Psychology Of Reconstructive Memory Research Paper

1. Background

The hypothesis that remembering should be viewed as reconstructive dates to an important book by Sir Frederic Bartlett (1932). Bartlett contrasted reproductive memory (veridical, rote forms of memory, such as reproducing a telephone number) with reconstructive memory and argued that the latter was more typical of our uses of memory outside laboratory and educational circumstances. That is, we are rarely faced with the task of remembering something exactly the way it happened, but more typically need only to get the essence of the event right. If a friend asks you, ‘What did Kathleen tell you last night?’ the request is not for a literal rendering of last night’s conversation, but rather for the gist of what was said.

Bartlett argued that perceiving and comprehending events do not simply happen automatically, but that every event of comprehension involves the mental construction of one’s understanding of the event in the world. We all struggle with the ‘effort after meaning’ in comprehending the events in the world around us. If encoding or perceiving is a construction, then when one wants to recall the events later, the attempt is to reconstruct the event.

Bartlett argued that recollection is guided by schemas, or general organizing structures, which aid encoding and retrieval. If you see a scene at the beach and are asked to recall it later, you might recall seeing a beach umbrella even if none was present in the actual scene itself, because it is consistent with the general schema of items that belong in a beach scene. Therefore, although schema can aid encoding and retrieval of information, they can also lead to errors.

A large amount of research is consistent with the idea that remembering is reconstructive. The standard textbook account holds that certain forms of remembering are reconstructive whereas others are reproductive. However, a strong case can be made that all remembering is reconstructive. What appears to be reproductive memory occurs in situations in which the reconstruction is quite accurate (Roediger and McDermott 1995). This research paper considers various forms of memory as they are experimentally studied and discusses evidence for reconstructive processes at work.

2. Sensory And Short-Term Memory

The greatest challenge to the claim in the previous paragraph perhaps comes from studies of sensory memory, the borderline between perceiving and remembering. However, many theorists have argued that perceiving is itself a constructive activity, with fascinating phenomena of visual and auditory illusions used as evidence for this claim (see Gregory 1998, among many others). Sensory memory refers to the temporary persistence of information that has struck the senses, which lingers briefly as it is being comprehended. Visual persistence is called iconic memory and auditory persistence is labeled echoic memory. It would seem that iconic memory—essentially a fleeting after-image of the scene from the outside world- —would surely be a form of reproductive memory. Yet even in these situations errors arise, showing that at least the retrieval processes from this type of memory involve reconstruction. For example, in Sperling’s studies of iconic storage, in which people had to report letters that they had briefly seen on a screen, a common error when people missed a letter was to report another letter than either looked like or sounded like the original letter (Sperling 1967). This type of error indicates that people may code even such simple items as single letters into visual patterns and associated sounds. When people miss a b in these experiments, they may substitute a (which sounds like b) or a p or d, which share both similar appearance (a long line and a curve) and similar sound (the letters rhyme with b). Even reports from iconic memory may show reconstructive tendencies.

Short-term memory or working memory (see related entries on these topics) hold information for longer than sensory memories, but people are still ordinarily accurate in retrieving information from short-term stores, if no interference occurs. Does this accuracy reflect a rote, reproductive process? The answer seems to be no, because when the short-term memory system is challenged by having people operate under, for example, fast rates of presentation, errors occur. Errors are often (but not always) phonological in nature. That is, if someone tries to recall letter strings and misses a letter, similar sounding letters are confused (Conrad 1964). In the case of words, those that share visual and phonemic (sound-based) features are confused (Crowder 1976, Chap. 4). Therefore, even though short-term memory processes are often considered quite accurate (and they can be), recall in these situations typically occurs under conditions (short unfilled delays between study and test) that make for accurate reconstructions. Stress the system by presenting material fast, or by creating interference, and the characteristic error pattern indicative of a reconstructive process appears.

3. Long-Term Memory

By far the greatest amount of research on reconstructive memory has occurred in the domain of longterm, episodic memory. Many different kinds of experiments have been conducted and we will consider some of the main types here, which have used a wide variety of materials: word lists, pictures and scenes, sentences, prose passages, and videotape. Research with all these materials reveals remembrance to be quite prone to error and to filling in the gaps.

3.1 Word Lists

Roediger and McDermott (1995) developed a paradigm in which people heard lists of 15 words that were generated as associates to another word, which was not itself presented. So, for example, subjects heard the list ‘hot, snow, warm, winter, ice, wet, frigid, chilly, heat, weather, freeze, air, shiver, Arctic, frost.’ All these words are associated to the word ‘cold,’ which was the generator of the list, but ‘cold’ was not presented. Immediately after hearing the list, subjects were asked to recall the items and were urged to be sure that the words they recalled had actually been presented in the list. Despite this warning not to guess, Roediger and McDermott (1995) found that their subjects recalled missing words like ‘cold’ at very high levels; in fact, probability of recalling the critical item often equals probability of recall of list items in this paradigm. After recall of many lists, subjects received a recognition test in which they were instructed to pick out only items that had actually appeared in the list. The test included words from the list (‘winter,’ ‘chilly’), completely unrelated words (‘automobile’), and the critical words such as ‘cold.’ Subjects examined each word and, if they judged it to be old (or studied), they were asked to make a further remember know judgment. This judgment (Tulving 1985) asks people to report if they can remember the moment of the item’s occurrence in the study list (a remember judgment) or if they just know it was on the list but cannot remember its moment of occurrence (a know judgment).

The recognition results are shown in Fig. 1, where it can be seen that subjects were good at distinguishing studied words from unrelated lures or distracters. However, the critical lures such as ‘cold’ were recognized at the same level as were the words in the list, about 80 percent. People recognize the related but nonstudied word at the same level as they do items in the list. In addition, subjects claimed to remember the moment of occurrence of the critical words (that, of course, were not actually presented) at the same level as for the words that were actually studied). In short, in this paradigm subjects recall, recognize, and claim to remember the occurrence of words that were never presented at the same level as of the words that really were presented, which provides a compelling demonstration of illusory memory. The implication is that subjects reconstruct the list using their general knowledge and yet insert a quite specific memory—they remember ‘cold’ as having occurred—because it fits with the general theme or schema of the list.

3.2 Pictures And Scenes

Miller and Gazzaniga (1998) developed a similar procedure with common pictorial scenes, ones that had been covers of the Saturday Evening Post magazine. They showed people the scenes with a critical object from the scene removed. To use our earlier example, people might view a beach scene with a beach umbrella either present in the scene or absent from it. Miller and Gazzaniga (1998) found that when people were given recognition tests on objects in the scene, they were very likely to recognize the object as having been present when in fact it was not. Again, in reconstructing the scene, the object is inserted and seems part of the memory.

3.3 The Misinformation Paradigm

One of the most popular paradigms for studying reconstructive processes in memory was developed by Elizabeth Loftus and her colleagues (e.g., Loftus et al. 1978) and is called the misinformation paradigm. It is modeled after the events that an eyewitness to a crime might experience. In the lab, subjects view a set of slides or a videotape, then later they receive questions or a narrative. In Loftus et al. (1978), subjects witnessed an automobile accident that occurred at an intersection where a car failed to stop at a stop sign. In the later series of questions, subjects would be asked either one of three types of question: ‘Did the car stop at the traffic sign?’ (control condition), ‘Did the car stop at the stop sign?’ (consistent information condition) or ‘Did the car stop at the yield sign?’ (inconsistent or misleading information condition). This question served as the experimental manipulation for a question asked at a later point in time: ‘What kind of sign was at the intersection, stop or yield?’

Results of this experiment are shown in Fig. 2, where it can be seen that, relative to the control condition in which the question was posed in a neutral manner, wording of the question that was consistent with the original information improved recognition whereas the conflicting information reduced recognition. This last finding is the misinformation effect —wrong information received after an event is often incorporated into one’s memory for the event. This misinformation effect has been demonstrated in hundreds of experiments, although the magnitude of the effect depends on the exact nature of the test.

The general principle to arise from experiments on the misinformation effect is that memory for an event is not encapsulated in time the way the event itself is. Rather, information provided after the event can modify our memories for the event itself. Consistent information improves our later reconstruction, whereas conflicting or misleading information is harmful. When we are trying to reconstruct events from our past, we draw on all the information and resources at our disposal and the reported memory may include ‘facts’ that were not really part of the event, but incorporated into it later. Practically, if police or other people ask suggestive questions or embed wrong information into questions asked of witnesses after the event, then the witnesses’ later recollections and testimony might be tainted.

3.4 Prose Recall

Perhaps the greatest body of research demonstrating reconstructive processes in retrieval has been carried out in prose recall. Brewer (1977) examined pragmatic implications in retention of sentences. A pragmatic implication is made when the person reading or hearing a sentence infers something that is neither explicitly stated nor logically implied by the sentence. For example, ‘The karate champion hit the cinder block’ implies that the block was broken. However, this is a pragmatic implication; it is perfectly possible that the block was struck but not broken. After people studied sentences of this type mixed with other sentences, they were given the first parts of the sentences as cues to recall the entire sentence. Brewer showed that in recall of sentences with pragmatic inferences, people were more likely to recall the implied verb than the verb actually used in the sentence. When reconstructing the action, they remember the champion as breaking the cinder block, not just hitting it. This result reflects a kind of inferential momentum, with people going beyond the literal truth of a statement to infer the events that probably occurred next. Often such inferences may be correct, but they are not necessarily so. When the inference is strongly made, but false, it can generate a false memory.

Sulin and Dooling (1974) also studied people’s remembrance of implications as facts. In one case, they presented subjects with short paragraphs about a troubled girl and later tested for recognition both of ideas from the paragraph and of nonpresented ideas that were either consistent or inconsistent with the theme of the passage. The paragraphs studied by two groups of subjects were identical, with the exception of the name above the story. In one case it was Helen Keller and in the other case it was Carol Harris. One item on the recognition test asked subjects if the following sentence had appeared in the narrative: ‘She was deaf, dumb, and blind.’ Students who had read the story about Helen Keller frequently made the error of saying that this sentence had appeared in the passage, whereas those reading about Carol Harris almost never made this error. In trying to reconstruct whether the sentence may have been in the story, the knowledge about Helen Keller caused subjects to fill in what ‘must have been’ in the story and led subjects erroneously to conclude that the sentence was there.

Sometimes people remember the sentences that were in a prose passage no better than sentences they have never read before, so long as the new sentences are consistent with the material that was actually presented. Bransford and Franks (1971) gave subjects short sentences such as ‘The rock rolled down the hill’ or ‘The rock crushed the hut.’ There were four simple ideas like this that, when put together, would make up the complex sentence: ‘The rock rolled down the hill and crushed the tiny hut.’ The experimenters gave their subjects various sentences to build up the complex idea unit. They later tested subjects by asking them to pick out sentences they had actually studied from among other sentences. Two types of distractor sentences were used: some were consistent with the idea unit but had not been studied, whereas other sentences were unrelated to the idea unit. The test sentences themselves could vary in the number of idea units present, from one to four. Subjects rated their belief that the sentences had been previously studied on a 10-point scale, from 5 (sure the sentence was new or nonstudied) to 5 (sure the sentence was old or studied). The sentences that were not consistent with the idea units were accurately rejected as not having been studied.

The results for the other two types of items are shown in Fig. 3 and reveal that the more idea units embedded in the test sentence, the higher the confidence rating that the sentence had been studied. Interestingly, this relation held both for sentences that had been studied and for new, nonstudied, sentences that were consistent with the overarching idea unit. Remarkably, except for sentences expressing only one idea, subjects were equally confident in the sentences that had not been presented in the study phase as for those that had been presented! Therefore, consistent with Bartlett’s ideas, people seemed to retain the meaning of the sentences, but did not retain the literal wording of the sentences.

Bartlett (1932), who began this tradition of work, was interested in repeated recollection of material. He presented subjects with a native American folktale, ‘The War of the Ghosts,’ which he had them read twice. The story is a bit disjointed, has supernatural elements, and is rather difficult to understand. He found that subjects frequently made errors in recalling the story and that these errors increased over repeated retellings of the story. The stories got shorter, reflecting forgetting, but more importantly for present purposes, the story changed in systematic ways by subjects introducing errors that served to rationalize its bizarre elements. For example, people might ‘remember’ elements that had not been present to explain some of the unusual features in the story. Interestingly, this tendency to rationalization—to reconstruct the story more in line with the rememberers’ own schema— seemed to grow over time and repeated retellings. Although Bartlett’s original studies were rather informal, the work has been replicated under more systematic conditions (Bergman and Roediger 1999).

4. Conclusion

This research paper has reviewed selectively some key evidence that permits us to draw the conclusion that remembering the past should be viewed as reconstructing it. To draw on a metaphor from Hebb (1949), we can think of the process of remembering the past as we can conceive of paleontologists’ reconstruction of a dinosaur from bone fragments and chips. The archeologist recovers a partial skeleton, but the finished product in a museum is shown as complete, with new bones added, old ones refinished or enhanced, and the entire skeleton constructed based on knowledge of what the animal must have looked like. Similarly, Hebb (1949) argued, remembering the past involves recollection of specific facts and details (the bones) that are woven together into a complete story of the event (like the skeleton). However, the story about the event might involve considerable constructive activity on the rememberer’s part and the rememberer’s present knowledge and goals may shape and determine how he or she remembers the past. A quote from Bartlett captures the essence of this reconstructive approach:

… the one overwhelming impression produced by this more ‘‘realistic’’ type of memory experiment is that human remembering is normally exceedingly subject to error. It looks as if what is said to be reproduced is, far more generally than is commonly admitted, really a construction, serving to justify whatever impression may have been left by the original. It is this ‘‘impression,’’ rarely defined as with much exactitude, which most readily persists. So long as the details which can be built up around it are such that they would give it a ‘‘reasonable’’ setting, most of us are fairly content, and we are apt to think that what we have built we have literally retained (Bartlett 1932, pp. 175–6).

Bibliography:

1. Bartlett F C 1932 Remembering: A Study in Experimental and Social Psychology. Cambridge University Press, Cambridge, UK
2. Bergman E T, Roediger H L 1999 Can Bartlett’s repeated reproduction experiments be replicated? Memory & Cognition 27: 937–47
3. Bransford J D, Franks J J 1971 The abstraction of linguistic ideas. Cognitive Psychology 2: 331–50
4. Brewer W F 1977 Memory for the pragmatic implications of sentences. Memory and Cognition 5: 673–8
5. Conrad R 1964 Acoustic confusions in immediate memory. British Journal of Psychology 55: 75–84
6. Crowder R G 1976 Principles of Learning and Memory. Erlbaum, Hillsdale, NJ
7. Gregory R L 1998 Eye and Brain. Oxford University Press, Oxford, UK
8. Hebb D O 1949 The Organization of Behavior: A Neuropsychological Theory. Wiley, New York
9. Loftus E F, Miller D G, Burns H J 1978 Semantic integration of verbal information into a visual memory. Journal of Experimental Psychology: Human Learning and Memory 4: 19–31
10. Miller M B, Gazzaniga M S 1998 Creating false memories for visual scenes. Neuropsychologia 36: 513–20
11. Roediger H L, McDermott K B 1995 Creating false memories: remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition 21: 803–14
12. Sperling G 1967 Successive approximations to a model for short-term memory. Acta Psychologica 27: 285–92
13. Sulin R A, Dooling D J 1974 Intrusion of a thematic idea in the retention of prose. Journal of Experimental Psychology 103: 255–62
14. Tulving E 1985 Memory and consciousness. Canadian Psychology 26: 1–12