demonstration of the inverted-U hypothesis in relation to memory in particular, in 2013 Thomas Schilling10 and his colleagues at the University of Trier in Germany published a study on how the stress hormone cortisol impacts memory performance. Cortisol is released into the bloodstream when acute stress or arousal activates the HPA axis in the brain – the hypothalamic-pituitary-adrenal cortex. The hormone then
crosses into the brain and contributes to the regulation of our stress response, helping to determine how strong and how long it will be.

  Schilling’s team first asked participants to come in to see 18 pictures of male faces accompanied by a brief description of each person, such as ‘he likes to get drunk at parties and then becomes aggressive’. Once the researchers were sure that the participants had learned the associations between the faces and descriptions, they sent them home. A week later, they were asked to come back into the lab. This time they were injected with one of five different levels of cortisol, from none to 24mg. They then had their memories of the associations between the faces and their descriptions tested. The results supported the inverted-U hypothesis, with memory performance increasing up to a moderate level of cortisol and then steadily decreasing after an optimal level was passed.

  So the inverted-U hypothesis does seem a good general model for understanding memory performance in relation to arousal. However, there may not be a one-size-fits-all solution. In a summary of the science behind this, published by the American Psychological Association, memory scientists Mara Mather and Matthew Sutherland from the University of Southern California11 claim that the inverted-U hypothesis does not tell the whole story. They argue this because ‘the findings we reviewed here indicate that emotional arousal makes things that are perceptually salient stand out even more and makes any high priority information even more memorable. At the same time, arousal reduces processing of low priority information. This increase in selectivity under arousal is likely to be adaptive in many situations, and can explain why sometimes arousing stimuli impair memory for nearby stimuli and sometimes they enhance it.’

  In other words, as our arousal increases, our memory focus generally narrows. We become better at remembering critical information about the incident that made us aroused, but we often become worse at remembering contextual information. If we were present at a bank robbery, for example, we will likely be great at remembering that there was a gun pointed at us, but terrible at remembering much else. Unfortunately, arousal states such as fear do not necessarily focus us on things we will need to remember later. In our bank robbery example, it would be far more prudent to remember the faces of the bank robbers rather than the gun, but according to the much-researched ‘weapon-focus effect’ our arousal is going to make it difficult to remember anything other than the gun in much detail.

  Things are further complicated by research demonstrating that individual characteristics such as age, gender and personality may also play a role in how arousal affects memory – the take-home message is that one size does not fit all when we talk about exactly how memory and arousal are intertwined.

  I’ve saved my favourite association between arousal and memory for last. One way to use this association between arousal and memory to our advantage is to appreciate what is called state-dependent memory, a phenomenon that has been repeatedly demonstrated and validated. What it means is that we generally remember things better if we are in the same state during the recall of a memory as during the encoding of it.

  In 1990, Shirley Pearce and her colleagues at University College London12 demonstrated this in two extremely interesting experiments. The first involved giving both chronic-pain patients and non-patients who were not suffering any pain a list of words to remember. They found that those who reported suffering chronic pain were much better at remembering words associated with pain than any other kinds of words. This is in line with the idea that ‘mood congruency’ matters; that we are better at encoding and retrieving information that fits with our mood. But this is not enough, as the exact state we are in can change, so Pearce and her team wanted to know whether a temporary state can also influence our memory.

  To do this they inflicted pain on some of their participants by asking them to submerge their hands in ice water, while others got pleasantly warm water. If you have never held your hand in ice water for any length of time, this is a surprisingly terrible experience. Right after this water bath experience, the participants were given a list of words to remember. They were then either given another painful ice-water bath, or were asked to put their hands in warm water, before being tested on their memory of the list.

  The researchers found that when participants experienced the same state at the time of learning as at recall, they performed significantly better. So those who experienced pain right before they learned the word list performed better if they experienced pain again right before they needed to recall the information. Similarly, those who experienced nice warm water before learning did better at the memory test if they had just placed their hands in warm water again. If we follow this through, it means that if we know we learned or experienced something during a particular type of arousal, by recreating that state we should be able to remember it better. Torturous ice-water baths not your thing? Here’s a more pleasant example: if you always drink a cup of coffee right before you study, your memory should be better if you drink a cup of coffee right before your exams.

  All this research clearly shows that our stress and arousal levels matter for what we are able to store as memories, and how we are able to retrieve them later on. So our memories can be affected not only by uncontrollable parts of our external environment, but also by largely uncontrollable elements of our internal environment.

  Time travellers

  Also dependent on our arousal and emotional state is our perception of time. As we all know, the more aroused or excited we are the faster time seems to pass. Popular sayings like ‘time flies when you are having fun’ or ‘this is like watching paint dry’ suggest that how much we are engaged by an activity has a clear effect on how we remember it temporally.

  Think about it. How long do you think it took you to read the last paragraph? Did it take a long time or a short time? How about assigning a specific value? Ten seconds? One minute? How accurate do you think you are? What do you think your answer is based on – how do you know how long it took?

  Of course, we do not typically ask ourselves these kinds of questions, and mostly take our perception of time for granted. We too often seem to think that we have some sort of magical internal clock that keeps time for us in a relatively objective manner. However, if we think about those situations in which time seemed to drag on forever during an unpleasant task, or sped by because we were so excited about something, we know that nothing could be further from the truth.

  Sometimes referred to as the fourth dimension – an extension of our 3D physical reality – time is something that could be considered a primarily internal phenomenon. It is characterised by linearity, sequentiality and change; by growth or destruction. Our subjective perception of time is known as chronesthesia,13 and it is studied by researchers from fields as diverse as neurophysiology, psychology and philosophy. And what all of these scientific disciplines have demonstrated is that, perhaps unsurprisingly, memory is vital for our ability to perceive time.

  One line of research argues that the way we perceive the passage of time is through our sense of chronology. In other words, we remember the order in which events happened, which then allows us to infer when and for how long an event took place. Obviously to do this, we need to remember what things happened, and in which order. Time is memory; memory is time.

  Nobel laureates and behavioural economists Daniel Kahneman and Amos Tversky have done a great deal of work on how we estimate and value time, particularly focusing on things from a memory perspective. They would say that many individuals, particularly those with time prediction problems, engage in the ‘planning fallacy’,14 which means they overly focus on what they refer to as ‘singular’ information, information that is associated with a single task.

  For example, if you were a doctor trying to predict how long an Alzheimer’s patient will live, the relevant singular information would include how old the patient is, how sick they are, and what their personal medical history is. These are all important pieces of information, but they only really bec
ome useful when we place them within the context of ‘distributional’ information. Distributional information refers to a wider set of information, including in general how long, on average, 70-year-old patients with Alzheimer’s live. The singular information helps you see how this patient may be different from others, their unique risk factors, while the distributional information helps you use that information to make predictions based on the averages seen in others with those kinds of characteristics.

  Distributional information is, of course, reliant on your ability to remember and access the past occasions when you had patients like this, and when you learned that the average life expectancy for Alzheimer’s is eight to ten years. Having the ability to contextualise singular information within our distributional information in this way greatly improves our ability to make accurate predictions about how long things will take in the future (or, in our example, how long a patient will survive).

  We all have that friend (or family member, or co-worker) who, when they organise events or plan their day, engages in deeply flawed time estimation. The ‘Oh, it’ll only take me five minutes to get there!’ kind of person. We might say such people are ‘optimistic’ in their time estimates. But we might also say that they are potentially bad at remembering how long it actually took them to do things in the past. They are worse at using their distributional information to ask ‘How long does it normally take me to get there?’ Sure, Google Maps says five minutes, but that is not accounting for fixing their hair, finding their keys, putting on a coat, going down four flights of stairs, and finding the right doorbell upon arrival. A memory science view of this prediction inadequacy suggests that some people may be late because their memory and time perception systems give them both a poor sense of their past experiences and poor ‘prospective’ memory – the ability to plan the future based on past experiences.