|chapter 8||Remembering Complex
Memory Errors, Memory Gaps Where did you spend last summer? What country did you grow up in? Where were you five minutes ago? These are easy questions, and you effortlessly retrieve this information from memory the moment you need it. If we want to understand how memory functions, therefore, we need to understand how you locate these bits of information (and thousands of others just like them) so readily.
But we also need to account for some other observations. Sometimes, when you try to remember an episode, you draw a blank. On other occasions, you recall something, but with no certainty that you’re correct: “I think her nickname was Dink, but I’m not sure.” And sometimes, when you do recall a past episode, it turns out that your memory is mistaken. Perhaps a few details of the event were different from the way you recall them. Or perhaps your memory is completely wrong, misrepresenting large elements of the original episode. Worse, in some cases you can remember entire events that never happened at all! In this chapter, well consider how, and how often, these errors arise. Let’s start with some examples. Memory Errors: Some Initial Examples In 1992, an El Al cargo plane lost power in two of its engines just after taking off from Amsterdam’s Schiphol Airport. The pilot attempted to return the plane to the airport but couldn’t make it. A few minutes later, the plane crashed into an 11-story apartment building in Amsterdam’s Bijlmermeer neighborhood. The building collapsed and burst into flames; 43 people were killed, including the plane’s entire crew.
Ten months later, researchers questioned 193 Dutch people about the crash, asking them in particular, “Did you see the television film of the moment the plane hit the apartment building?” More than half of the participants (107 of them) reported seeing the film, even though there was no such film. No camera had recorded the crash; no film (or any reenactment) was shown on television. The participants seemed to be remembering something that never took place (Crombag, Wagenaar, & van Koppen, 1996).
In a follow-up study, investigators surveyed another 93 people about the plane crash. These people were also asked whether they’d seen the (nonexistent) TV film, and then they were asked detailed questions about exactly what they had seen in the film: Was the plane burning when it crashed, or did it catch fire a moment later? In the film, did they see the plane come down vertically or did it hit the building while still moving horizontally at a considerable with no forward speed speed? Two thirds of these participants reported seeing the film, and most of them were able to provide details about what they had When asked about the plane’s speed, for example, only 23% said that they couldn’t remember. The others gave various responses, presumably based on their “memory” of the (nonexistent) film.
Other studies have produced similar results. There was no video footage of the car crash in which Princess Diana was killed, but 44% of the British participants in one study recalled seeing the footage (Ost, Vrij, Costall, & Bull, 2002). More than a third of the participants questioned about a nightclub bombing in Bali recalled seeing a (nonexistent) video, and nearly all these participants reported details about what they’d seen in the video (Wilson & French, 2006).
It turns out that more persistent questioning can lead some of these people to admit they actually don’t remember seeing the video. Even with persistent questioning, though, many participants continue to insist that they did see the video-and they offer additional information in the film (e.g., Patihis & Loftus, 2015; Smeets et al., 2006). Also, in all about exactly what they sav these studies, let’s emphasize that participants are thinking back to an emotional and much- discussed event; the researchers aren’t asking them to recall a minor occurrence.
Is memory more accurate when the questions come after a shorter delay? In a study by Brewer and Treyens (1981), participants were asked to wait briefly in the experimenter’s office prior to the procedure’s start. After 35 seconds, participants were taken out of this office and told that there actually was no experimental procedure. Instead, the study was concerned with their memory for the room in which they’d just been sitting. Participants’ descriptions of the office were powerfully influenced by their prior beliefs. Surely, most participants would expect an academic office to contain shelves filled with books. In this particular office, though, no books in view (see Fiqure 8.1). Even so, almost one third of the participants (9 of 30) reported seeing books in the office. Their recall, in other words, was governed by their expectations, not by reality. How could this happen? How could so many Dutch participants be wrong in their recall of the plane crash? How could intelligent, alert college students fail to remember what they’d seen in an office just moments earlier? Memory Errors: A Hypothesis In Chapters 6 and 7, we emphasized the importance of memory connections that link each bit of knowledge in your memory to other bits. Sometimes these connections tie together similar episodes, so that a trip to the beach ends up connected in memory to your recollection of other trips. Sometimes the connections tie an episode to certain ideas-ideas, perhaps, that were part of your understanding of the episode, or ideas that were triggered by some element within the episode.
It’s not just separate episodes and ideas that are linked in this way. Even for a single episode, the elements of the episode are stored separately from one another and are linked by connections. In fact, the storage is “modality-specific,” with the bits representing what you saw stored in brain areas devoted to visual processing, the bits representing what you heard stored in brain areas specialized for auditory processing, and so on (e.g., Nyberg, Habib, McIntosh, & Tulving, 2000; Wheeler Peterson, & Buckner, 2000; also see Chapter 7, Figure 7.4, p. 245).
With all these connections in place-element to element, episode to episode, episode to related ideas-information ends up stored in memory in a system that resembles a vast spider web, with was the each bit of information connected by many threads to other bits elsewhere in the web. This idea that in Chapter 7 we described as a huge network of interconnected nodes. However, within this network there are no boundaries keeping the elements of one episode separate from elements of other episodes. The episodes, in other words, aren’t stored in separate “files,” each distinct from the others. What is it, therefore, that holds together the various bits within each episode? To a large extent, it’s simply the density of connections. There are many connections linking the various aspects of your “trip to the beach” to one another; there are fewer connections linking this event to other events.
As we’ve discussed, these connections play a crucial role in memory retrieval. Imagine that you’re trying to recall the restaurant you ate at during your beach trip. You’ll start by activating nodes in memory that represent some aspect of the trip-perhaps your memory of the rainy weather. Activation will then flow outward from there, through the connections you’ve established, and this will energize nodes representing other aspects of the trip. The flow of activation can then continue from there, eventually reaching the nodes you seek. In this way, the connections serve as retrieval paths, guiding your search through memory.
Obviously, then, memory connections are a good thing; without them, you might never locate the information you’re seeking. But the connections can also create problems. As you add more and more links between the bits of this episode and the bits of that episode, you’re gradually knitting these two episodes together. As a result, you may lose track of the “boundary” between the episodes. More precisely, you’re likely to lose track of which bits of information were contained within which event. In this way, you become vulnerable to what we might think of as “transplant” errors, in which a bit of information encountered in one context is transplanted into another context. In the same way, as your memory for an episode becomes more and more interwoven with other thoughts you’ve had about the event, it will become difficult to keep track of which elements are were actually part of the episode itself, and which are linked merely because they were associated with the episode in your thoughts. This, too, can produce linked to the episode because they transplant errors, in which elements that were part of your thinking get misremembered as if they were actually part of the original experience. Understanding Both Helps and Hurts Memory It seems, then, that memory connections both help and hurt recollection. They help because the connections, serving as retrieval paths, enable you to locate information in memory. But connections can hurt because they sometimes make it difficult to see where the remembered episode stops and other, related knowledge begins. As a result, the connections encourage intrusion errors-errors in which other knowledge intrudes into the remembered event.
To see how these points play out, consider an early study by Owens, Bower, and Black (1979). In this study, half of the participants read the following passage:
Nancy arrived at the cocktail party. She looked around the room to see who was there. She went to talk with her professor. She felt she had to talk to him but was a little nervous about just what to say. A group of people started to play charades. Nancy went over and had some refreshments. The hors d’oeuvres were good, but she wasn’t interested in talking to the rest of the people at the party. After a while she decided she’d had enough and left the party.
Other participants read the same passage, but with a prologue that set the stage:
Nancy woke up feeling sick again, and she wondered if she really was pregnant. How would she tell the professor she had been seeing? And the money was another problem.
All participants were then given a recall test in which they were asked to remember the sentences as exactly as they could. Table 8.1 shows the results-the participants who had read the prologue (the Theme condition) recalled much more of the original story (i.e., they remembered the propositions actually contained within the story). This is what we should expect, based on the claims made in Chapter 6: The prologue provided a meaningful context for the remainder of the story, and this helped understanding. Understanding, in turn, promoted recall.
At the same time, the story’s prologue also led participants to include elements in their recall that weren’t mentioned in the original episode. In fact, participants who had seen the prologue made four times as many intrusion errors as did participants who hadn’t seen the prologue. For example, they might include in their recall something like “The professor had gotten Nancy pregnant.” This idea isn’t part of the story but is certainly implied, so will probably be part of participants’ understanding of the story. It’s then this understanding (including the imported element) that is remembered. The DRM Procedure Similar effects, with memory connections both helping and hurting memory, can be demonstrated with simple word lists. For example, in many experiments, participants have been presented with lists like this one: “bed, rest, awake, tired, dream, wake, snooze, blanket, doze, slumber, snore, nap, peace, yawn, drowsy.” Immediately after hearing this list, are asked to recall as many of participants the words as they can.
As you surely noticed, the words in this list are all associated with sleep, and the presence of this theme helps memory: The list words are easy to remember. It turns out, though, that the word “sleep” is not itself included in the list. Nonetheless, research participants spontaneously make the connection between the list words and this associated word, and this connection almost always leads to a memory error. When the time comes for recall, participants are extremely likely to recall that they heard “sleep.” In fact, they’re just as likely to recall “sleep” as they are to recall the actual words on the list (see Figure 8.2). When asked how confident they are in their memories participants are just as confident in their (false) recall of “sleep” as they are in their (correct) memory of genuine list words (Gallo, 2010: for earlier and classic papers in this arena. see Deese, 1957; Roediger & McDermott, 1995, 2000). This experiment (and many others like it) uses the DRM procedure, a bit of terminology that honors the investigators who developed it (James Deese, Henry Roediger II, and Kathleen McDermott). The procedure yields many errors even if participants are put on their guard before the procedure begins-that is, told about the nature of the lists and the frequency with which they produce errors (Gallo, Roberts, & Seamon, 1997; McDermott & Roediger, 1998). Apparently, the mechanisms leading to these errors are so automatic that people can’t inhibit them. Schematic Knowledge Imagine that you go to a restaurant with a friend. This setting is familiar for you, and you have some commonsense knowledge about what normally happens here. You’ll be seated; someone will bring menus; you’ll order, then eat; eventually, you’ll pay and leave. Knowledge like this is often referred to with the Greek word schema (plural: schemata). Schemata summarize the broad pattern of what’s normal in a situation-and so your kitchen schema tells you that a kitchen is likely to have a stove but no piano; your dentist’s office schema tells you that there are likely to be magazines in the waiting room, that you’ll probably get a new toothbrush when you leave, and so on.
Schemata help you in many ways. In a restaurant, for example, you’re not puzzled when someone keeps filling your water glass or when someone else drops by to ask, “How is everything?” Your schema tells you that these are normal occurrences in a restaurant, and you instantly understand how they fit into the broader framework. Schemata also help when the time comes to recall how an event unfolded. This is because there are often gaps in your recollection-either because you didn’t notice certain things in the first place, or because you’ve gradually forgotten some aspects of the experience. (We’ll say more about forgetting later in the chapter.) In either case, you can rely on your schemata to fill in these gaps. So, in thinking back to your dinner at Chez Pierre, you might not remember anything about the menus. Nonetheless, you can be reasonably sure that there were menus and that they were given to you early on and taken away after you placed your order. On this basis, you’re likely to include menus within your “recall” of the dinner, even if you have no memory of seeing the menus for this particular meal. In other words, you’ll supplement what you actually remember with a plausible reconstruction based on your schematic knowledge. And in most cases this after-the-fact reconstruction will be correct, since schemata do, after all, describe what happens most of the time. Evidence for Schematic Knowledge Clearly, then, schematic knowledge helps you, by guiding your understanding and enabling you to reconstruct things you can’t remember. But schematic knowledge can sometimes hurt you, by promoting errors in perception and memory. Moreover, the types of errors produced by schemata are quite predictable. As an example, imagine that you visit a dentist’s office, and this one happens not to have any magazines in the waiting room. It’s likely that you’ll forget this detail after a while, so what will happen when you later try to recall your trip to the dentist? Odds are good that you’ll rely on schematic knowledge and “remember” that there were magazines (since, after all, there usually are some scattered around a waiting room). In this way, your recollection will make this dentist’s office seem more typical, more ordinary, than it actually was. Here’s the same point in more general terms. We’ve already said that schemata tell you what’s typical in a setting. Therefore, if you rely on schematic knowledge to fill gaps in your recollection, you’ll fill those gaps with what’s normally in place in that sort of situation. As a result, any reliance on schemata will make the world seem more “normal” than it really is and will make the past seem more “regular” than it actually was.
This tendency toward “regularizing” the past has been documented in many settings. The classic demonstration, however, comes from studies published long ago by British psychologist Frederick Bartlett. Bartlett presented his participants with a story taken from the folklore of Native Americans (Bartlett, 1932). When tested later, the participants did reasonably well in recalling the gist of the story, but they made many errors in recalling the particulars. The pattern of errors, though, was quite systematic: The details omitted tended to be ones that made little sense to Bartlett’s British participants. Likewise, aspects of the story that were unfamiliar were often changed into aspects that were more familiar; steps of the story that seemed inexplicable were supplemented to make the story seem more logical.
Overall, then, the participants’ memories seem to have “cleaned up” the story they had read- making it more coherent (from their perspective), more sensible. This is exactly what we would expect if the memory errors derived from the participants’ attempts to understand the story and with that, their efforts toward fitting the story into a schematic frame. Elements that fit within the frame remained in their memories (or could be reconstructed later). Elements that didn’t fit dropped out of memory or were changed. the same spirit, consider the Brewer and Treyens study mentioned at the start of this chapter- the study in which participants remembered seeing shelves full of books, even though there were none. This error was produced by schematic knowledge. During the event itself (while the participants were sitting in the office), schematic knowledge told the participants that academic offices usually contain many books, and this knowledge biased what the participants paid attention to. (If you’re already certain that the shelves contain books, why should you spend time looking at the shelves? This would only confirm something you already know-see Vo & Henderson, 2009.) Then, when the time came to recall the office, participants used their schema to reconstruct what the office must have contained- a desk, a chair, and of course lots of books. In this way, the memory for the actual office was eclipsed by generic knowledge about what a “normal” academic office contains.
Likewise, think back to the misremembered plane crash and the related studies of people remembering videos of other prominent events, even though there were no videos of these events. Here, too, the memory errors distort reality by making the past seem more regular, more typical, than it really was. After all, people often hear about major news events via a television broadcast or Internet coverage, and these reports usually include vivid video footage. So here, too, the past as remembered seems to have been assimilated into the pattern of the ordinary. The event as it unfolded was unusual, but the event as remembered becomes typical of its kind-just as we would expect if understanding and remembering were guided by our knowledge of the way things generally unfold. e. Demonstration 8.1: Associations and Memory Error This is a test of immediate memory. Read List 1; then close the list and try to write down, from memory, as many words as you can remember from the list. Then expand the list to read List 2, close the list, and try to write down as many of its words as you can remember. Then do the same for List 3. When you’re all done, read the material that follows.
List 1 List 2 List 3
Door Nose Sour
Glass Breathe Candy
Pane Sniff Sugar
Shade Aroma Bitter
Ledge Hear Good
Sill See Taste
House Nostril Tooth
Open Whiff Nice
Curtain Scent Honey
Frame Reek Soda
View Stench Chocolate
Breeze Fragrance Heart
Sash Perfume Cake
Screen Salts Tart
Shutter Rose Pie Don’t read beyond this point until you’ve tried to recall each of the three lists!
Each of these lists is organized around a theme, but the word that best captures that theme included in the list. All of the words in List 1, for example, are strongly associated with the word “window” but that word is not in the list. All of the words in List 2 are strongly associated with “smell” and all in List 3 are strongly associated with “sweet”; but again, these theme words are not in the lists. In your recall of the lists, did you include seeing “window” in List 1? “Smell” in List 2? “Sweet” in List 3?
This procedure, described in the chapter, is called the DRM procedure, in honor of the researchers who have developed this paradigm (Deese, Roediger, and McDermott). IIn this situation, often as many as half of the people tested do make these specific errors-and with considerable confidence. Of course, the theme words are associated with the list in your memory, and it’s this association that leads many people into a memory error.
Perhaps you read through the material after reading the text’s description of the DRM procedure. Did you make the expected error anyway? Research suggests that these errors appear even when research participants are warned about the DRM pattern, just as you were. Did you show that pattern? Or did you manage to avoid the errors?
Demonstration adapted from McDermott, K., & Roediger, H. (1998). False recognition of associates can be resistant to an explicit warning to subjects and an immediate recognition probe. Journal of Memory and Language, 39, 508-520. See also Roediger, H., &McDermott, K. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory and Cognition, 21(4), 803-814. The Cost of Memory Errors There’s clearly a “good news, bad news” quality to our discussion so far. On the positive side, memory connections serve as retrieval paths, allowing you to locate information in storage. The connections also enrich your understanding, because they tie each of your memories into a context provided by other things you know. In addition, links to schematic knowledge enable you to supplement your perception and recollection with well-informed (and usually accurate) inference.