Reliability of Long term memory
Effects of repeated tests
Information doesn’t just stay in the system indefinitely (or does it?). What determines the loss of information from memory?
Accuracy of personal memory
Evidence on the accuracy of autobiographical memories is conflicting. People can be highly accurate over very long periods of time, or highly inaccurate, depending on the exact circumstances.
Bahrick et al. (1975) tested participants on their memory for the names and faces of people they had been with at college. Different groups were tested between 3 mo. and about 50 yr after graduating. In the free recall test: pp's remember as many of their classmates as possible. In the picture recognition test, photographs of classmates and non-classmates had to be discriminated. In the picture matching test, pp's pick the face that matched a classmate’s name.
Bahrick et al. found that there was steady forgetting on the free recall test, but virtually none on the recognition or matching tests, until 30 or so years had elapsed (by which time ageing may have played a role). Of course, exposure to class mates involves repeated “learning trials” and hugely distributed practice, but neither of these factors can be critical because studies of single-event memories can be equally accurate.
Wagenaar (1986) recorded 1 or 2 events that happened to him each day for 6 yr. Each event was recorded in terms of WHO WHAT WHERE. At the end of the 6 yr period, Wagenaar tested himself by providing either 1, 2, or 3 cues.
At the end of the 6 yr period, Wagenaar tested himself on each day. The recording of who, what and where developed three cues to help him remember the memories. He then tested himself on these memories, using these retrieval cues. He found that over time, memory did slowly decline, however using the retrieval cues allowed nearly all memories to be recalled. For the events that were not recalled despite the use of all three cues, Wagenaar contacted the other people involved in the incident, who were then able to provide him with more cues which resulted in memory recollection. Thus with sufficient cues, accurate memories tend to be recalled.
Flashbulb memories
Other data on “flashbulb” memories tend to reveal quite the opposite pattern of results. Neisser and Harsch (1992) studied students’ memories of the Challenger space-shuttle disaster. The day after the accident, participants described the circumstances in which they heard the news. They answered questions such as: what time was it? how did you hear about it? where were you? who else was there? etc. Then 2½ years later they were questioned again.
The data revealed an amazing amount of misremembering. In fact, only 3/44 participants remembered accurately what they were doing. Most were mistaken in some respect, and 11/44 participants were entirely wrong and gave reports that
were complete inventions.
Similarly, Bartlett’s famous studies of the “War of the Ghosts” story showed that people can be very poor at accurately remembering a story. With repeated tests, memory includes more and more schema-consistent elements that were not part of the original story, and fewer and fewer true elements. Although Bartlett’s studies were rather informal, the main findings were replicated by Bergman and Roediger (1999). Remembering appears under some circumstances to be a highly reconstructive process.
What is the explanation of the discrepancy between Wagenaar’s (1986) and Bahrick et al.’s (1975) results on the one hand and Neisser and Harsch’s (1992) and Bartlett’s on the other? There are at least 3 likely factors (and probably many
more).
One is the interval between repeated tests. A second is the extent to which subsequent thinking about the event provides an opportunity for the construction of a socially-shared narrative with which the initial memory becomes integrated, and that this narrative leads to assimilation and alteration.
Because flashbulb events make for a good story, people tend to add spurious details, particularly as a result of social interaction. Thirdly, the familiarity of the event or materials can lead to lesser or greater degrees of schema-based reconstruction. Can we test these ideas?
A clear prediction is that the interval between tests should be important.
With longer intervals there should be greater opportunity for distortions to intrude than with short intervals. Indeed, repeated
testing at short intervals may lead to substantial increases in the amount of information correctly recalled (hypermnesia). Bartlett’s studies, in contrast, tended to include longer intervals (e.g., a week) between tests.
Wheeler and Roediger (1992) obtained evidence for both patterns within a single experiment, showing that, indeed, whether hypermnesia or forgetting occurs depends on the interval between tests. They presented participants with 60 pictures. For half the participants, the pictures were accompanied by their names, for the other half they were connected by a story. Then at
various delays, forced recall tests were administered. All groups received 3 consecutive tests one week after the learning
phase, but they had either 0, 1, or 3 consecutive tests immediately after the study stage.
The results showed that compared to an immediate test, a test administered after 1 week showed considerable forgetting (Bartlett’s result). However, tests administered close together in time produced the exact opposite, namely hypermnesia. In general, repeated tests are very beneficial for memory accuracy, whatever the inter-test delay. Whether or not the materials could
be linked by a story was irrelevant.
Nature of information to be remembered
Bartlett placed emphasis on the idea that individuals interpret events in terms of pre-existing schemas. Familiar events or materials will be straightforwardly consistent with the schemas they evoke. Unfamiliar events or materials however will evoke a pre-existing schema that is likely to colour or distort memory. Thus we might anticipate greater distortion of memory when the materials are unfamiliar, and the evidence seems to point to that conclusion.
For example, in Wynn and Logie’s (1998) study, students showed good recall across 6 months in details about specific lectures they had attended, even when tested repeatedly in the manner of Bartlett’s experiments. The nature of the event is important: a flashbulb event is the sort of thing we turn into a story, whereas an everyday event is not
Hence the discrepancy between studies which show sharp forgetting (Neisser & Harsch, Bartlett) and those which don’t (Bahrick et al., Wagenaar) seems to be explicable in terms of several factors, including test interval (with long intervals between tests providing greater opportunity for intrusions and reconstruction), narrative reinterpretation, and the scope for schematization
Emotionally arousing information
Similarly, research on ‘flashbulb’ memories has suggested that some information is remembered much more reliably than other. In particular, memories of emotionally significant events, such as a natural disaster, tend to be more accurately recalled than everyday activities (Neisser, 1996).
This enhanced memory of arousing incidents is believed to be an adaptive behaviour, ensuring we remember dangerous and life-threatening information (Roozendaal, McEwen & Chattarji, 2009).
Cahill and McGaugh (1995) demonstrate this phenomenon in a study in which participants are presented with one of two stories. They are the same length and equally as complex, yet one story includes more emotive language and events; there is a car accident and the person involved is rushed to hospital. Results showed that memories regarding the particularly emotional details of the accident are more reliably recalled. On the other hand, the story eliciting low arousal does not result in enhanced memory.
The mechanism behind this reliable recollection is believed to lie in the stress response. McGaugh and Roozendaal (2002) investigated how hormones released in the face of a stressful situation affect memory. They found that the release of glucocorticoids is associated with information of a highly emotive nature being more accurately remembered. This is believed to be due to their action of the amygdala. This suggests that enhanced memory consolidation occurs when information is emotionally arousing due to the amygdala facilitating memory consolidation in the hippocampus. These long-lasting memories are then more easily recalled.
Narrative reconstruction
When an individual engages in social interaction with another person, dialogue and the desire to construct a shared narrative version of events can lead to major memory distortions.
For example, Gabbert et al. (2003) created two slightly different versions of a video showing a girl entering an unoccupied office to return a book. To reflect the fact that the same event might be seen from different perspectives by different people, some details were altered in the two versions. For instance, in version A (but not B) the title of the book was visible and the girl is seen throwing a note in the waste bin. In version B (but not A) she checks the time on her watch and steals £10. Some participants watched the videos in pairs, unaware that they were seeing slightly different versions. Others saw one of the versions individually. The paired subjects then jointly answered a questionnaire about the events in the video as if they were witnesses preparing their testimony, while participants who had seen a video individually answered the same questionnaire on their own. Finally, all participants individually attempted to recall the original events as accurately as possible.
Gabbert et al. found that the opportunity to construct a shared version of events led to significant levels of distortion in the final recalls. Over 70% of participants who had been in a pair recalled a detail provided by their co-witness, that they themselves had not observed. More strikingly, when asked to report whether the girl was guilty, when they had seen no evidence of that, over half the participants in pairs judged that she was. Participants who had seen the video in isolation, by contrast, recalled accurately. Thus memory can be distorted to include details seen by another witness but not by the individual him/herself (and vice versa).
The social relationship between the two people who share information about the event can also have an influence on accuracy of recollection. Incorporation of information from another person is more likely to occur if the other person is a friend or partner (Hope, Ost, Gabbert, Healey, & Lenton, 2008).
Since many of our lifetime experiences are within a social context, this may be a particularly important factor to consider. Furthermore, it has practical legal complications. This research suggests that co-witnesses of a crime may influence each others recollection of the event if they are able to discuss the event. This could have serious implications on whether or not someone is arrested, or even sentenced.
Inducing false memories
We have seen that memory can not only suffer errors of omission (i.e., forgetting something that happened) but also errors of commission (i.e., reporting something that did not), the latter being false memories. False memories can have enormous applied significance, sometimes leading for example to unsafe criminal convictions based on mistaken eyewitness testimony. In legal settings, errors of omission (forgetting) tend to reduce the likelihood of convicting a guilty person whereas errors of commission (false memory) can increase the likelihood of convicting an innocent person, and we generally regard the latter as less tolerable than the former.
1. False memory of a strong associate
2. imagination and source confusion
3. imagination inflation
False memories can arise during narrative or schema-based reconstruction. But what exactly are the cognitive processes underlying their creation? How is it that memory can be tricked in this way? It turns out to be quite easy to induce people to falsely remember something. This is in accord with the theory that says that all experiences of memory are attributions
Roediger and McDermott (1995) presented participants with lists of 12 words such as water, stream, lake which were all associates of a critical nonpresented word (river). When asked to immediately recall the words, 65% of list words were remembered, but 40% of the critical nonpresented words were falsely remembered. Participants were then given a recognition test including old words, critical nonpresented words, and unrelated lures, and gave confidence ratings for their responses.
A number of studies have shown that imagining an event can lead to it falsely being recalled as having actually occurred. This is called source confusion. For example, in Weinstein and Shanks’s (2008) procedure, participants first studied and named coloured or black and white pictures. Next, they saw the names again and had to report whether the corresponding picture was in colour or black and white in the study phase. Critically, some of the object names in this phase were not shown in the first phase, but participants were told that if they couldn’t recall an item, they should imagine what it might have looked like. Finally, a recognition test was presented containing
Hyman and Pentland (1996) have gone on from the basic imagination/source confusion effect to show imagination inflation, the finding that repeated imaginings can strengthen false memories still further. They asked the parents of college students to recall events that happened to their children prior to age 6 in each of 10 categories (e.g., going to hospital, getting lost, a holiday, etc.). The students were then interviewed on 3 consecutive occasions at intervals of 2 days. In each interview, they were asked to recollect each of the parent-provided memories and describe as much additional detail as possible.
Again, participants falsely believed the critical nonpresented words had been on the study list, and moreover often reported high confidence in their decisions. Thus even in a simple immediate memory task, confident false memories are easy to elicit. The most likely explanation is that the critical lures were processed fluently because of their semantic relatedness to the study words, and this fluency was then unconsciously attributed to repetition.
Weinstein and Shanks’s results show substantial evidence of false memory. In the test, participants often reported seeing a picture 20 min earlier which they did not in fact see. Weinstein and Shanks (2010) have also shown that even when offered the choice between responding ‘imagined’ versus ‘seen and imagined’, participants still recall the critical items as having been seen. This source confusion effect can extend to forced-choice recognition: when shown a seen picture and an imagined one and asked to select the previously seen one, participants often choose the imagined picture, even for highly individuated scenes.
They were also asked to recollect a fictitious event, an occasion at age 5 at a wedding reception when they had accidentally spilled a punch bowl over the parents of the bride. This event was presented just like the true memories as having been reported by the parents. Based on the students' descriptions, each event was classified as either remembered (e.g., if the student acknowledged it as such or generated additional detail about the event) or not remembered.
In the first interview, most true events were remembered and most fictitious events were not. However, the interesting aspect of the study concerns changes across the 3 interviews. In each interview, if an event was not recalled, participants in the imagery group were told to imagine what the event might have been like and to describe details of their image (e.g., what objects and people looked like). They were told that this imagery might increase the likelihood of recalling the event. Hyman and Pentland found that this imagery process considerably increased the likelihood of participants falsely recalling the fictitious event. This effect did not occur in a control group that did not receive the imagery task. Moreover, participants in the imagery group became nearly as confident that the false memories were veridical as the true ones. The results clearly imply that false memories can be implanted via suggestion and imagery.
Bernstein et al. (2005) have used suggestion to implant the false memory that individuals had an unpleasant childhood experience, namely falling ill after eating strawberry ice cream. After the false memory induction, some of these individuals avoided eating strawberry ice cream. Hence a false memory can have real effects on behaviour
Finally....Repression and recovered memories
The topic of recovered memory has become very contentious recently as a result of legal cases in which the primary evidence of (say) childhood abuse is from memories that have allegedly been repressed and then recovered. To the extent that long-term memory can be unreliable, the idea of recovered memories is rather implausible. Perhaps some of the people who claim to have recovered memories of childhood abuse are being tricked by their memories in the same way as participants in laboratory experiments can be?
What is the relevant evidence? The best reason for believing that recovered memories might be veridical is that, as we shall see in Lecture 9, it is quite hard to show that post-event information actually destroys the original memory trace. On the other hand, there are plenty of reasons to be sceptical about claims of recovered memory and about the notion that memories can be repressed. For instance, Loftus (1993) cites a study of children who had witnessed a parental murder but which was unable to find a single child who had repressed the memory.
On the contrary, the children were continually besieged by thoughts about the murder. Pope and Hudson (1995) review published research and conclude, controversially, that no convincing evidence of recovered memory of abuse currently exists.