The (Mis)information Effect: Trace Alteration or Coexistence?

Draft 8/14/02

Heather Flowe

University of California, San Diego

Addrss ms. correspondence to hflowe@psy.ucsd.edu

Abstract

Reviews the research evidence that is often used to support the argument that misleading postevent information alters existing memory traces. A meta-analysis of published experiments (N = 95) found that the size of the misinformation effect depends on age of subject and type of to-be-remembered stimuli. Adult performance in event memory studies on the modified test was not influenced by misleading postevent information, whereas children's responses were influenced on both the standard and modified tests. Additionally, interference and spontaneous recovery effects were observed only in the adult studies that used unrelated lists of pictures as the to-be-remembered stimuli. Taken together, the results suggest that both the original and misled items co-exist in memory and support a response biases and/or a blocking account of the misinformation effect.

The (Mis)information Effect: Trace Alteration or Coexistence?

Sequestering witnesses while others testify has long been relied upon by police investigators and in court as a means to preserve the independent character of witness testimony (6 Wigmore §§ 1837-1838; Federal Rules of Evidence § 615; California Rules of Evidence, § 777). These practices are enacted in the legal system to help deter social conformity and misconduct among eyewitnesses. What is more, keeping witnesses apart while others are giving testimony may help prevent memory impairment. According to some memory theorists, exposure to new information after witnessing an event could permanently alter memory (Loftus & Hoffman, 1989). Memory distortion is thought to be especially likely if an eyewitness later learns information inconsistent with what he or she remembers perceiving.

Loftus' (1979) updatable memory hypothesis is perhaps the single most influential account of how eyewitness memory distortion could arise following exposure to inconsistent postevent information. In her seminal paper (Loftus, Miller, & Burns, 1978), participants were presented with a slide sequence that featured an auto-pedestrian accident. Following its depiction, a written test was administered, ostensibly querying the participant's memory for the event. One of the questions, however, incorrectly suggested to participants in the experimental group that the red Datsun depicted in the slides stopped at a stop sign, when in fact a yield sign was present. Control participants received consistent information about the traffic sign on the questionnaire rather than misleading information. A two-alternative forced-choice recognition test was given in the final phase of the experiment. For the question that asked what kind of traffic sign was shown, two slides of the Datsun, one featuring a yield sign and the other a stop sign, were shown side by side as response alternatives. Misled participants incorrectly indicated that a stop sign was shown in the original slide sequence more often than control participants. Loftus and Hoffman (1989) later termed this experimental outcome the misinformation effect. Others have since replicated these results, presenting participants with a to-be-remembered event followed by a questionnaire or a written narrative imbedded with incorrect event suggestions (e.g., Bekerian & Bowers, 1983; Bonto & Payne, 1991; Toglia, Payne, & Anastasi, 1991; Belli, 1993; Welch-Ross, 1999).

Following on from the report by Loftus, Miller and Burns (1978), a number of researchers began to investigate the conditions under which the misinformation effect obtains. In particular, much research on the misinformation effect has focused on determining whether the constituents of memory for a witnessed event are overwritten and permanently lost following the presentation of misleading information. Loftus interpreted her findings as evidence that memory representations can be altered (e.g., Loftus & Loftus, 1980). According to her updatable memory hypothesis, forgetting occurs because stored memories can be overwritten with new information. The following explanation has been offered to explain how this might work (Loftus, 1979, p. 112-115; for other similar but less precise theoretical statements of how misinformation alters memory, see Loftus, Schooler & Wagenaar, 1985; Loftus & Hoffman, 1989; Wright & Loftus, 1997; Wright & Loftus, 1998; Braun & Loftus, 1998): Memory is the product of reconstruction and what a witness sees is continually being interpreted. To illustrate, a witness to the auto-pedestrian accident, who is accepting of the misinformation that is imbedded in the interpolated questionnaire or narrative, might visualize the misleading detail as part of the scene. The witness will then store these fragments of misinformation in memory. The event will be reconstructed using all of the fragments, misinformation included, during subsequent recalls. This reconstructive process in turn leads the witness to incorrectly respond on the final memory test that they saw a yield sign when in fact a stop sign was present. Loftus has provided a bolder statement of the updatable memory hypothesis elsewhere: "Give us a dozen healthy memories well-formed, and our own specified world to handle them in. And we'll guarantee to take any one at random and train it to become any type of memory that we might select--hammer, screwdriver, wrench, stop sign, yield sign, Indian chief--regardless of its origin or the brain that holds it." (Loftus & Hoffman, 1989, p. 103).

According to some theorists, however, postevent information might co-exist with the original memory, rather than overwrite it. For instance, McCloskey and Zaragoza (1985) challenged the evidence supporting the destructive updating account by questioning whether the typical misinformation effect testing procedures biased participants' responses rather than their memories. They argued that participants who remembered both the original and misled items might select the incorrect alternative on the test because they feel obliged to select the alternative suggested by the experimenter. The second objection they raised concerned whether misled participants who did not encode the original information appearing in the slide sequence might have used the postevent information to "fill in the blanks" of their memory. They argued that if this were the case, then the higher error rate observed for the misled group reflects misinformation acceptance, rather than memory overwriting. Assuming that 40% of control participants correctly recognize the stop sign and the remaining 60% simply guess on the final test, total accuracy expected for the control group would be 70%. A lower accuracy rate, however, is expected for misled participants. Once again it is assumed that 40% recognize the stop sign. For the 60% who did not encode the stop sign, however, 40% might select the yield sign because they remember having read about one. The remaining subjects will simply guess on the final test and by chance alone half will correctly select the stop sign. As such, expected overall accuracy would be only 30% for the misled group, significantly lower than that expected for the control group.

To address these issues, Mcloskey & Zaragoza (1985) developed a modified recognition test, which provided both the correct item and a novel alternative as response choices. The postevent narrative presented misleading information about two critical items and neutral information about two control items. Participants were given either the standard or modified recognition test in the final phase of the experiment. McCloskey and Zaragoza (1985) reasoned that if the destructive updating view is correct in its assertion that memory is impaired by misleading information, misled participants would guess on the modified test, thereby performing more poorly than control subjects. This is not what they found however. They obtained the misinformation effect with the standard testing procedure, a result that is in keeping with past research; however, in the modified test condition, correct recognition rates did not differ for misled and control items. Since these findings contradicted predictions based on the destructive updating view, McCloskey & Zaragoza (1985) concluded that misleading information has no effect on memory at all.

In response to these findings, Loftus argued that the modified test is not sufficiently sensitive to detect small memory impairments (Loftus, Schooler, & Wagenaar, 1985). The modified test, she argued, is not appropriate for detecting blend memories, which are composed of fragments of both the original and interpolated information. Given that the modified test does not provide the misinformed item as a response option, subjects will be biased to select the original item because it has some features that resemble their blend memory. If such composite objects really existed, like "stop-yield" signs, misled subjects would select them given the option. In support of this argument, Loftus cited work on illusory conjunctions (e.g., Treisman & Schmidt, 1986) and her own work, which has demonstrated that people who are shown a green object but later misinformed that it was blue will tend to indicate on a final recognition test that the object they saw was bluish-green (Loftus, 1979).

The intent of this paper is to evaluate the claim that memory can be overwritten by misleading postevent information. The argument that will be advanced is that after 25 years of investigation, there is little solid evidence to support the notion that misleading information erases existing memory traces. Rather, retrieval based explanations of the misinformation effect seem more consistent with the empirical evidence gathered to date. Accordingly, this review examines the boundary conditions of the misinformation effect and the factors that have been shown to attenuate the size of the effect. A meta-analysis of the studies that have used the standard and modified testing procedure is presented to further evaluate these factors and to determine whether the observed performance decrements that follow misleading information reflect encoding failures, temporary memory lapses, or complete storage losses. Potential methodological problems that need further attention by researchers who use the misleading information experimental paradigm to study eyewitness memory impairment are explored and the applied implications of this research are considered.

The following summarizes the results of studies that have identified key factors that influence the misinformation effect. The research evidence is critically evaluated to determine whether it supports the updatable memory hypothesis or other retrieval based explanations of the misinformation effect. Key factors include the effects of retention interval, the timing of misinformation, item centrality, retrieval cues, and age.

Retention Interval and the Timing of Misinformation. The recent debate over whether erroneous postevent information causes eyewitnesses to "unlearn" event details is reminiscent of the classic debate that memory theorists in the verbal learning tradition have had over whether forgetting reflects storage or retrieval failures. An early consolidation theory of memory proposed by Müller and Pilzecker (1900) argued that learning continues even after the rehearsal period ends. A process they termed "perseveration" acts long after the practice period to consolidate memory traces. Consolidation will be prevented and recall of the critical information will not be possible if perseveration is interrupted. Hebb (1945) later formulated a theory of the biological mechanisms that underlie perseveration, proposing that assemblies of neurons are modified either through biochemical or structural changes during learning, which in turn affect communication between neurons. Evidence supporting the idea that a period of inactivity or uninterruption is needed to consolidate information and form long term memory has been widely reported (e.g., Ebbinghaus, 1885; Jenkins & Dallenbach, 1924; Minami & Dallenbach, 1946; Miller & Springer, 1972; Lisanby, Maddox, Prudic, Devanand & Sackeim, 2000) For example, Minami and Dallenbach (1946), conditioned Periplaneta americana, or the cockroach, to elicit an avoidance response and found greater retention if the cockroaches were immobilized rather than forced to engage in activity following the learning period. They inferred from these results that the immobilized cockroaches were able to consolidate the information, whereas those who were interrupted could not because their perseveration process had been cut short. In humans, evidence for consolidation theory comes from the observation that retrograde amnesia often follows electroconvulsive therapy, or any adverse event that inflicts trauma to the brain. Consolidation theory would argue that trauma interrupts the perseveration process and prevents the consolidation of information learned prior to the traumatic episode. However, strong forms of consolidation theory, which argue that if consolidation is prevented then the to-be-learned information will never be recalled, have generally not been supported (see Crowder, 1982). For example, the retrieval of memory following interrupted consolidation has been demonstrated experimentally (e.g., Muller & Springer, 1972). Thus, interruption of the preservation process in some cases might render information in memory irretrievable only temporarily rather than permanently.

McGeoch (1932) argued that forgetting reflected a temporary lapse in remembering rather than a storage problem. In his view, other material that is also being processed at the same time to-be-remembered information is being processed causes interference. The other information competes with the to-be-retrieved information, and thereby leads to response failures. For instance, Underwood (1957) showed that the more previous experience subjects have had in learning lists of words, the worse performance is on the current trial, an effect that has come to be known as proactive interference. Melton and Irwin (1940), however, argued that more than just response competition seems to be driving these response failures. To demonstrate, in an A-B A-D experimental paradigm, they had control subjects learn A-B paired associates and then do nothing else for 30 minutes. Those in the experimental condition also learned the stimulus pairs, but in a second phase of the experiment they learned A-D paired associates. The experimental group was also divided into subgroups, depending on the number of practice trials (5 to 40) they received. Finally, all groups were then tested for their memory of the A-B pairs. The key prediction from the response competition formulation of interference theory is that the number of B response errors made should be equal to the number of D response intrusions for the experimental group. Stated differently, the amount of retroactive interference should equal the number of intrusion errors. As expected, the performance of experimental group was worse than the control group, thereby demonstrating a retroactive interference effect. However, the number of intrusion errors fell short of the amount of retroactive interference observed, with intrusion errors falling to almost zero between trials 10 and 40. They surmised that some unknown factor, which they termed "Factor X", must be getting stronger as the number of trials increased. They concluded on the basis of these findings that subjects had unlearned the original A-B pairs. To account for these and other results, Postman and Underwood (1973) offered a two-factor theory of forgetting, which combined the ideas of response competition and unlearning in order to account for the effects of proactive and retroactive interference. As shall be seen later, many misinformation effect theorists seem to have adopted a two-factor theory of forgetting to explain performance decrements.

The early study by Loftus et al. (1978) demonstrated the effects of timing and retention interval manipulations on response accuracy (see Figure 1). To find out whether misinformation has a greater impact depending on whether it is presented immediately after the event or just before the final recognition test, the timing of misinformation presentation was manipulated along with retention interval, which was varied by giving the final test either 20 minutes, 1 day, 2 days or 1 week after the interpolated event. Additionally, subjects were provided with either consistent, misleading or no information about the stop sign during the interpolated phase. Based on the results, Loftus argued that providing subjects with consistent information enhanced their memory for the stop sign, as the consistent group performed better than the group receiving no information. In contrast, misleading information negatively affects memory, as the misinformed group performed significantly worse than the baseline group. Furthermore, presenting misleading information immediately after the critical event has less of an impact than presenting it just before the final test because its easier to introduce misleading information as memory fades. A closer look at the data, however, suggests that there are other possible explanations for the observed effects.

Figure 1. The effects of when misinformation is presented in relation to the critical event (immediate versus delayed) and retention interval (time between the to-be-remembered event and the final test) on response accuracy. During the interpolated phase subjects were given either misleading information, no information, or consistent information about the stop sign appearing in the original slide sequence. (From Loftus et al. (1978).)

First, consider the effect of misinformation when it immediately followed the critical event. When there is no delay between the interpolated event and the final recognition test, performance for misled subjects is poorer than that of subjects who received either consistent or no misleading information. Over time, however, performance climbed for the misinformed group. After one week has passed, misled subjects were performing a bit better than subjects who had received no information and their performance was more similar to the consistent group than it had been earlier. Performance should not improve over time if misleading information overwrites memory traces. These data suggest there is response recovery over time, thereby providing evidence against the updatable memory hypothesis. On the other hand, this might be an illusion. Memory for the misinformation might have decayed over time, leading subjects to perform at chance levels on the final recognition test. Additional evidence is in order.

Chandler (1991) also found that over time the performance of subjects improves. In a manner similar to the misleading post-event design, she presented a list of pictorial stimuli in the critical phase of the experiment. In the interpolated phase, a list of items that were highly similar to those subjects encountered in phase 1 were presented. Accuracy on the modified test was significantly less for misled items than for control items when the test was given after a short retention interval. However, at a longer retention interval, performance on misled items did not differ from control items. Windschitl (1996) was also able to replicate the effect with faces as the to-be-remembered stimuli. Other experiments in the facial memory domain have demonstrated that performance decrements in facial identification accuracy that typically occur after giving a verbal description of a face are eliminated with time, an effect that is called "release from verbal overshadowing" (see Miessner & Brigham, 2001 for a review). Thus, it seems as though there is much evidence to support a retrieval based explanation of the Loftus et al. (1978) retention interval results.

The second result of interest from Loftus et al. (1978) (see Figure 1) is that when misinformation is provided immediately before the final recognition test, performance for the misled group is worse than that of the groups receiving either consistent or no information during the interpolated phase. An interesting aspect of these results is that the performance for the misled and consistent groups appears to be fairly symmetrical around the no information (or baseline) group at every retention interval. Such symmetry would not be expected if misinformation overwrites memories. In other words, consistent information seems to improve performance just about as much as inconsistent information decreases it. Loftus (1979) argued that memory for the stop sign was improved by presenting information about the stop sign during the interpolated phase. However, the increase in performance observed for the consistent group relative to the baseline group might indicate that subjects who did not encode the stop sign during the critical event accepted the information about the sign on the experimenter's questionnaire. Encoding failure can also partly explain the decreased performance observed for the misled group. Additionally, about a third of Loftus' misled subjects reported in a follow-up study that they remembered seeing the stop sign, but accepted the misinformation and reported it on the final questionnaire. Thus, encoding failure coupled with misinformation acceptance might fully account for the misinformation effects observed.

Another notable aspect of these results is that they suggest that lesser contiguity between the critical and misleading stimuli produces greater memory impairment. However, from a memory consolidation point of view, this result is surprising because misinformation should affect memory storage all the more the closer in time it is to the critical event.

Belli, Windschitl, McCarthy and Winfrey (1992) also argued that the misinformation effect should be more likely with long rather than short retention intervals. They noted that with adult subjects, the modified test results in equal performance rates for misled and control items, no matter whether the retention interval is long or short (McCloskey & Zaragoza, 1985; Ceci, Ross & Toglia, 1987; Loftus, Donders, Hoffman & Schooler, 1989; Bonto & Payne, 1991). Children, however, perform less well on misled items compared to control items, but only when the retention interval is long (Loftus et al., 1978; Belli, Windschitl, McCarthy & Winfrey, 1992; Ceci, Ross & Toglia, 1987; Toglia, Ross, Ceci & Hembrooke, 1992). They reasoned that the crucial factor controlling the difference in performance between adults and children was that experimenters arrange better stimulus conditions for children. They argued that since many of the adult studies not finding a misinformation effect at long retention intervals used slides that presented the crucial items in the periphery of the scene, encoding might be "less than perfect" for adults compared to children.

To determine whether a misinformation effect could be produced using the modified test under better encoding conditions, they created stimulus materials in which the crucial items were centrally presented on the slides. As they had hypothesized, a misinformation effect was not found at a retention interval of 15 minutes (accuracy for both the misled and control items was 94%); but, one was produced at the 5 day and 1 week retention interval. They proposed a storage based explanation for their findings: The bonds that connect the features of memories together weaken over time. The introduction of misinformation causes additional weakening, which results in worse performance on misled items at long retention intervals. However, given that the misinformation was presented approximately 10 minutes before subjects were given the memory test, response blocking might also explain the results.

Given that the misled items were presented just before the final test, they might have been stronger competitors relative to the correct information that was presented many days before. Models of blocking propose that competition robs activation from the target, and as a result, the competitor is sampled rather than the target (Rundus, 1973; Bjork & Bjork, 1992). The observation that no misinformation effect is found for the 15 minute retention interval is not necessarily problematic for the blocking account. At the immediate retention interval, the strength of the original and misled items might have been equal, especially since care was taken to produce memorable stimuli. The finding of the misinformation effect at a short retention

Figure 2. Critical slides used by Loftus et al. (1978) and Belli et al. (1992), presented on the left and right, respectively.

interval by Loftus et al. (1978) supports this claim. In her slide sequence, the yield sign (or stop sign, depending on condition) was located in the periphery and partly hidden from view by a tree branch (see Figure 2). Thus, weak encoding of the critical item (and the fact that the misled item was presented more recently than the original item) might increase the likelihood that the misled item is produced on the final test.

An additional issue to consider from Belli et al. (1992) is the analytical method they used to determine whether misleading items had impaired memory. They reasoned that memory for control items could be estimated by the formula p(correct) = R + .5 (1 - R), with R equal to the proportion of items remembered in the control condition and .50 equal to the guessing rate. To estimate memory for misled items, they used the same equation, but of course substituted in the proportion of misled items remembered for R. Based on their data, they concluded that since the misled condition resulted in 16% and 33% fewer correct responses compared to the control condition at the 5 day and 1 week intervals, respectively, there was evidence for the memory impairment hypothesis. However, as noted by McCloskey and Zaragoza (1985), there is good reason to suspect that the rate of selecting the original response when the original information is forgotten (or not encoded in the first place) will not be equal to .50 for those in the misled condition. Instead, the rate of selecting the correct item might be less than .50 because subjects who do not remember the original information might select the incorrect item if they remember that it was mentioned during the interpolated phase.

In order to resolve these issues, what is needed are baseline measures of accuracy for the items presented during the critical and interpolated phases. Without such measures, whether items presented during the interpolated phase produce "stronger" memories compared to the items shown in the original event all other things being equal will remain an issue. Not one of the event memory studies found for the meta-analysis presented later, however, provided a baseline measure of accuracy for the information presented during the interpolated phase. Nevertheless, in the meta-analysis that will follow, an attempt was made to determine whether misleading and consistent information produce symmetrical changes about baseline accuracy for the original information. If memory is being updated by misinformation, a greater change from the baseline might be expected for the misled condition compared to the control condition.

Central Versus Peripheral Details. Whether using simulated crime sequences or mundane scenarios as the to-be-remembered stimuli, experiments on the misinformation effect provide information about event details that are relatively peripheral to the main action being described. Subjects are misled about such objects as hardware, magazine covers, and soft drink brands, rather than being told that the man who got hit by a car was really a woman or the red Datsun that hit the pedestrian was really a yellow school bus. Dritsas and Hamilton (1977) (see Loftus, 1979) questioned whether it was easier to mislead participants on peripheral details rather than on central details, which they defined as stimuli that are easily perceived. To investigate, they presented during the interpolated phase a questionnaire that had misleading suggestions for both central and peripheral details. The critical event was three fairly graphic industrial accidents shown on videotape. In one, a man is hit in the eye by a flying metal chip, in another a man is thrown to the ground after being struck in the back by a spinning metal rod, and in a third, a woman gets her hand caught in a punch press. Centrality was determined by having another group of people rate how central or peripheral the event items were. The results demonstrated that accuracy for control items was higher for central (M = .81) rather than peripheral details (M = .47). As for the misled items, accuracy for central details (M = .53) was higher than accuracy for peripheral details (M = .31). Although the manner in which Dritsas and Hamilton (1977) measured centrality might be questioned (i.e., since the accuracy rate for their control central items was lower than the accuracy rate for the control group in Loftus et al. (1978), participants who rated the details might have had their own ideas of what centrality means), these data lend some support the notion that it is harder to mislead people on original event details that are encoded.

In other areas of eyewitness memory research, centrality has been defined in relation to accuracy (see Christianson, 1992). Better performance might be expected on central items because attention is directed to them, thereby leaving less attentional resources for processing peripheral aspects of the scene. For instance, research on the weapon focus effect has shown that participants remember fewer peripheral event details when a weapon is present compared to when one is not (Kramer, Buckhout, & Eugenio, 1990; Pickel, 1998; Maas & Kohnken, 1989). Alternatively, post-stimulus elaboration, which refers to thinking about information after it has been presented, might also account for enhanced recall if central items are thought about more often than peripheral ones (Christianson, 1992; Loftus, Hoffman & Loftus, 1991). This idea is similar to the reasoning behind the repetition effect, which is that simple repetition of an item leads to better recall performance.

Putting aside the issue of what psychological mechanisms give rise to better memory for central items, the storage based and retrieval based views of the misinformation effect make different predictions regarding the effect of memory strength for the critical event details on forgetting over time (see Figure 3). It seems clear that memory cannot be overwritten by postevent information if the traces for the critical event are not laid down in the first place. In other cases, however, an item might have a "weak" memory representation, owing to the fact that not much attention was directed to it during its presentation or because the duration of exposure for the item was too short. Both the updatable memory hypothesis and the blocking hypothesis would predict that increasing the strength of memory for the original item would reduce the size of the misinformation effect. Blocking argues that response competition in this case will not be strong, and the critical item will be retrieved. According to the updatable memory hypothesis, weak

Figure 3. Illustration of the blocking hypothesis compared to the updatable memory hypothesis for the retrieval of critical items encoded at different strengths over time. Both blocking and the updatable hypothesis predict that if memory for the critical item is strengthened (see gray line versus black line, which represent items that originally had strengths of .75 and .85, respectively) the size of the MI effect will decrease. However, blocking predicts (red line) that over time the size of the misinformation effect will decrease, whereas the updatable hypothesis predicts (green line) that the misinformation effect will be more stable over time.

memory traces are more likely to be overwritten. Therefore, if the original item is strong, the critical item will not be overwritten and it will be retrieved. However, the updatable memory and blocking hypothesis make different predictions about response accuracy over time. The updatable memory hypothesis predicts that the critical information will be permanently lost. From a blocking point of view, however, the competing information has a lower probability of being chosen over time because it is losing strength. Thus, the blocking account predicts that a correct response will be more likely with time.

The strength of the interpolated items relative to the critical items is an important issue that has not received attention in the misinformation effect literature. For instance, duration of exposure to each of the items presented on the narrative or questionnaire during the interpolated event is never carefully timed. Subjects read the questionnaire or the narrative however they want to read it within the allotted time (typically 5 or 10 minutes). If most critical items in studies of the misinformation effect are peripheral, then strength of memory might be relatively weak compared to memory for interpolated items--especially if one imagines the possibility that some subjects might read and then re-read the information on the narrative. In comparison, subjects typically see the critical slide during phase 1 for only 3 to 5 seconds. Research by Mandler and Johnson (1976) suggests that longer exposure durations might be needed. In this study, memory for pictures was tested using a Yes/No recognition test. The test included target and lures, which were exactly the same as the pictures studied except that one of the items in the picture was substituted for another of the same size, shape and conceptual class. They found that a presentation rate of 5 seconds resulted in a correct recognition rate of .46 on average. The rate of correct recognition significantly improved when the study phase was longer than 5 seconds. Moreover, Loftus et al. (1978) performed a follow-up experiment to determine whether subjects were encoding the critical details from the slide sequence. Subjects were shown the event sequence and then asked to draw the event details on a diagram. Only about half of the subjects included the critical traffic sign in their diagram.

This also raises the issue of how does one define central and peripheral items on a written narrative or questionnaire? If centrality is defined in terms of performance, then baseline accuracy measures of performance are needed for both the critical and interpolated event to demonstrate equal rates of performance (i.e., memory strength) for the critical and misled items presented in different formats (i.e., pictorial and written). To date, no misinformation effect study has measured baseline performance for items presented during the interpolated event. Furthermore, if central details are defined as those occurring in the middle of the slide (Belli et al. 1992), what is the equivalent of that on the written questionnaire or narrative? On some level, answering a question makes the item central at least momentarily and everything written in the narrative could be construed as "important" or central from a subject's point of view.

Retrieval Effects. A retrieval failure view of forgetting takes the position that performance decrements arise when the cues available at test lead to the retrieval of the wrong information. For instance, research on "permastore" memories illustrates that memory performance is enhanced by cued recognition tests, which unlike recall tests, provide important cues that can lead to impressive rates of accuracy. Even with long retention intervals of up to 50 years, people can identify high school classmates (Bahrick, Bahrick & Wittlinger, 1975), retain Spanish vocabulary (Bahrick, 1984) and remember city streets and locations (Bahrick, 1983) with a high degree of accuracy. The stimulus conditions, however, under which these types of permastore memories are acquired are arguably different from those found in experiments on the misinformation effect or in real eyewitnessing situations, in which events typically last only minutes (Flowe & Ebbesen, 2001) and happen only once. If it could be demonstrated in studies using the misinformation effect testing paradigm that the effect still persists when the appropriate cues are available at test, a retrieval based explanation would seem less plausible compared to one that posits that stored information is being overwritten.

Bowers and Bekerian (1983) performed one such test using the Loftus et al. (1978) stimulus materials. They found that the order in which the items were presented on the final test influenced memory for misled items. A misinformation effect was obtained if the items were presented in random order, but not if the items were presented sequentially. The order of the items purportedly served as a cue aiding in the recovery of the original item. In a follow-up study, Bekerian and Bowers (1984) replicated these findings and also discovered that the appearance of their testing effect was conditional on the order in which the inconsistent items were presented during the interpolated phase. A misinformation effect was not produced in the sequential test condition if the misinformed items were ordered randomly on the interpolated questionnaire. However, if the misinformation was introduced sequentially on the interpolated questionnaire followed by a sequential test, a misinformation effect was obtained. Kroll, Ogawa, and Neiters (1988) also found evidence of response recovery when the original order of events was reinstated prior to taking the final test. Subjects were shown a slide sequence twice and tested each time. Those who saw the second slide sequence in a sequential rather than a random order were more likely to correct any errors they made on the first test. These results suggest that sequence is an important cue used to differentiate information learned from the to-be-remembered and interpolated event. This evidence also provides support for the view that both misleading and original event information co-exist in memory because it shows that misinformed subjects will correctly recall details of the critical event provided that there are cues to help them at test.

Other retrieval cue manipulations have produced similar results. Ables and Morton (1999) hypothesized that reinstating at test the modality in which the to-be-remembered event had been presented would provide retrieval cues that would eliminate the misinformation effect. To test for this possibility, they used a reverse testing procedure, which has been used by other investigators to determine whether misinformation overwrites existing memories (e.g., Lindsay & Johnson, 1989). The reverse testing procedure involves presenting a misleading narrative to subjects first, followed by a to-be-remembered event and then a final memory test. The logic behind the using the procedure is that if misinformation overwrites existing memory traces, then presenting misleading items before the critical event should have no effect on memory. Lindsay and Johnson (1989) found a misinformation effect using the reverse testing procedure along with a final yes/no recognition test, and concluded that information from different sources might become integrated in memory. Ables and Morton (1999), however, manipulated the format of the final test and found that misleading information had a differential effect on memory, depending on whether the test was written or pictorial. If the test was pictorial, no misinformation effect was found. However, if the test was written, subjects performed worse on misled items than on control items. The updatable memory hypothesis has difficulty accounting for these findings because performance on misled items seems to have hinged on whether pictorial or verbal cues were available at test. Similar effects can be found in other reports, in which investigators have held constant the modality of the information presented along with the format of the test (Bowman & Zaragoza, 1989; Toglia, Payne, & Anastasi, 1991). Another outcome of the Ables and Morton (1999) experiment was that written probes produced a higher hit rate for consistent items (items that appeared in the to-be-remembered picture and narrative) compared to baseline items (items that appeared only in the to-be-remembered picture). In contrast, consistent items and baseline items were remembered equally well if a pictorial probe was used. They provided an dual-encoding interpretation of these findings, arguing that pictorial probes access pictorial representations, whereas verbal probes can access either pictorial or verbal representations. These results further support the idea that subjects exposed to misleading information can remember the critical item if they have cues that allow them to access it in memory.

Researchers have also examined retrieval cue effects in a more general way by examining how contextual cues influence memory for misled items. Context effects have influenced memory reports in other research domains. For instance, unfamiliar faces are remembered better if the recognition test is given in the same context in which the faces were studied (Shapiro & Penrod, 1986; Cutler, Penrod, & Martens, 1987; Gibling & Davies, 1988; Krafka & Penrod, 1985). The effect of testing subjects in a context different from which the misinformation was acquired, however, has produced mixed results. Bonto and Payne (1991) ran each phase of their misinformation effect experiment in either different rooms or in the same room. They obtained an overall misinformation effect on the standard test but not on the modified test. However, although misled and control subjects performed more similarly when the contexts were dissimilar, they did not find a significant context effect. Lindsay (1990) did find a context effect, but due to confounds, such as retention interval, which also varied between conditions, the validity of the results is compromised.

Taken together, these experiments examining how characteristics of the testing environment influence the misinformation effect provide some strong evidence against the updatable memory hypothesis, as they suggest that misled subjects' memory for a critical event detail can be "recovered." Encoding of the original event appears to be multidimensional, in that the stimuli encountered at study are associated with environmental cues that are available in the experimental situation. If the conditions at test exclude cues that would allow subjects to discriminate between the original and interpolated event, then permanent forgetting will seem to have occurred. Forgetting in the studies reported here seem to be characterized best as a temporary memory lapses rather than as a permanent losses.

Age. Similar to the studies with adults, misinformation effects in children are less reliable on the modified test compared to the standard test (for a review see Zaragoza, Dahlgren, & Muench, 1992). On the modified test, misinformation effects with children are more likely to be found at longer instead of shorter retention intervals (see Belli et al. 1992 for a review). Older adults are also more likely to base their answers on misinformation than young adults, though this effect is inconsistently reported (see Ceci & Bruck, 1993 for a review). Since most of these studies present the misleading information immediately before the final test, one explanation for these findings is that children and older adults might be less hesitant to accept misleading suggestions compared to young adults (Dempster, 1992). For instance, Jacoby (1999) tested the memory of young and older adults for paired associates using a cued word fragment completion task. Before each test item, a valid prime (i.e., same word as the stimulus word seen earlier), an invalid prime or no prime was presented. The results showed that older adults were more likely to base their answers on an invalid prime compared to young adults. Older adults also showed a larger positive priming effect when a valid prime was presented. However, if participants engaged in a divided attention task during the study phase, age effects were eliminated. These effects seem to further support the notion that misinformation alters responses rather than memories.

Payne, Toglia and Anastasi (1994) completed the only meta-analysis to date of studies that have used the misleading post-event paradigm to study memory impairment. The intent behind their work was to determine whether response accuracy and retention interval length influenced the size of the misinformation effect. The results (based on 48 cases from 11 research articles) showed that misleading suggestions significantly impaired performance on both the standard and modified tests. On the modified test, they reported that the mean recognition level in the control group (75.8%) was significantly greater than for the misled group (71.7%). Furthermore, studies using longer rather than shorter retention intervals were more likely to show impaired performance for misled items. This finding emerged because performance was found to be higher for control items in the studies that used long compared to shorter retention intervals. They argued that this result suggests that investigators alter the stimulus conditions in the long interval studies in order to prevent floor effects. Based on the totality of their results, they concluded that the misinformation effect is a "real memory phenomenon."

These results, however, far from settle the issue. Any inferences made based on their data might be seriously flawed because age of subject and retention interval are confounded. Child subjects contributed to 64% (n = 8) of the data points in the long retention interval studies and to only 17% (n = 5) of the data points in the short retention interval studies. Re-analyzing their data by age showed that there are differences between adults and children as a function of retention interval on the modified test. The average effect size for the adult studies using a short retention interval was .03 and for children it was .02. However, in the long retention interval studies, the effect size for adults was .03 and for children it was over three times larger. Therefore, the long retention interval effects noted by Payne et al. (1994) might have to be qualified by age of subject.

An additional problem with their conclusions is that they did not separately analyze the data from Chandler (1989, 1991). These data might have overly influenced their results. Both of Chandler's reports use lists of unrelated pictures as stimuli in both the critical and interpolated phase. The remaining studies in the meta-analysis portrayed a to-be-remembered event. Even though only 34% of the adult data is based on Chandler's studies, her data account for 54% of the difference between control and misled items on the modified test with adults. Specifically, the average difference between control and misled performance is .05 if based on Chandler's data, and .02 if based on the event studies. Given that the intent of most misinformation effect research is to draw conclusions about the behavior of real eyewitnesses, the fact that the type of to-be-remembered stimuli might influence the size of the misinformation effect is important information for psychologists who testify in the courtroom about the effect to know about.

The following meta-analysis analyzed the standard and modified test data, grouping the analyses based on age and type of to-be-remembered stimuli. Furthermore, the effect of retention interval on the misinformation effect was examined using response data from the standard and modified tests in order to test key predictions made by the blocking and updatable memory hypotheses. In keeping with the criteria used by Payne et al. (1994), a short retention interval was defined as that shorter than or equal to 60 minutes.

Performance data from 95 cases reported in 34 research articles that used the misleading information experimental paradigm were collected by entering the terms "misinformation effect" in the PsychInfo Database and by querying the Social Science Research Citation index using the Loftus et al. (1978) and the McCloskey and Zaragoza (1985) reports as search terms. In order for a study to be included, recognition performance had to be assessed using either the standard and/or modified recognition test. (Studies that used a Yes/No recognition test or that measured performance based on the mean number of control and misled items endorsed by subjects were excluded.) A total of 45 cases had adult subjects witness some sort of event (11 used only the modified test, 22 used only the standard test and 12 studies used both) For children, 20 cases were found, all of which used events as to-be-remembered stimuli (16 used only the standard test, 3 the standard test, and 1 used both). Lastly, a total of 30 cases were from studies that used pictorial stimuli as the to-be-remembered event (26 used only the modified test, and 4 used both the modified and standard test). All of the pictorial cases involved adult subjects.

If misleading items are altering the content of memory, then one prediction is that there should be a positive correlation between the size of the misinformation effect on the modified test and standard tests. If the memory alteration account is correct, then low performance rates on misled items relative to control items on the standard test should be accompanied by lower performance rates on the modified test. However, for the adult studies that used both the standard and modified test, there was not a significant relationship between peformance on the standard and modified tests (r = -.33, p > .05, N = 12). Given the relationship was negative, these results might suggest that subjects' responses were biased on the standard test rather than their memories. (The relationship could not be investigated for children, as there was only one child study that used both the standard and modified test. Similarly, there were only 4 pictorial stimuli studies that used both the standard and the modified test; however, there was some evidence for a positive relationship between peformance level on the standard and modified test (r = .27, p < .05) in the picture studies.)

Figure 4 displays mean performance found for the control and misled conditions on the standard and modified tests. As shown, the difference between control and misled performance is larger overall on the standard test compared to the modified test. For the adult event memory studies, performance for the misled group on the standard test--but not on the modified test--was negatively affected (Standard M Difference = .28, SEM = .018, p < .001, r = .41; Modified M Difference = .01, SEM = .009; p = .14, r = .24). For the child event memory studies,

Figure 4. Response accuracy on the standard and modified test as a function of stimulus type and age.

performance on both the standard and modified tests was significantly affected by misinformation (Standard M Difference = .21 SEM = .018, p < .001, r = .94; Modified

M Difference = .07, SEM = .018, p < .01, r = .69). Maybe there is some brain based developmental explanation for these results (e.g., the frontal lobes, which have been implicated in facilitating source monitoring judgements, are still developing). On the other hand, these findings could suggest that children are more likely to go along with the misinformation presented during the interpolated phase compared to adults.

For the picture memory studies, the control and experimental conditions were significantly different on both the standard (M Difference = .17, SEM = .019, p < .001, r = .98) and the modified (M Difference = .04, .001, p < .001, r = .999) tests. One possible reason why the misinformation effect is found on the modified test for the pictorial studies but not in the event memory studies is that there is more room for social influence to bias the results in the latter. The modified tests controls for these biases, and as a result, subjects in event memory studies select the correct response on the modified test. In contrast, a real memory phenomenon of some sort, whether memory overwriting or interference, might be occurring in the pictorial studies. The basis for this claim is that the pictorial studies included in the analysis used experimental designs that are highly similar to the A-B A-D design used in the retroactive interference research tradition. In both the critical and interpolated phases, lists of nature scenes (Chandler 1989, 1991, 1998) or faces (Windschitl 1996) were presented. Social influences that can bias responding seem like they would be minimal in the pictorial stimuli studies compared to the event memory studies. In the event memory studies, misinformation is introduced by way of humans: Subjects read a questionnaire written by the experimenter or read a narrative written ostensibly by another subject (or some other person, depending on what the cover story is). After using the modified test to control for these biases, memory apparently recovers in the event memory studies. This reasoning, however, is highly speculative. If demand characteristics are really driving the observed difference between the pictorial and event studies on the modified test--and not some other differentiating characteristic of the stimulus conditions--is an empirical issue awaiting an answer.

Payne et al. (1994) presented evidence that the misinformation effect was increasingly likely as performance in the control condition increased (r = .54, p < .001). Looking at the data by age and stimulus type, however, alters this conclusion. There was no significant relationship between control response accuracy and effect size for adults (r = .30, p = .17) or for children (r = .37, p = .13). If pictures are the to-be-remembered stimuli, however, there is a significant relationship between control accuracy and the size of the effect (r = .57, p < .001). The correlation between control performance and effect size suggests that another factor might be involved. One hypothesis is that retention interval is influencing the effect. For the studies that used a short retention interval, the blocking account predicts that performance on control items will be higher than performance on misled items. However, performance is expected to recover if a longer retention interval is used. As a result, control performance at a short retention interval will be strongly related to effect size, while for longer retention intervals, control performance will be weakly related to effect size. This prediction received support. Looking just at the short retention interval conditions showed that control performance and effect size were significantly related (r = .52, p < .05, n = 21). The control-effect size relationship was weaker at longer retention intervals (r = .37, p = .33, n = 9).

To investigate this relationship further, additional analyses were conducted for the pictorial stimuli studies. As suggested by the above analyses, the size of the effect was larger on the modified test if a relatively short compared to long retention interval was used (Short Interval M Difference = .06, SEM = .007, n = 21; Long Interval M Difference = .00, SEM = .011, n = 9), F (1,28) = 21.14, p < .0001, r = .65.)

The nature scenes used are arguably unique stimuli. Compared to faces, contact with these stimuli outside of the laboratory during the retention interval might happen less frequently (see Figure 5 for examples of the stimuli used in the nature and face studies). Faces seen outside of the experiment might affect performance at long retention intervals. According to the trace alteration explanation, seeing additional faces should impair performance to an even larger extent for experimental compared to control items, because the traces for the original items have been weakened by interpolated information. On the other hand, if control and misled accuracy is similar following a long retention interval, this would be evidence favoring an interference based explanation of the misinformation effect.

Figure 5. Example of stimuli used in Windschitl (1996) and Chandler (1989, 1991, 1998), on the left and right, respectively.

As shown in Figure 6, control accuracy for the nature scenes did not change between the long and short retention interval conditions (accuracy is 79% for both). If memory was tested right after the interpolated phase, accuracy on the misled items (73%) is significantly less than control items (79%). After a delay, however, the response seemed to have recovered, and accuracy for misled items returns to 79%--the same rate obtained for the control long and short interval conditions. For faces, a slightly different picture emerges. Control accuracy is not the same for the short and long retention intervals (76% and 69%, respectively). At the short retention interval condition, control performance is significantly better than misled performance (69%). At a longer retention interval, however, performance rates are identical for the control and misled items (accuracy is 69% for both), indicating response recovery. These results suggest that there was memory decay over the retention interval for faces but not for nature scenes. A possible cause for the decay could be other faces that the subject saw during the retention interval outside of the experiment. These results lend some additional support to the notion that interference is affecting the modified test results, not memory overwriting.

Figure 6. Effects of retention interval on accuracy for misled and control items on the final test by stimulus type.

Additional effects of retention interval were investigated for the adult and child event memory studies; but, the results are highly preliminary because the number of studies available for analysis was small in number. For the adult event studies, the misinformation effect on the modified test was larger at longer retention intervals compared to shorter retention intervals (Long M = .06, SEM = .025; Short M = .00, SEM = .009), F (1, 23) = 3.96, p = .06). These results are open to a response blocking explanation, however, because all but one of the long retention interval studies presented the misinformation just before the final test. There was some evidence that timing of the misinformation had an effect in the long retention interval studies (though confidence in this evidence is minimized by the fact that the number of studies is small). If the misinformation was presented just before the test, the size of the effect was larger (M = .09, n = 2) than if it was presented in closer contiguity with the to-be-remembered event (M = .00, n = 1). (This was also true for the standard test: Before Test M = .47, n = 2; Closer to Phase 1 M = .25, n = 1.) For children, longer retention intervals also produced a larger effect compared to shorter retention intervals (Long M = .10, SEM = .02, n = 12; Short M = .02, SEM = .03, n = 6), F (1,16) = 5.09, p < .05. However, for children, every long retention interval study presented the misinformation just before the final test.

A final issue concerned whether consistent items enchanced performance while misled items reduced peformance at similar rates relative to the baseline. Across the adult event studies, the control items were either referred to either in neutral language (e.g., the screwdriver seen during phase 1 was referred to as a "tool"), consistently named (e.g, "screwdriver") or not mentioned (i.e., accuracy for phase 1 items in the absence of further information served as a baseline measure) in the interpolated phase. Figure 7 displays the performance rates for the studies using each type of control condition. The size of the misinformation effect did not differ by the type of control used, F(1, 31) = 2.19, p = .12 . The size of the effect appeared to be somewhat greater, however, for the studies studies that used either consistent or neutral control items compared to those that did not mention the critical control item during the interpolated

Figure 7. Accuracy on the standard test by the type of control item.

phase (p = .05). It is also surprising that the studies referring to the control item in consistent terms did not result in greater accuracy compared those that did not mention the item or instead reffered to it in neutral terms.

Figure 8 displays the results from studies that included data from consistent, baseline and misled conditions. The outcome measure of interest was accuracy on the standard test. First, the pattern of results does not seem to support the contention that larger misinformation effects occur when accuracy for control items is high. Second, there is no apparent symmetry about the baseline for consistent and misled items, as 4 out of the 6 cases presenting consistent information did not enhance performance as much as misled items decreased performance. However, encoding failure coupled with response bias (since the data are all from the standard test) might explain the lack of symmetry.

Figure 8. Performance on the standard test for the 6 cases that reported baseline data for the original item and manipulated whether consistent or misleading information was presented during the interpolated event.

Overall, the results from the meta-analysis suggest that misinformation affects responses rather than an adult's memory for a to-be-remembered event. Age and type of stimuli influenced performance on the modified test and thereby qualify the conclusions reached by Payne et al. (1994). Only children performed more poorly on the critical event item on both the standard and modified test. For adults, however, peformance was affected only for pictorial stimuli, not event stimuli. Furthermore, the results found evidence for the occurrence of spontaneous recovery in studies using pictorial stimuli and longer retention intervals.

Procedural differences might explain why interference was observed for the pictorial studies but not for the adult event memory studies on the modified test. The pictorial studies more closely followed the A-B A-D design than the event memory studies. Though the typical misinformation effect paradigm is often thought similar to the classic A-B A-D retroactive interference design, the similarites might be only on the surface. First, the human element is more pervasive in the event memory studies compared to the pictorial studies. Misinformation is delivered using other people in the event studies. Social influence, therefore, can lead subjects to comply with the demands of the experiment and select the incorrect alternative on the standard test. When the misled item is not an option, peformance on misled items is equal to that of control items on the modified test.

Second, event memory studies do not tightly control for extraneous variables that might differentially affect memory for the control and misled items, such as duration of exposure, pre-experimental familiarity or where in the slides (or narrative) subjects focus their attention. Under these conditions, reliable misinformation effects on the modified test might be difficult to obtain across studies. For example, in Loftus et al. (1978), the results for the stop and yeild sign critical items are not separately reported. Even though type of traffic sign is counterbalanced across control and misled conditions, one might surmise that subjects who missed the critical detail in phase 1 would more readily guess that they saw a stop sign rather than a yeild sign. Afterall, in the real world, how often do yeild signs stand in guard of crosswalks? As such, because of pre-existing differences in familiarity, it seems as though control subjects would pick stop sign more often than not on the final test and that it would be easier to mislead subjects that the critical "yeild" sign was a stop sign. As another example, Loftus (1979) found that subjects exposed to "blatant" misinformation (e.g., a red wallet is referred to as brown) are less likely to report the misleading suggestion on the final test. An additional finding in this study was that subjects who encountered the false information early during the interpolated event were less likely to be misled by other incorrect details. Thus, the interference effects observed for the pictorial studies on the modified test might not have been obtained in the meta-analysis of event studies because of extraneous factors varying across studies.

Third, in the typical misinformation effect study, retrieval conditions are not the same as learning conditions. That is, the to-be-remembered event is a pictorial display and both the interpolated event and recognition test are usually written. In the pictorial studies, modality of presentation is held constant across all phases of the experiment. Unfortunately, no misinformation effect experiment has ever varied in a factorial design the presentation format of both the misinformation and the final test. The few studies, however, that have held presentation modality constant across all phases of the experiment have not found retroactive interference for prose stimuli (McGeoch & McKinney, 1934; Hall, 1955; Ausubel, Robbins, & Blake, 1957, Ausubel, Stager, & Gaite, 1968) or misinformation effects for event stimuli on the modified test (Bowman & Zaragoza, 1989; Toglia, Payne, & Anastasi, 1991). These findings suggest that the modality of presentation is not the best candidate for explaining why the misinformation effect is found for the pictorial studies but not for the event memory studies on the modified test.

Fourth, event memory studies using long retention intervals confound retention interval with the timing of the misinformation. Misleading information is presented immediately before the final test. As a result, it is not clear whether the misleading information blocks the retrieval of the original information or affects memory for the original item. In contrast, the pictorial studies with long retention intervals presented the interpolated information immediately after phase 1. Future studies will need to vary the timing of the misinformation to determine whether spontaneous recovery can also be observed at long retention intervals in the event memory studies. The results of Loftus et al. (1978) supports the expectation that spontaneous recovery will be found.

Loftus (1979) has argued that the spontaneous recovery of the original information happens in only some cases. She maintained that under certain conditions (though she did not specify exactly what those conditions are), fragments are not permanently stored, but rather altered by other information. Several theorists, including Loftus, have proposed that there is evidence to support the trace alteration view. This section reviews research evidence that has been proferred to show that memory traces are impaired by misinformation, including studies that have reportedly found memory impairment using more sensitive tests, studies that have claimed that memories can be implanted wholesale, and studies that have seemingly produced blend memories.

Testing Procedures. Whether the modified test is sufficiently sensitive to detect memory impariment was one of Loftus' key concerns following the McCloskey and Zaragoza (1985) report. She argued that other methods that have been used to measure impairment lent support to the updatable memory hypothesis (for a review of this research, see Loftus et al. 1985). For instance, the betting recognition test procedure allows subjects to assign probability estimates to each of the alternatives (usually 4 are presented) to indicate the likelihood that the item was shown during the critical event. Benzing (1985) (see Loftus 1979) tested subjects using the modified or standard procedure, but altered the test format such that subjects could bet on each of the response alternatives. Performance decrements were found on both the standard and modified test following exposure to misinformation. The size of the effect, however, was smaller for those taking the modified test. Benzing (1985) concluded that this was a better way to test for memory impairment, since subjects are able to weight their responses by confidence. Other research that has provided monetary incentive for providing a correct answer has reported similar results, with the misled condition being less accurate than the control condition (Loftus, 1979).

Another piece of evidence thought to support the updatable memory hypothesis comes from experiments that have measured response times and confidence on the final test (Tousignant, Hall, & Loftus, 1986; Loftus, Donders, & Hoffman, 1989; Bonto & Payne, 1989). The logic behind these experiments is that misled subjects should confidently respond with the incorrect answer and do so quickly on the final test if their memories had been altered. Loftus, Donders, & Hoffman (1989) found a significant misinformation effect on the standard test, and also showed that on misled items, subjects responded more quickly when they gave an incorrect rather than a correct answer. In contrast, subjects responded more quickly on control items when they were correct rather than incorrect. Response deliberation was proposed to account for these results: When subjects do not know the correct answer they ponder over the response choices. They argued further that when subjects gave an incorrect response to a misled item, they must have believed in their memory for the item because for misled items, subjects were as confident in their wrong answers as they were in their right answers.

On the modified test, however, no misinformation effect was obtained. Furthermore, subjects responded faster on control items compared to misled items. Subjects were also faster and more confident when their responses were correct rather than incorrect. In a second experiment, Loftus, Donders & Hoffman (1989) strengthened the neutral items (e.g., screwdriver was referred to as tool) and misleading items by presenting the interpolated narratives three times in an effort to produce a misinformation effect on the modified test. Additionally, the number of slides shown during the first experiment was reduced and each slide was shown for a longer period of time. They replicated the effects found with the standard test in the first experiment. However, a misinformation effect was still not produced on the modified test. They concluded the results overall supported the view that new memories had been created. They also maintained that the longer response times and higher accuracy rates obtained on the modified test were not problematic for their interpretation. For instance, misled subjects might encounter the response alternatives on the modified test in "perplexed bewilderment," since their memory for the original item had been erased and the novel distracter was not at all familiar.

However, these results might also be explained without invoking a memory overwriting explanation. Inferring from the bar graphs displayed in the report, for the first experiment, the average reaction times for misled items was 2000 msec and 2500 when subjects gave an incorrect and correct response, respectively. In the second experiment, however, when the information presented during both phases of the experiment was strenghthened, response latencies were less than those in the first experiment, with latencies of 1500 msec for wrong answers and 1700 msec for right answers. On the basis of these reaction time data, it seems as though a response blocking explanation can also accommodate the results. Strenghtening the information resulted in subjects answering more quickly overall in experiment 2. Furthermore, correct answers in both experiments took longer than incorrect ones, but the difference between the control and misled response latencies was smaller in experiment 2 by strengthening the items in both phases of the experiment. However, although the means are in the predicted direction, control performance was not significantly improved (68% correct for experiment 1 and 74% in experiment 2) and misled performance was not significantly reduced (36% correct in experiment 1 and 31% in experiment 2). Therefore, further research is needed to determine whether item accessibility influences retrieval times for misled items. (This research should manipulate the strength of the information presented during both phases and include baseline measures of accuracy for the original and misled events as well as a condition in which consistent information is provided during the interpolated phase.)

Tversky and Tuchin (1989) also criticized the modified test for a lack of sensitivity in detecting the misinformation effect. They presented participants with a slide sequence followed by a narrative that introduced misleading information. A Yes/No recognition test was used to obtain separate measures of memory for the original information, the misleading information and novel distracters, which had not been previously presented but belonged to the same general category as the original and misled items. Misled and control subjects were equally good at rejecting the novel distracters. Evidence of the misinformation effect, however, was found in that misled subjects were poorer than control subjects in rejecting the misleading distracters. Misled subjects were also significantly worse than control subject in recognizing the items that had been present in the original event. Furthermore, since memory for each item was tested separately, a sizeable percentage of misled subjects indicated that they recognized both the original and misleading items from the slides. Tversky and Tuchin (1989) concluded from these results that a retrieval based interpretation of the misinformation effect seemed feasible, but also acknowledged that their results could not rule out the possibility that original memory was overwritten.

Results for the Yes/No procedures are similar to source monitoring tests, in which subjects indicate for each item either whether they saw it during the critical event or read about it during the interpolated phase. Lindsay (1990) tested recall for control and misled items, manipulating the discriminability of the critical and interpolated information. He also used the "logic of opposition" procedure, which involves telling subjects just before the final test that the information they read during the interplated phase was incorrect. Subjects were warned that if they based their answers on the narrative information, their answers would be wrong. Under conditions in which it was easy for subjects to discriminate between the original and interpolated event, no misinformation effect was obtained. However, a misinformation effect was obtained when the original and interpolated event were harder to discriminate. These results suggest that the misinformation effect may be due in part to source confusion errors.

Increasing the perceptual similarity between memories has been shown to increase source monitoring errors. Johnson, Foley & Leach (1988) found that participants could differentiate words that they imagined themselves saying from words that were read by a confederate. However, participants were less accurate at distinguishing words that they themselves read from those that they had imagined hearing in a confederate's voice. Semantic similarity has also been shown to increase souce monitoring errors. Subjects find it difficult to attribute a statement to a speaker if they hear the speakers describe the same event rather than different events (Lindsay & Johnson, 1991). Based on data such as these, Johnson, Hashtroudi and Lindsay (1993) have argued that source attributions should be accurately made if memories are richly detailed and have unique source characteristics. Given that the misinformation effect event studies in the meta-analysis usually presented critical and interpolated stimuli that differed on many perceptual dimensions (e.g., the critical event is pictorially presented and and the interpolated event is written), it seems somewhat enigmatic that subjects would not be able to discriminate between the information presented during the two phases. One explanation of why this is so is that the original items are being being encoded differently than the interpolated items.

There is also additional evidence that memory for the critical and misled items might co-exist. Performance on implicit memory tasks is not differentially affected by misleading information (see Loftus, Feldman & Dashiell, 1995). Although misled subjects perform worse on the final explicit memory test compared to control subjects, priming effects for the original items (relative to baseline performance) have been obtained for both misled subjects and control subjects. These results suggest that memory for the critical items was not completely overwritten.

While the results presented in this section suggest that the misinformation effect can be obtained using other testing procedures, a retrieval based interference explanation is still a viable alternative in every case. An important question for future research is whether there is response recovery over time.

The Implanting of New Memories. Historian and Pulitzer Prize winner Dr. Doris Goodwin was recently accused of plagiarizing whole passages in her book "The Fitzgeralds and the Kennedys". Perhaps she unknowingly failed to credit the original authors because she incorrectly attributed the passages to herself. In investgations of "cryptomnesia", or unconscious plagiarism, people are more likely to falsely take credit for the words of others if they are concentrating on what they are going to say or distracted by other task demands (e.g., Marsh, Landau, & Hicks, 1997). Similar to other findings in the source monitioring area, perceptual similarity and retrieval cues have been found to affect the occurrence of cryptomnesia (e.g., Macrae, Bodenhausen, & Calvini, 1999). However, whether people can come to remember the correct origin of the information, whether the true source of the information was encoded to begin with, or whether people's memories are overwritten by their belief that they authored the information are issues that have yet to be resolved in this research area.

UFO abductions, encounters with angels, and recovered memories of childhood sexual abuse are other examples of types of false testimonials rendered by people who have come to believe in "false memories". To demonstrate that false memories could be implanted for childhood experiences that never occurred, Loftus and Pickrell (1995) interviewed the relatives of 24 subjects, asking them to recall information about 3 events that took place during the participant's childhood. Relatives were also asked whether the subject had been lost in a shopping mall at the age of 5 and to provide information concerning a plausible shopping trip to the mall that the participant might have taken as a child. Participants were then given a booklet containing information about three real childhood events and a false event--getting lost in a shopping mall at the age of 5. They were asked to describe each of the events or to indicate "I don't remember this." The following is an example of one of the false incidents detailed to subjects, "You, your mom, Tien and Tuan, all went to the Bremerton K-Mart. You must have been five years old at the time. Your Mom gave each of you some money to get a blueberry ICEE. You ran ahead to get into the line first, and somehow lost your way in the store. Tien found you crying to an elderly Chinese woman. You three then went together to get an ICEE."

A total of 7 participants indicated that they remembered the time when they were lost in a shopping mall. During the first follow-up interview 1 to 2 weeks later, 1 of the 7 misled participants recanted, saying that she did not remember the episode. At the final interview, 19 out of 24 subjects picked the "lost in the shopping mall" episode when asked to indicate which of the events was a false memory. Loftus and Pickrell (1995) concluded that the demonstration was an "existence proof" that false memories for childhood events could be implanted. Crook and Dean (1999), however, in their critique of the experiment, were on the mark when they countered that the experimenters' results were "an example of behavior consistent with false memory formation", rather than proof that false memories could be implanted.

That some people were seemingly convinced they were lost in a shopping mall is not all surprising. Afterall, people often rely on and trust their relatives to share with them correct information about their childhood. Furthermore, the critical feature of the event itself--getting lost--might have happened to participants in other places besides a shopping mall or at other ages (neither participants nor relatives were asked whether the participant ever experienced being lost at any age as a child). As such, participants might have made a source attribution error, correctly recalling they were lost at some point during their childhood, but falsely recalling exactly where or at what age they were lost. Another possible problem that complicates the interpretation of these results is that participants might have been lost in a mall at the age of 5, yet their relatives might not have remembered. Since there is no way to independently verify that the participants were never lost in a mall--or to find out whether the participants were trying to appease their relatives and the experimenters by falsely reporting that the event happened--the conclusion that memories were implanted by the experimenters seems highly premature, and is only one out of several competing explanations for the participants' reports.

Other research has suggested that the plausibility of the to-be-implanted event is an important factor influencing whether subjects come to report events that never happened to them. Pezdek (1995) found that although 3 out of 20 subjects falsely reported being lost in a shopping mall, none of the subjects were convinced by the experimenters that they had received a painful enema from a parent as a child. Furthermore, given that the majority of participants in Loftus and Pickrell (1995) and in Pezdek (1995) did not come to endorse false childhood events, it is plausible that individual differences in suggestibility might account for these results (e.g., Loftus, Levidow & Duensing, 1992).

Blend Memories. Research on the suggestibility of color memory has found that subjects will sometimes report that the color of an object presented during the critical and interpolated event was a composite blend of the two presentations (Loftus, 1977; Braun & Loftus, 1998). Furthermore, if subjects are given the opportunity to pick another color, they will pick the color of the misleading item as their second choice (Loftus, 1979).

Loftus, Schooler and Wagenaar (1985) have argued that these results can be extended to classes of objects. In other words, if some object in the world existed as a composite blend of a yeild sign and a stop sign, misled subjects would select it over other response options on the final test. To support their argument, they cited the report by Weinberg, Wadsworth and Baron (1983), in which a sequence of slides featuring a car stopped at a yellow "yeild" sign was used as the critical event. Following the interpolated phase, subjects tested with the standard test were given a yellow stop sign and a yellow yield sign as response alternatives. Those tested with the modified test were given a yellow "yeild" sign and a red "yeild" sign as response options. Misled subjects performed worse on both the standard and modified test compared to those who did not receive inconsistent information.

As noted by McCloskey & Zaragoza (1985), however, these findings could indicate that subjects deliberately choose a compromise response to comply with the demands of the experiment. Furthermore, research on illusory conjunctions, or other phenomena such as the McGurk effect, probably do not reveal anything about whether memory can be overwritten by misleading information. The basis for illusions such as these are attentional and perceptual in nature, rather than higher level cognitive processes.

Loftus testified in Easter v. Strainer that "a witness's memory begins to decay immediately after an event takes place, and that it can be altered or supplanted by intervening events such as a suggestive line-up." (No. C-92-1635). Similarly, in United States v. Curry, Loftus advised the court that "post-event phenomena may distort or supplant original memory, and memory is easily distorted by leading questions or other manipulations." (977 F.2d 1042; 1992 U.S. App. LEXIS 23386). Based on the totality of the research reviewed in the present report, it seems as though the expert information she provided in these cases was an overstatement of the research evidence at best.

Loftus, however, is not the only eyewitness memory expert in the country who testifies before juries that eyewitness reports are tainted by postevent information. In a recent survey of psychological experts, 94% said that the effect of postevent information on eyewitness reports was reliable enough to testify about in the courtroom (Kassin, Tubb, Hosch, & Memon, 2001). Additionally, 83% said that they would testify in the courtroom about the effect under the right circumstances, and 98% said that their opinions were based on peer reviewed, scientific research. The meta-analysis reported in the present paper suggests that the effect is not as reliable as some experts might think.

Furthermore, the conditions under which the misinformation effect is studied in the laboratory may be highly different from those that real eyewitnesses encounter. For instance, Flowe and Ebbesen (2001) examined the characteristics of 721 closed felony cases and found that the modal robbery lasted for 5 minutes, and the modal rape or assault lasted for 10 minutes. In contrast, the slide sequence or video lasted only 1.5 minutes on average (mode) in the adult misinformation effect studies. Additionally, critical details, such as the perpetrator and the weapon, were viewed for 60 seconds on average (mode) in the archives, whereas the critical detail in the slide sequence was visible typically for only 5 seconds (mode) in the studies. The factors examined in the real world cases also encompassed a considerable range of values, far more vast than those implemented in the laboratory studies. These results suggest that multiple factors in combination have to be considered in order to determine whether the research generalizes to any given real world case.

Finally, in real world cases, experts will always seem to have a claim for testifying about postevent information. For instance, in the archival cases, 86% involved eyewitnesses who knew one another before the crime was committed. While it is nearly certain that witnesses will talk to one another about what they witnessed, what is not clear is how often misleading information or consistent information will be introduced in these conversations. Thus, psychologists testifying in the courtroom are faced with a dilemma: Should they discuss the "memory enhancing" effects that Loftus et al. (1978) found after participants were provided with consistent information during the interpolated event, an "information effect" of sorts, or should they testify on the impaired accuracy observed for critical items when misleading information was immediately presented and memory was tested after a 5 minute retention interval?

In a chapter written to further advance the idea that memory is impaired by misleading information, Loftus, Feldman and Dashiell (1995) opened with the following:

Though they later conceded the type of memory implants that they had in mind were "smaller" and "less exotic", the reader is left with the impression that it has been proven beyond any reasonable doubt that misleading information can supplant true memory.

The research reviewed in this paper suggests that the day when artificial memories are created is not here yet, as the misinformation effect does not appear to be a reliable memory phenomenon. The results of the meta-analysis indicated that any conclusions drawn from the research must be highly qualified. The memory reports of children compared to adults appear to be more influenced by misleading postevent information. Furthermore, on the modified test, there is no evidence that anything has happened to adult memory for the critical event, as accuracy on control and misled items did not significantly differ. Finally, in cases where a misinformation effect is found, more likely than not it is an interference effect that will diminish with time, not a storage based effect that is permanent.

References

Abeles, P., & Morton, J. (1999). Avoiding misinformation: Reinstating target modality. Quarterly Journal of Experimental Psychology: Human Experimental Psychology, 52(3), 581-592.

Ausubel, D. P., Stager, M., & Gaite, A. J. (1968). Retroactive Facilitation in Meaningful Verbal Learning. Journal of Educational Psychology, 59(4), 250-255.

Bacon, W. F., & Egeth, H. E. (1997). Goal-directed guidance of attention: Evidence from conjunctive visual search. Journal of Experimental Psychology: Human Perception & Performance, 23(4), 948-961.

Bahrick, H. P., Bahrick, P. O., & Wittlinger, R. P. (1975). Fifty Years of Memory for Names and Faces: A Cross-Sectional Approach. Journal of Experimental Psychology: General, 104(1), 54-75.

Bahrick, H. P. (1984). Fifty years of second language attrition: Implications for programmatic research. Modern Language Journal, 68(2), 105-118.

Bahrick, H. P. (1984). Semantic memory content in permastore: Fifty years of memory for Spanish learned in school. Journal of Experimental Psychology: General, 113(1), 1-29.

Bekerian, D. A., & Bowers, J. M. (1983). Eyewitness testimony: Were we misled? Journal of Experimental Psychology: Learning, Memory, & Cognition, 9(1), 139-145.*

Belli, R. F., Windschitl, P. D., McCarthy, T. T., & Winfrey, S. E. (1992). Detecting Memory Impairment With a Modified Test Procedure - Manipulating Retention Interval With Centrally Presented Event Items. Journal of Experimental Psychology-Learning Memory and Cognition, 18(2), 356-367.*

Belli, R. F. (1993). Failure of Interpolated Tests in Inducing Memory Impairment With Final Modified Tests - Evidence Unfavorable to the Blocking Hypothesis. American Journal of Psychology, 106(3), 407-427.*

Blank, H. (1998). Memory states and memory tasks: An integrative framework for eyewitness memory and suggestibility. Memory, 6(5), 481-529.

Blank, H. (1998). Memory states and memory tasks: An integrative framework for eyewitness memory and suggestibility. Memory, 6(5), 481-529.

Bonto, M. A., & Payne, D. G. (1991). Role of environmental context in eyewitness memory. American Journal of Psychology, 104(1), 117-134.

Bowers, J. M., & Bekerian, D. A. (1984). When will postevent information distort eyewitness testimony? Journal of Applied Psychology, 69(3), 466-472.*

Bowman, L. L., & Zaragoza, M. S. (1989). Similarity of encoding context does not influence resistance to memory impairment following misinformation. American Journal of Psychology, 102(2), 249-264.*

Braun, K. A., & Loftus, E. F. (1998). Advertising's misinformation effect. Applied Cognitive Psychology, 12(6), 569-591.

California Rules of Evidence. [Online]. Available: http://expertpages.com/state_rules_of_evidence.htm

Ceci, S. J., Ross, D. F., & Toglia, M. P. (1987). Suggestibility of children's memory: Psycholegal implications. Journal of Experimental Psychology: General, 116(1), 38-49.*

Ceci, S. J., & Bruck, M. (1993). Suggestibility of the child witness: A historical review and synthesis. Psychological Bulletin, 113(3), 403-439.

Ceci, S. J., Huffman, M. L. C., Smith, E., & Loftus, E. F. (1994). Repeatedly Thinking About a Non-Event - Source Misattributions Among Preschoolers. Consciousness and Cognition, 3(3-4), 388-407.

Chandler, C. C. (1989). Specific retroactive interference in modified recognition tests: Evidence for an unknown cause of interference. Journal of Experimental Psychology: Learning, Memory, & Cognition, 15(2), 256-265.*

Chandler, C. C. (1991). How memory for an event is influenced by related events: Interference in modified recognition tests. Journal of Experimental Psychology: Learning, Memory, & Cognition, 17(1), 115-125.*

Chandler, C. C., & Gargano, G. J. (1998). Retrieval processes that produce interference in modified forced-choice recognition tests. Memory & Cognition, 26(2), 220-231.*

Christianson, S.-A. (1992). Emotional stress and eyewitness memory: A critical review. Psychological Bulletin, 112(2), 284-309.

Crook, L. S., & Dean, M. C. (1999). "Lost in a shopping mall"--A breach of professional ethics. Ethics & Behavior, 9(1), 39-50.

Crowder, R. G. (1982). General forgetting theory and the locus of amnesia. In L. S. Cermak (Ed.), Human memory and amnesia . Hillsdale: Erlbaum.

Cutler, B. L., Penrod, S. D., & Martens, T. K. (1987). Improving the reliability of eyewitness identification: Putting context into context. Journal of Applied Psychology, 72(4), 629-637.*

Dempster, F. N. (1992). The rise and fall of the inhibitory mechanism: Toward a unified theory of cognitive development and aging. Developmental Review, 12(1), 45-75.

Dodson, C., & Reisberg, D. (1991). Indirect testing of eyewitness memory: The (non)effect of misinformation. Bulletin of the Psychonomic Society, 29(4), 333-336.*

Ebbinghaus, H. (1885). Uber das Gedachtnis: Untersuchungen zur experimentellen Psychologie (Reprinted as H.E. Ebbighaus (1964). Memory: A contribution to experimental psychology (H.A. Turger, Trans.). New York: Dover., Trans.). Leipzig: Dunker and Humboldt.

Federal Rules of Evidence for United States Courts and Magistrates. (2001-2002). St. Paul: West Publishing Company.

Flowe, H. D., & Ebbesen, E. B. (2001). Testing the reflection assumption: An examination of the external validity of published studies on lineup identification accuracy. Unpublished Manuscript.

Frost, P. (2000). The quality of false memory over time: Is memory for misinformation "remembered" or "known"? Psychonomic Bulletin & Review, 7(3), 531-536.

Gathercole, S. E. (1998). The development of memory. Journal of Child Psychology and Psychiatry and Allied Disciplines, 39(1), 3-27.

Gibling, F., & Davies, G. (1988). Reinstatement of context following exposure to post-event information. British Journal of Psychology, 79(1), 129-141.

Hall, J. F. (1955). Retroactive Inhibition in Meaningful Material. Journal of Educational Psychology, 46, 47-52.

Hebb, D. O. (1945). The organization of behavior: A neuropsychological theory. New York: Wiley.

Henkel, L. A., Franklin, N., & Johnson, M. K. (2000). Cross-modal source monitoring confusions between perceived and imagined events. Journal of Experimental Psychology: Learning, Memory, & Cognition, 26(2), 321-335.

Holliday, R. E., & Hayes, B. K. (2002). Automatic and intentional processes in children's recognition memory: The reversed misinformation effect. Applied Cognitive Psychology, 16(1), 1-16.*

Jackson, S., & Crockenberg, S. (1998). A comparison of suggestibility in 4-year-old girls in response to parental or stranger misinformation. Journal of Applied Developmental Psychology, 19(4), 527-542.

Jacoby, L. L. (1999). Deceiving the elderly: Effects of accessibility bias in cued-recall performance. Journal of Cognitive Neuroscience, 16, 417-436.

Jenkins, J. G., & Dallenbach, K. M. (1924). Obliviscence During Sleep and Waking. American Journal of Psychology, 35, 605-612.

Johnson, M. K., Raye, C. L., Wang, A. Y., & Taylor, T. H. (1979). Fact and fantasy: The roles of accuracy and variability in confusing imaginations with perceptual experiences. Journal of Experimental Psychology: Human Learning & Memory, 5(3), 229-240.

Johnson, M. K., Foley, M. A., Suengas, A. G., & Raye, C. L. (1988). Phenomenal characteristics of memories for perceived and imagined autobiographical events. Journal of Experimental Psychology: General, 117(4), 371-376.

Johnson, M. K., Foley, M. A., & Leach, K. (1988). The consequences for memory of imagining in another person's voice. Memory & Cognition, 16(4), 337-342.

Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitoring. Psychological Bulletin, 114(1), 3-28.

Karpel, M. E., Hoyer, W. J., & Toglia, M. P. (2001). Accuracy and qualities of real and suggested memories: Nonspecific age differences. Journals of Gerontology Series B-Psychological Sciences and Social Sciences, 56(2), P103-P110.

Kassin, S. M., Tubb, V. A., Hosch, H. M., & Memon, A. (2001). On the "general acceptance" of eyewitness testimony research. American Psychologist, 56(5), 405-416.*

Krafka, C., & Penrod, S. (1985). Reinstatement of context in a field experiment on eyewitness identification. Journal of Personality and Social Psychology, 49(1), 58-69.*

Kramer, T. H., Buckhout, R., & Eugenio, P. (1990). Weapon focus, arousal, and eyewitness memory: Attention must be paid. Law & Human Behavior, 14(2), 167-184.

Kroll, N. E., Ogawa, K. H., & Nieters, J. E. (1988). Eyewitness memory and the importance of sequential information. Bulletin of the Psychonomic Society, 26(5), 395-398.*

Lindsay, D. S., & Johnson, M. K. (1989). The eyewitness suggestibility effect and memory for source. Memory & Cognition, 17(3), 349-358.

Lindsay, D. S. (1990). Misleading suggestions can impair eyewitnesses' ability to remember event details. Journal of Experimental Psychology: Learning, Memory, & Cognition, 16(6), 1077-1083.

Lindsay, D. S., & Johnson, M. K. (1991). Recognition memory and source monitoring. Bulletin of the Psychonomic Society, 29, 203-205.

Lisanby, S. H., Maddox, J. H., Prudic, J., Devanand, D. P., & Sackeim, H. A. (2000). The effects of electroconvulsive therapy on memory of autobiographical and public events. Archives of General Psychiatry, 57(6), 581-590.

Loftus, E. F. (1977). Shifting human color memory. Memory & Cognition, 5(6), 696-699.

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 & Memory, 4(1), 19-31.*

Loftus, E. F. (1979). Eyewitness testimony. Cambridge: Harvard University Press.

Loftus, E. F., & Loftus, G. R. (1980). On the permanence of stored information in the human brain. American Psychologist, 35(5), 409-420.

Loftus, E. F., Schooler, J. W., & Wagenaar, W. A. (1985). The fate of memory: Comment on McCloskey and Zaragoza. Journal of Experimental Psychology: General, 114(3), 375-380.

Loftus, E. F., Donders, K., Hoffman, H. G., & Schooler, J. W. (1989). Creating new memories that are quickly accessed and confidently held. Memory & Cognition, 17(5), 607-616.*

Loftus, E. F., & Hoffman, H. G. (1989). Misinformation and memory: The creation of new memories. Journal of Experimental Psychology: General, 118(1), 100-104.

Loftus, E. F. (1991). The glitter of everyday memory . . .and the gold. American Psychologist, 46(1), 16-18.*

Loftus, E. F., Levidow, B., & Duensing, S. (1992). Who remembers best? Individual differences in memory for events that occurred in a science museum. Applied Cognitive Psychology, 6(2), 93-107.

Loftus, E. F., Feldman, J., & Dashiell, R. (1995). The reality of illusory memories. In D. L. Schacter (Ed.), Memory Distortion . Cambridge: Harvard University Press.

Loftus, E. F., & Pickrell, J. E. (1995). The formation of false memories. Psychiatric Annals, 25(12), 720-725.

Lynn, S. J., Rhue, J. W., Myers, B. P., & Weekes, J. R. (1994). Pseudomemory in Hypnotized and Simulating Subjects. International Journal of Clinical and Experimental Hypnosis, 42(2), 118-129.

Maass, A., & Koehnken, G. (1989). Eyewitness identification: Simulating the "weapon effect.". Law and Human Behavior, 13(4), 397-408.*

Macrae, C. N., Bodenhausen, G. V., & Calvini, G. (1999). Contexts of cryptomnesia: May the source be with you. Social Cognition, 17(3), 273-297.

Magner, E. S., Markham, R., & Barnett, C. (1996). Would reading an account of an event refresh your memory? Journal of Applied Psychology, 81(6), 769-776.

Mandler, J. M., & Johnson, N. S. (1976). Some of the Thousand Words A Picture Is Worth. Journal of Experimental Psychology: Human Learning & Memory, 2(5), 529-540.

Marsh, R. L., Landau, J. D., & Hicks, J. L. (1997). Contributions of inadequate source monitoring to unconscious plagiarism during idea generation. Journal of Experimental Psychology: Learning, Memory, & Cognition, 23(4), 886-897.

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

McGeoch, J. A. (1932). Forgetting and the law of disuse. Psychological Review, 39, 352-370.

McGeoch, J. A., & McKinney, F. (1934). Retroactive inhibition in the learning of poetry. American Journal of Psychology, 46, 19-33.

Meissner, C. A., & Brigham, J. C. (2001). A meta-analysis of the verbal overshadowing effect in face identification. Applied Cognitive Psychology, 15(6), 603-616.

Melton, A. W., & Irwin, J. M. (1940). The influence of degree of interpolated learning on retroactive inhibition and the overt transfer of specific responses. American Journal of Psychology, 53, 173-203.

Minami, H., & Dallenbach, K. M. (1946). The effect of activity upon learning and retention in the cockroach, Periplaneta americana. American Journal of Psychology, 59, 1-58.

Moldoveanu, M., & Langer, E. (2002). False memories of the future: A critique of the applications of probabilistic reasoning to the study of cognitive processes. Psychological Review, 109(2), 358-375.

Muller, G. E., & Pilzecker, A. (1900). Experimentalle Beitrage zur Lehre vom Gedachtnis. Zeitschrift fur Psychologie, 1, 1-300.

Nelson, D. L., Reed, V. S., & Walling, J. R. (1976). Pictorial Superiority Effect. Journal of Experimental Psychology: Human Learning & Memory, 2(5), 523-528.

Nelson, D. L., & Reed, V. S. (1976). On the Nature of Pictorial Encoding: A Levels-Of-processing Analysis. Journal of Experimental Psychology: Human Learning & Memory, 2(1), 49-57.

Ober, B. A., & Stillman, R. C. (1988). Memory in chronic alcoholics: Effects of inconsistent versus consistent information. Addictive Behaviors, 13(1), 11-15.*

Payne, D. G., Toglia, M. P., & Anastasi, J. S. (1994). Recognition Performance-Level and the Magnitude of the Misinformation Effect in Eyewitness Memory. Psychonomic Bulletin & Review, 1(3), 376-382.

Pezdek, K., & Roe, C. (1995). The effect of memory trace strength on suggestibility. Journal of Experimental Child Psychology, 60(1), 116-128.

Pickel, K. L. (1998). Unusualness and threat as possible causes of "weapon focus". Memory, 6(3), 277-295.*

Poole, D. A., & Lindsay, D. S. (1995). Interviewing Preschoolers - Effects of Nonsuggestive Techniques, Parental Coaching, and Leading Questions On Reports of Nonexperienced Events. Journal of Experimental Child Psychology, 60(1), 129-154.

Postman, L., & Underwood, B. J. (1973). Critical Issues in Interference Theory. Memory & Cognition, 1(1), 19-40.

Powers, P. A., Andriks, J. L., & Loftus, E. F. (1979). Eyewitness Accounts of Females and Males. Journal of Applied Psychology, 64(3), 339-347.*

Qin, J., Raye, C. L., Johnson, M. K., & Mitchell, K. J. (2001). Source ROCs are (typically) curvilinear: Comment on Yonelinas (1999). Journal of Experimental Psychology: Learning, Memory, & Cognition, 27(4), 1110-1115.

Rantzen, A., & Markham, R. (1992). The reversed eyewitness testimony design: More evidence for source monitoring. Journal of General Psychology, 119(1), 37-43.

Rundus, D. (1973). Negative Effects of Using List Items as Recall Cues. Journal of Verbal Learning & Verbal Behavior, 12(1), 43-50.

Shapiro, P. N., & Penrod, S. (1986). Meta-analysis of facial identification studies. Psychological Bulletin, 100(2), 139-156.

Shaughnessy, J. J., & Mand, J. L. (1982). How Permanent Are Memories For Real Life Events. American Journal of Psychology, 95(1), 51-65.*

Sporer, S. L. (1995). Experimentally induced person mix-ups through media exposure and ways to avoid them. In E. Graham Davies, E. Sally Lloyd-Bostock, & et al. (Eds.), Psychology, law, and criminal justice: International developments in research and practice. (pp. xix, 606): Walter De Gruyter.

Springer, A. D., & Miller, R. R. (1972). Retrieval Failure Induced by Electroconvulsive Shock: Reversal with Dissimilar Training and Recovery Agents. Science, 177(4049), 628-630.

Springer, A. D., & Miller, R. R. (1972). Retrieval Failure Induced by Electroconvulsive Shock: Reversal with Dissimilar Training and Recovery Agents. Science, 177(4049), 628-630.

Suengas, A. G., & Johnson, M. K. (1988). Qualitative effects of rehearsal on memories for perceived and imagined complex events. Journal of Experimental Psychology: General, 117(4), 377-389.

Toglia, M. P., Payne, D. G., & Anastasi, J. S. (1991). Recognition level and the misinformation effect: A meta-analysis and empirical investigation. Paper presented at the Psychonomic Society, San Francisco.

Toglia, M. P., Ross, D. F., Ceci, S. J., & Hembrooke, H. (1992). The suggestibility of children's memory: A social-psychological and cognitive interpretation. In M. L. Howe, C. J. Brainerd, & V. F. Reyna (Eds.), The development of long term retention . New York: Springer-Verlag.*

Tousignant, J. P., Hall, D., & Loftus, E. F. (1986). Discrepancy detection and vulnerability to misleading postevent information. Memory & Cognition, 14(4), 329-338.*

Treisman, A., & Schmidt, H. (1982). Illusory conjunctions in the perception of objects. Cognitive Psychology, 14(1), 107-141.

Tversky, B., & Tuchin, M. (1989). A reconciliation of the evidence on eyewitness testimony: Comments on McCloskey and Zaragoza. Journal of Experimental Psychology: General, 118(1), 86-91.

Underwood, B. J. (1957). Interference and forgetting. Psychological Review, 64, 49-60.

Ward, R. A., & Loftus, E. F. (1985). Eyewitness Performance in Different Psychological Types. Journal of General Psychology, 112(2), 191-200.*

Weinberg, H. I., Wadsworth, J., & Baron, R. S. (1983). Demand and the Impact of Leading Questions On Eyewitness Testimony. Memory & Cognition, 11(1), 101-104.*

Welch-Ross, M. K. (1999). Interviewer knowledge and preschoolers' reasoning about knowledge states moderate suggestibility. Cognitive Development, 14(3), 423-442.

Welch-Ross, M. K. (1999). Preschoolers' understanding of mind: Implications for suggestibility. Cognitive Development, 14(1), 101-131.*

Wigmore, J. H. (1940). Treatise on Evidence. (3rd ed.): Little, Brown & Co.

Windschitl, P. D. (1996). Memory for faces: Evidence of retrieval-based impairment. Journal of Experimental Psychology-Learning Memory and Cognition, 22(5), 1101-1122.

Wright, D. B., & Loftus, E. F. (1998). How misinformation alters memories. Journal of Experimental Child Psychology, 71(2), 155-164.

Wright, D. B., & Stroud, J. N. (1998). Memory quality and misinformation for peripheral and central objects. Legal & Criminological Psychology, 3(Part 2), 273-286.

Zaragoza, M. S., McCloskey, M., & Jamis, M. (1987). Misleading postevent information and recall of the original event: Further evidence against the memory impairment hypothesis. Journal of Experimental Psychology: Learning, Memory, & Cognition, 13(1), 36-44.*

Zaragoza, M. S. (1991). Preschool children's susceptibility to memory impairment. In J. Doris (Ed.), The suggestibility of children's recollections: Implications for eyewitness testimony . Washington, D. C: American Psychological Association.*

Zaragoza, M. S., Dahlgren, D., & Muench, J. (1992). The role of memory impairment in children's suggestibility. In C. J. B. M. L. Howe, & V. F. Reyna (Ed.), The development of long-term retention . New York: Springer-Verlag.*

* Indicates paper was included in meta-analysis.