On the Inability to Generalize Misleading Information Research to Actual Crime Situations

Why We Cannot Generalize Conclusions from Source Misattribution (or Misleading Post-Event Information) Studies to Actual Crime Situations

Ebbe B. Ebbesen

Department of Psychology, UCSD

September, 2001

In progress draft

Considerable research has examined the effects that misleading information has on the accuracy of memory for observed events (e.g., Weingardt, Troland, & Loftus, 1994). In most of these studies the misleading information is presented after exposure to the "to be remembered" material.[1] Such studies examine the effects of "post-event" misleading or suggestive information on accuracy of memory for the event. In the typical three-stage research design, subjects are first presented with the material to be remembered, then they are either mislead or not, and finally their memory for the event they saw in the first stage is tested. The information in the original event is generally presented in the form of a slide show with an accompanying narration (although sometimes videotapes of events are used, e.g., Zaragosa & Mitchell, 1996; sometime directly experienced events provide the information to be remembered, e.g., Roberts, Lamb, & Sternberg, 1999). For example, in one study (Weingardt, Troland, & Loftus, 1994), subjects were shown a slide sequence of a college student walking through the college bookstore looking at and selecting objects from different shelves and then watching a handyman do some work. In one slide, the student selected a can of Coca-Cola (or 7-Up) from the shelf, in another slide he put a copy of Esquire (or GQ) magazine in his pack, and in another he watched the handyman use a screwdriver (or wrench).

The events in the second stage determine whether and about what the subjects are being mislead. Generally half of the subjects are presented with misleading information about a few "minor" aspects of what they saw in the slide sequence and the other half are not mislead. Sometimes all subjects are misled on some critical items and not on others. When this design is used, different subjects are misled on different critical items. In virtually all of the research done in this area, the "new" information presented in the second stage describes a number of specific features of the event correctly but, for the misleading items, also describes a few of the features incorrectly. In most studies the proportion of misled items is much less than the proportion of correctly described features. In the typical study, the second stage occurs (usually more than an hour and many times more than a week) after the initial presentation. Sometimes the "suggestive" information is in the form of questions that also add information (e.g., after seeing the college student take an Esquire magazine from the shelf, the subjects might be asked, "When the student took the GQ magazine from the shelf, did he open his backpack first?") Other times the information is presented in narrative form (e.g., "The college student stopped at the magazine rack, took the GQ magazine from the shelf and put it in his backpack.").

In virtually all of the studies, the narrative is presented at another time, but in the same experimental context as the original "to be remembered" information. The person presenting the narrative, i.e., the experimenter, presumably has special knowledge of the actual contents of the slide sequence and therefore it is likely that subjects would believe that the narrative is a completely factual account of what was actually in the slide sequence (provided they did not have a clear recollection of what they originally saw at the time that they hear the narrative).[2] Evidence supporting a subject's belief in the accuracy of the narrative is the fact that the narrative typically provides descriptions that are consistent with the large majority of "items" that the subject might remember from the event.

The third stage in these studies consists of a memory test. It is generally, but not always, a test for the information presented in the first stage only. The subject is asked to remember what they saw in the slides and not what they heard in the narrative. For example, subjects might be given a two-alternative forced-choice test that presents pairs of slides depicting events in the original set. The choice might be between a slide that was actually seen and one showing the same context but with the misleading information present. In the bookstore example, one slide might show the student taking a GQ magazine and the other might show him taking an Esquire from the shelf. Sometimes other measures, e.g., confidence, are also taken (Zaragosa & Lane, 1994; Zaragosa & Mitchell, 1996). Sometimes items that are presented contain neither the originally seen nor the misled item. Sometimes a cued-recall test is given similar to the following: "The student took a can of soda from the shelf. What was the brand of soda shown in the slides?" (e.g., Lindsay, 1990). Researchers then compare the accuracy of the subjects' performance on the "critical" misled and control items. Accuracy of memory for the non-critical items is usually ignored.

Table 1: Typical Design of Misleading Information Experiment

	*Stages of the Experimental Procedure*
	*Stage 1*	*Stage 2*	*Stage 3*
*Post-Event Condition*	Initial event	Post-event	Memory Test
*Misleading Information*	Subject sees a coke	Subject hears "7up"	Did you see coke or 7up?
*Consistent or Neutral Information*	Subject sees a coke	Subject hears "coke" or "can of soda"	Did you see coke or 7up?

Table 1 shows the typical design of a misleading information experiment. Some subjects read a narrative that misleads and other subjects read one that does not (or misleads on some critical items and not others). The content of the narrative in the non-misleading information condition tends to be either "consistent" information or "neutral" information. In the consistent case the narrative correctly describes some key items of the slide sequence, as in, "He took a can of Coca-Cola from the shelf." In the neutral case, the narrative is less specific and describes key elements in a way that is consistent with both what was seen and the misleading information. In the example being used here, a "neutral" condition might describe the student as taking a can of soda from the shelf without specifying the type of soda.[3]

Note that in these studies, the misleading narrative does not claim that the student took out a gun, shot and killed a woman, and then ran away drinking a can of 7-Up even though such a description would be inconsistent and would clearly be misleading. The narrative does not even describe the student as taking and drinking a container of milk from the refrigerator section of the store rather than taking a can of soda from the shelf. Nor does the narrative describe the male student seen in the slide as a 45-year-old female. Instead the narrative describes all of the actions taken by the student in the slides in a correct manner but substitutes a few objects for other equally plausible, highly similar, and often "inconsequential" objects.[4] Other slides in the sequence might even have shown cans of 7-Up on the shelves right next to the Coke cans as the student moved down the shelf. The point is that the misleading information in these studies is often "peripheral" to the main content of the slide sequence, especially since the subjects do not know beforehand what features of the slide sequence they will be expected to describe or recognize later. [5] In addition, subjects are misled on many fewer items than on which they are not misled. Finally, the nature of misleading information is very close to information originally seen in the slide sequence, so close, that were a free-recall test given and the subjects described the student's actions as, "He took a can of soda from the shelf," or "He watched a handyman use a tool to fix something," they would be correct (actually absolutely correct -- though not as detailed as they could have been) regardless of the narrative or slide sequence that they saw.

The most intensively studied theoretical issue in this area is whether the misleading information changes some response/decision mechanism, e.g., a guessing strategy, or actually causes the subjects' memories to blend or become transformed (Loftus, 1989; Weingardt, Troland, & Loftus, 1994).[6] A related sub-issue concerns the relative roles that encoding, retention, and retrieval processes might play in producing the observed results. For example, will the misleading effect only occur if the original content was poorly encoded? Focusing on the retrieval stage, McCloskey & Zaragoza (1985) suggested the possibility that different subjects in these kinds of experiments could be in any of several different memory states at the time of the memory test. In particular, a subject might recall that she saw a Coke but be unable to recall what she heard in the narrative about the type of soda. Alternatively, the subject might recall quite clearly that she heard about a can of 7-Up and know that she had forgotten what she saw in the slides. Other subjects might have independent recollections of both seeing a Coke and hearing about a 7-Up. Still others might not recall the source of information about the type of soda but have a clear recollection that a can of soda was removed. Still others might not even recall seeing or hearing about the college student taking a can of soda from the shelf. Finally, other subjects might have transformed their memory of what was seen as a result of reading the narrative and actually have a recollection of seeing the student removing the type of soda that they heard about rather than saw. The latter has come to be known under the more general term, source misattribution.

Source Misattribution

Source misattribution begins with the idea that people sometimes have clear recollections of events but can't recall the source of those events. One might recall a funny line from a play but be unable to recall where it was first heard or read. In the context of post-event misleading information, source misattribution occurs when the subject misattributes the source of hearing about the 7-Up to seeing it (rather than the Coca-Cola) in the slide. Another effect of misleading information might be to reduce the accuracy of memory for the original event without source misattribution. That is, after hearing the misleading narrative, subjects might be less likely to perform accurately on a test of their memories for what was seen even if they fail to remember seeing the misled items.

In contrast to the source misattribution explanation, the guessing strategy hypothesis suggests that subjects who claim to have seen the critical misleading items in the narrative might be guessing that they saw a 7-Up because that is what the experimenter told them they saw in the narrative.[7] Subjects might be reporting something other than the exact content of their recollections of what they saw. Many different experimental strategies have been employed in an attempt to determine whether the post-event information actually changes the content of memories for what was seen or the guessing strategies that subjects use. Most likely, aspects of both views are correct for some subjects on some items in some conditions.

What does the misleading post-event literature tell us about actual witness memory?

Regardless of the final resolution of the key theoretical issue in this area, some researchers have concluded that the results have applied significance for the legal system (e.g., Lindsay, Loftus, Weingardt). After all, misled witnesses are more likely to report having seen something that they did not actually see. What does it matter whether the underlying reason for those false reports is a memory failure or motivated guessing? In fact, this is exactly the kind of argument that we have heard experts make in court. They argue that research has repeatedly demonstrated that post-event misleading information increases the odds that people will incorrectly respond in memory tests (compared to the non-misleading condition). Generalizing, they suggest that the memories of witnesses to actual crimes who have discussed with other witnesses or with the police what they saw or who have seen information about the suspect in the media have been irrevocably compromised to the point that their reports of what happened can no longer be trusted. The claim is that the act of discussing what they saw with others or of hearing what others said they saw is the equivalent of the procedures (used in the experimental studies) of hearing narratives with misleading information about a few bits of information.

Unfortunately this conclusion is unjustified given the nature of the procedures that have been used in the research, the nature of the decisions facing jurors and prosecutors in actual cases, and the results from key studies. Consider the claim that the source of the errors, whether from memory or decision-making, can be ignored because in the real world the only thing that jurors hear are the errors. If the errors that misleading information produces are errors resulting from a reconstruction, blending, or transformation of the contents of memory, then motivational differences between witnesses in the real world and those in laboratory studies might be ignored (assuming that such memory processes are independent of witness motivation).[8] Simply hearing inconsistent information would be enough to affect recollections of the original event. On the other hand, errors might result not from memory distortions but from college students guessing what they believe is the most appropriate response given the social conditions (see, for example, Highhouse & Bottrill, 1995) created in the study, e.g., they can recall both the original event and the misleading information but choose the latter because this is what they believe the experimenter wants them to report. Were this case, the ability to generalize to the real world would depend on whether the motivational conditions present in the real world caused actual witnesses to guess in the same way. After all, actual witnesses might have an independent recollection of both the original and the misleading events and report both of them rather than choose one. Such outcomes seem especially likely because examination of police and investigative interviews in the real world suggests that the contents of witnesses' memories are not generally assessed with two-alternative forced-choice or cued-recall procedures. Instead, witnesses are asked in a relatively free recall procedure what they heard and saw.[9] In short, whether the findings from post-event misleading information studies can be generalized depends on the proportion of errors that are "true" memory errors and the proportion that are strategic errors (the person privately recalls correctly but publicly reports otherwise). Unless we know this proportion, we do not know how to take motivational differences between laboratory and real world witnesses into account.

An aspect of motivation that is all but ignored by researchers who claim generality for the misleading information effect is the reluctance of witnesses to testify in actual crime cases. Because many prosecuted crimes involve gang members, a frequent complaint by the police and prosecution is that witnesses are reluctant to testify about the events that they saw out of fear of or loyalty to gang members. If these claims are accurate, such motivational effects on the willingness of witnesses to report information that might be misleading is far from known because it has not been studied.

The issue of motivation is important from a different point view. In particular, the "credibility" or "expertise" of the source of the misleading information might play a role in the strategic decisions people make as well as the contents of their memories (e.g., Okamoto & Sugahara, 1986). As a result, this variable might easily limit the applicability to actual witnesses of findings for subjects in post-event memory studies. For example, strategic decisions about what to report as the correct answer might be affected by the credibility of the source of the misleading information. If the source is another witness who admits that she is completely uncertain or who explains that he only got a brief glimpse, actual witnesses might be much less likely to report such content, even if they completely forgot the relevant detail in the original event. Unless we know how credibility and expertise affect memory and reporting strategies and unless one is able to measure the "expertise" of the source of misleading information, generalization from misleading event research to conclusions about the accuracy of actual witnesses who might have discussed all or part of what they saw with others seems unwise.

Not only is the "expertise" of the source of the misleading information a potential moderating variable, so is the strength of the witness's memory for the initial information at the time that the misleading information is presented. It seems reasonable to assume that the presentation of misleading information will have less of an effect the stronger the memory for the initial event. Although this issue has not been thoroughly explored, it seems intuitively reasonable to assume that misleading information will have a smaller effect or even no effect if the initial memory for the original event is very strong at the time that the misleading information is heard.[10] If the original memory were particularly strong, the witness might even attempt to correct others who were attempting to mislead (e.g., "No, he didn't take a 7-Up, I distinctly remember seeing that he took a Coca-Cola can from the shelf."). Until parametric work is done examining the relatively impact of different degrees of credibility interacting with the strength of memory for the initial event, I can't see how one would know whether a witness's having heard misleading information in a given context in the real world would or would not have any effect on the accuracy of the witness's reports of what he or she recalled.

The fact that the "items" that are tested in the misleading paradigm might represent "minor" features of the entire event has potential significance for our ability to generalize the results from these studies to witnesses and victims of actual crimes. It is important to realize that when taken at face value, these studies (presumably demonstrating the "reconstructive" nature of human memory) do not demonstrate that misleading information completely changes people's entire memories for the events that they have observed. Women who have been raped are not being misled into remembering that the rape never occurred or that the rape was actually a visit to the opera with the culprit. These studies demonstrate that misleading information can alter the accuracy of reports about possibly minor, "unimportant," features of the event. They do not demonstrate that misleading information causes entire memories of events to be reconstructed in completely different forms. In short, when researchers testify in court, there is the possibility that they are over-generalizing from the "non-central" informational domains typically studied in the laboratory to the information remembered by an actual witness. Is the information that might be most probative in determining a defendant's guilt likely to be "central" or "minor" in actual crimes? Until appropriate data are available from actual crimes, generalization seems unwise.

In addition, as Lindsay (1990) has noted, the misleading information has to be very similar to the original information for the effect to occur. Again, a can of Coke changes to a can of 7-Up not a hotdog or even a can of Bubble Bee tuna fish. Unfortunately, parametric studies examining the impact of the degree of similarity between the misleading and original information in the context of different strengths of the original memory and different degrees of credibility of the source of the misleading information have not been done. As a result, we do not really know how to generalize findings from studies showing that an effect can occur under the right conditions to situations in which the issue is did an effect occur in a particular witness for a particular type of overheard information. I have heard the testimony and read transcripts of the testimony of many eyewitness experts. Not one has qualified his or her conclusions about the impact of misleading information on the accuracy of memory by noting the roles that credibility of the source, strength of initial memory, similarity between the content of the original information and the misleading information might have on the size of the effect.

Lindsay (1990) reported results of a manipulation that reduced the impact of the misleading information on accuracy. He varied the "ease of discrimination" between the original and the misleading information. In the "easy discrimination" condition the misleading information was presented in a different location several days after the subjects saw the slides and a different person read the post-event narrative than read the narrative that accompanied the original slides. In the "difficult discrimination" condition, the same person who read the narrative that accompanied the slides read the post-event narrative. In addition, this person read the post-event narrative in the same room as the slides were shown and immediately after the slide show. Interestingly, the effect of the misleading information on the willingness of subjects to report that they saw the misleading items was eliminated in the easy discrimination condition. The fact that manipulations such as this one might eliminate the effect of misleading post-event information means that attempts to generalize will have to take into account "ease of discrimination" between the misleading and original information. Since crimes in the real world do not occur in laboratory rooms with narrations read by individuals whose voices might sound similar, it seems reasonable to ask whether watching a bank robbery and then discussing what one saw with another witness soon after they witnessed the crime would be an "easy discrimination" or a "difficult discrimination." Clearly, until additional research is conducted that examines exactly what features of the post-event experience make it easy or hard to distinguish from the original event, it would seem unwise to generalize the misleading event findings to actual crime situations.

Another potential limit on the generality of the findings from this research concerns the form of the misleading information in actual cases compared to the form used in laboratory studies. For example, although it is likely that misleading information might occur in the form of conversations that witnesses have with other individuals (investigators, prosecutors, other witnesses), the nature of these conversations might not be in the form of narratives that describe events. Even if they are similar to narrations, it is clear that the original event was not accompanied by a narration (as happens in many of the studies that present the original information in the form of slide shows). As such, the possibility that some of the effect of the misleading information might be the result of subjects hearing two different narratives rather than seeing information and then hearing different information later.[11] Finally, it is unclear whether in most actual crime situations witnesses hear only one other source of potentially misleading information. It is conceivable that when witnesses discuss things with others, they have more than one discussion. We know little about what might happen when witnesses hear different misleading events, e.g., the Coke is described as a 7-Up by one person and a Dr. Pepper by another compared to being described as a Coke by one and as a 7-Up by another. Just as important is the fact that in conversations, rather than in situations in which someone passively listens, the witness is hearing himself or herself say things. If these things are inconsistent with what the other witness is saying, it is unclear how to predict what the joint effect will be.

Another Interpretation of the Misleading Event Effect

As noted earlier, some researchers (e.g., Weingardt, Troland, & Loftus, 1994) argue that misleading information not only increases the odds that subjects will respond as if they remember the misleading information as having been seen, it also increases the odds that the originally seen critical items will not be remembered at all. Evidence for this comes from "yes/no" recognition memory tests in which the subjects are presented with both seen (Coke) and not seen (7-Up) critical items and asked to indicate whether they saw each item. One study (Troland, 1990) reported in Weingardt, Troland, & Loftus (1994), found that the hit rate for seen items was 73% when the subjects had not been mislead and 60% when they had been mislead. Similarly, the false alarm rate was 21% when they had not been mislead and 43% when they had been mislead. In other words, when the subjects were told that they had seen the 7-Up, the likelihood of them saying that they had seen the Coke decreased. This has been taken as evidence that memory for what was seen gets worse when people are mislead.

When this study is recast in terms of signal detection theory, another interpretation of the results is possible. In a signal detection theory of recognition memory, we can think of items in terms of their position on a subjective dimension of "evidence that the item was seen before" (e.g., Ebbesen and Wixted, 1996). Better-encoded and better-remembered items will presumably have higher values on this dimension. Items that have not been recently seen would, on average, have lower values on this dimension. One way to describe the effect of hearing misleading information on the likelihood of saying that one saw the misled item before would be to assume that hearing the misleading description moves the item up on the subjective evidence dimension. That is, it increases the average (over subjects and items) strength or familiarity of the unseen critical items. We can represent this in terms of distributions of items as in Figure 1. One of the consequences of the not seen distribution moving higher on the evidence dimension is that the decision criterion moves higher as well. Since the strength of the seen items does not change with the presentation of the misleading items, we would expect that distribution to remain in its original place. If this model of the effect of misleading information has validity, it could explain the reduction in hit rate reported when misleading information is presented. The interesting part of this model is that it does not require a reduction in the strength of memory for the seen items as the strength of memory for the mislead items increase in strength.

Figure 1: Signal detection representation of the effect of misleading information on the strength of the not seen and seen distributions. The decision criterion moves up as the strength of the not seen mislead items increase.

One of the interesting consequences of the model in Figure 1 is that it demonstrates why it might be premature to generalize conclusions about memory for information from studies examining the effect of misleading information. On the other hand, to be fair, at this point, this model has not been tested and may not apply to the misleading information paradigm.[12]

Logic Is the Ultimate Limitation on Generality

Even if the above potential limitations on the ability to generalize are rejected as irrelevant, there is one additional logical problem that not only cannot be dismissed but also virtually makes all of the misleading event research irrelevant to actual crime situations. We know from the studies in this area that subjects are much less likely to report misleading information for the critical items about which they heard consistent information. That is, when subjects see a Coke and then hear about a Coke (or a can of soda), they will rarely report that they saw a 7-Up. In fact, although the studies are generally not designed to test the following prediction, it seems reasonable to assume that seeing a Coke and then hearing about a Coke would make it more likely (assuming that memory is reconstructive in some circumstances) that subjects would report seeing a Coke compared to a condition in which they just saw a Coke and heard nothing. In other words, just as hearing misleading details increases the odds of reporting those details, one might argue that hearing correct details would increase the odds that subjects would report correct details (compared to someone who heard no additional details). If this reasoning is valid, it raises the possibility that hearing additional details might either reduce or enhance the accuracy of the witness's reports depending on the content of the overheard information.

Table 2: Information available to jury

	*Stages of the Criminal Justice Procedure*
	*Stage 1*	*Stage 2*	*Stage 3*
*Post-Event Condition*	Initial event	Post-event	Memory Test
*Misleading Information*	Witness sees ?????????	Witness overheard Coke (defense claims it was 7Up)	Witness claims saw Coke
*Consistent Information*	Witness sees ?????????	Witness overheard Coke (prosecution claims is was Coke)	Witness claims saw Coke

With this in mind, consider the decision facing a typical jury. Keep in mind that the jury does not see all three of the stages that occur in a typical laboratory study. At best they only see the last two stages. They hear a witness testify that she saw the defendant take a can of Coke from the shelf (stage 3). They also hear from another witness (stage 2) that he recalls having a conversation with the first witness in which he told her that he recalled seeing a can of Coke being removed from the shelf (although his angle of viewing prevented him from identifying the defendant). The jury also hears an expert testify about the many studies that provide overwhelming converging evidence for the misleading post-event memory effect. What should a rational jury decide? Should they assume that the first witness was obviously misled by the second witness and therefore her recollection that she saw a Coke is in error? Unfortunately, a rational decision requires that they know the odds that the second witness's statement to the first witness is consistent or inconsistent with what actually took place.[13] If the second witness's statement was not misleading but described the actual events, then the first witness is more likely to be correct. If the second witness's statement was misleading, then the first witness might be less likely to be correct. Unfortunately, there is nothing that tells the jury (in this hypothetical) about the consistent or misleading nature of the second witness's narrative. The prosecution argues that both witnesses have independent recollections of the actual event. The defense argues that the first witness was misled by the incorrect memory of the second witness. We cannot choose which version is correct without knowing the content of the critical event. The problem would be made even more difficult if the prosecution were to present evidence from a police report indicating that the first witness might have said in her initial interview, before speaking with the second witness, that she thought the soda might have been a Coke. What should a rational jury decide under these conditions?

A critic of the present argument might say that, at the least, it should be clear that the first witness's recollection should not be taken as independent evidence that it was a Coke because the possibility exists that her memory was contaminated. However, without knowing whether it really was a Coke, the possibility also exists that the witness had an independent memory that was merely strengthen by hearing the other witness confirm what she already knew to be true. Therefore, without knowing the baserate of confirming v misleading conversations among witnesses, the mere fact that misleading information can cause higher error rates does not caste doubt on the independence of the witness's memory.

One might argue that the absence of the conversation between two witnesses would supply stronger evidence that the first witness saw a Coke. But this is not necessarily correct. We would need to compare the error rates of a group of witnesses who did not have conversations with a group of witnesses who heard confirming and disconfirming evidence in proportion to their occurrence in actual crime cases. Suppose, for example, that confirming conversations occur 95% of the time and misleading ones occur only 5% of the time in actual cases. It is quite conceivable that witnesses who have discussed events with others are more accurate on average than witnesses who never discuss what they saw. For example, there is some evidence that the accuracy rate (proportion of correct items recalled to the total number of recalled items) is generally very high in free recall (Ebbesen & Rienick, 1998; Roberts, Lamb, & Sternberg, 1999; Yuille & Cutshall, 1986). That is, most of the content that people recall is accurate. If this is generally the case for witnesses to actual crimes, then most of the content of what witnesses tell each other about jointly witnessed events will be correct content. As a result, the effect of conversation might actually be to increase the overall accuracy of witness testimony (e.g., Hinsz, 1990; Vollrath, et. al., 1989). In fact, the more one thinks about the misleading event paradigm in the context of actual witness discussion, the less likely it seems that discussion would decrease the accuracy of witnesses. For example, some evidence suggests that recollections in which people are most confident are more likely to be correct than other things in which they are less confident (Ebbesen, 2000; Ebbesen & Rienick, 1998; Ebbesen & Wixted, 1996). If so, not only are those things that witnesses tend to hear others mention likely to be correct, but they are also likely to have a greater impact because of the added credibility of high confidence (Hinsz, 1990; Hinsz, Tindale, & Vollrath, 1997).

In its current state, research on the misleading information effects seems irrelevant with regard to judging the accuracy of actual eyewitness memory. We know virtually nothing about the relative rate at which misleading compared to consistent information occurs in overheard and direct conversations. Before we generalize results from laboratory studies of misinformation, we need to study those real world settings in which the witness's initial memory of the event is relatively weak and she is likely to accept the other person's version and/or be willing to report what the other witness said even when she might have clear recollection that this content came from the other witness and not from an independent recollection of the original event, Until relevant research is conducted, it seems premature to assume from what is currently known that talking with others means that a witness's recollections have been contaminated.

References

Bjorklund, D. F., Bjorklund, B. R., Brown, R. D., & Cassel, W. S. (1998). Children's susceptibility to repeated questions: How misinformation changes children's answers and their minds. Applied Developmental Science, 2, 99-111.

Ebbesen, E. B. (2000). Some thoughts about generalizing the role that confidence plays in the accuracy of eyewitness memory. Unpublished manuscript. University of California, San Diego.

Ebbesen, E. B., & Rienick, C. B. (1998). Retention interval and eyewitness memory for events and personal identifying attributes. Journal of Applied Psychology, 83, 745-762.

Ebbesen, E. B. & Wixted, J. T. (1996) A signal detection analysis of the relationship between confidence and accuracy in face recognition memory: Implications for eyewitness identification. Unpublished manuscript. University of California, San Diego.

Highhouse, S., & Bottrill, K. V. (1995). The influence of social (mis)information on memory for behavior in an employment interview. Organizational Behavior & Human Decision Processes, 62, 220-229

.Hinsz, V. B. (1990). Cognitive and consensus processes in group recognition memory performance. Journal of Personality & Social Psychology, 59, 705-718.

Hinsz, V. B., Tindale, R. S., & Vollrath, D. A. (1997). The emerging conceptualization of groups as information processes. Psychological Bulletin, 121(1), 43-64.

Lindsay, D. S. (1990) Misleading suggestions can impair eyewitnesses' ability to remember event details. Journal of Experimental Psychology: Learning, Memory and Cognition, 16, 1077-1083.

Loftus, E. F. (1989). Misinformation and memory: The creation of new memories. Journal of Experimental Psychology: General, 118, 100-104.

McCloskey, M., & Zaragoza, M. (1985). Misleading postevent information and memory for events: Arguments and evidence against memory impairment hypothesis. Journal of Experimental Psychology: General, 114, 1-16.

Okamoto, S. I., & Sugahara, Y. (1986). Effects of postevent information on eyewitness testimony. Japanese Psychological Research, 28(4), 196-201.

Roberts, K. P., Lamb, M. E., & Sternberg, K. J. (1999) Effects of timing of post-event information on preschooler's memory of an event. Applied Cognitive Psychology, 13, 541-559.

Vollrath, D. A., Sheppard, B. H., Hinsz, V. B., & Davis, J. H. (1989). Memory performance by decision-making groups and individuals. Organizational Behavior & Human Decision Processes, 43, 289-300.

Weingardt, K. R., Toland, H. K., & Loftus, E. F. (1994). Reports of suggested memories: Do people truly believe them? In D. F. Ross, J. D. Read M. P. Toglia (Eds.), Adult eyewitness testimony: Current trends and developments.: Cambridge University Press, New York, NY, 3-26.

Yuille, J. C., & Cutshall, J. L. (1986). A case study of eyewitness memory of a crime. Journal of Applied Psychology, 71, 291-301.

Zaragoza, M. S., & Lane, S. M. (1994). Source misattributions and the suggestibility of eyewitness memory. Journal of Experimental Psychology: Learning, Memory, & Cognition, 20(4), 934-945.

Zaragoza, M. S., & Mitchell, K. J. (1996). Repeated exposure to suggestion and the creation of false memories. Psychological Science, 7(5), 294-300.

Footnotes