Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2013 Jan 1.
Published in final edited form as: J Mem Lang. 2012 Jan;66(1):68–78. doi: 10.1016/j.jml.2011.07.002

Retrieval Failure Contributes to Gist-Based False Recognition

Scott A Guerin 1, Clifford A Robbins 1, Adrian W Gilmore 2, Daniel L Schacter 1
PMCID: PMC3223960  NIHMSID: NIHMS310599  PMID: 22125357

Abstract

People often falsely recognize items that are similar to previously encountered items. This robust memory error is referred to as gist-based false recognition. A widely held view is that this error occurs because the details fade rapidly from our memory. Contrary to this view, an initial experiment revealed that, following the same encoding conditions that produce high rates of gist-based false recognition, participants overwhelmingly chose the correct target rather than its related foil when given the option to do so. A second experiment showed that this result is due to increased access to stored details provided by reinstatement of the originally encoded photograph, rather than to increased attention to the details. Collectively, these results suggest that details needed for accurate recognition are, to a large extent, still stored in memory and that a critical factor determining whether false recognition will occur is whether these details can be accessed during retrieval.

Keywords: false memory, false recognition, gist, recognition memory

INTRODUCTION

Human memory is not a literal reproduction of the past, like a photograph or film, but rather a constructed representation of past experience that is influenced by a variety of factors related to the originally encoded event, including general knowledge, personal biases, information from other events, and inferences (Bartlett, 1932; Johnson, 1997; Johnson, Hashtroudi, & Lindsay, 1993; Loftus, 1979, 2003; Roediger, 1996; Schacter, Norman, & Koutstaal, 1998). These constructive processes presumably lead to functionally beneficial representations of the past, but they also cause memory to be prone to error (Schacter, 1999, 2001). The mistaken recognition of an item that is similar, but not identical, to a previously encountered item is a ubiquitous and robust memory error referred to as gist-based false recognition (Koutstaal & Schacter, 1997; cf., Reyna & Brainerd, 1995). For instance, people tend to mistakenly recognize a word that is a synonym of a studied word (Anisfeld & Knapp, 1968), an abstract shape that is structurally similar to studied shapes (Koutstaal, Schacter, Verfaellie, Brenner, & Jackson, 1999; Slotnick & Schacter, 2004), or a pictured object that has the same verbal label as a studied item (Koutstaal, 2006). In these cases, people fail to remember the specific details of an event but can remember more abstracted information — the “gist” — such as the superordinate category of an encountered object.

One possibility is that these errors occur because the original details have been lost from memory, either because they were not encoded originally or because the memory trace has degraded over time. Then, at retrieval, the system relies on more abstract information to reconstruct the lost details. This line of thinking has been implicit in much of the literature. For instance, in the Constructive Memory Framework put forward by Schacter et al. (1998), it was proposed that gist-based false recognition results primarily from a failure of pattern separation, a process that occurs during encoding. Thus, according to this proposal, gist-based false recognition occurs in large part because the details were not adequately encoded in the first place. Brainerd and Reyna’s (2002) Fuzzy Trace Theory proposed that verbatim details are forgotten more rapidly than gist information, a combination that contributes to gist-based false recognition (see also Reyna & Brainerd, 1995).

However, it is also known that people can store an impressive amount of information, particularly about recently encountered pictures or objects. For example, Standing (1973) demonstrated that people could recognize thousands of experimentally presented pictures. More recently, Brady et al. (2008) had participants study 2,896 pictures of objects shown for 3 s each. In a forced-choice test, participants were asked to make subtle distinctions based on memory: the foil was the same object as the studied item, but in a slightly different state (e.g., a bread box with the loaf of bread inside the box or outside the box). Participants scored 87% correct in this condition, suggesting that people can store a large amount of detailed information about recently encountered objects (see also Konkle, Brady, Alvarez, & Oliva, 2010).

These findings present a puzzle. On the one hand, a large body of data on false recognition suggests that the detail stored in episodic memory is limited and that memory relies heavily on constructive processes to compensate for this limitation. On the other hand, the findings concerning highly specific recognition of visual objects suggest that the level of detail stored in episodic memory is far greater than what would have been expected on the basis of the false recognition findings. Here we attempt to reconcile these seemingly contradictory results.

We suggest that, even when rates of false recognition are high, people do retain many details, but do not adequately utilize them at retrieval. In two experiments, we explored the hypothesis that high rates of false recognition occur when people do not attend to or do not retrieve the relevant perceptual details. To this end, we developed an experimental paradigm that encouraged participants to focus on the relevant perceptual details.

EXPERIMENT 1

Material and Methods

Rationale and Design

The conditions of the memory test are depicted in Figure 1. On each trial, the participant was presented with three pictures. Two of the pictures were related to one another because they were both exemplars of the same category and shared a common verbal label. It is important to note that the conditions did not differ systematically in terms of their perceptual presentation; they differed only in terms of the content of the participant’s memory. The participant’s task was to select one of the items as studied or reject all three items as new (“all new”). In the baseline target condition, one of the pictures was a target (studied item) and the other two items were not systematically related to any of the studied items. In the baseline foil condition, all three items were not systematically related to any of the studied items and the correct response was “all new”. In the single related item condition, one of the pictures was related to a studied item but no target was presented. The other two items were not systematically related to any of the studied items. The correct response was “all new”, but we anticipated that participants would falsely recognize the related item with high frequency, reflecting a standard gist-based false recognition effect. In the target and related item condition, the target was presented adjacent to the related item. The third item was not systematically related to any of the studied items. In this condition, the nature of the discrimination required was made explicit to the participant. Both the target and the related foil were likely to seem familiar, thus requiring that the participant systematically compared the target and related foil and identified features that distinguished them. In this condition, participants explicitly attended to the details that distinguished the target and related foil. Thus, if gist-based false recognition occurs because people fail to attend to or retrieve relevant perceptual details still stored in memory, then false recognition rates should be substantially reduced in this condition.

Figure 1.

Figure 1

Open in a new tab

Recognition test conditions in Experiment 1. The participant’s task was to select one of the three items as old (studied) or reject all three items as new. The correct response is to select the silver anchor when it is present and reject all three items when it is not. See Rationale and Design. Stimuli were presented in color.

Participants

32 college students (15 male, ages 18–29, mean 22) from the Boston metropolitan area served as participants and were paid $70 (participants were scanned with functional MRI during the experiment; the imaging data will be presented in a separate report). Candidates were excluded for participation that did not meet standard MRI safety criteria, required glasses, had strabismus or a history of eye surgery, or that were left handed. All participants provided informed consent as approved by the Institutional Review Board at Harvard University. Four participants were replaced: one because the participant failed to respond on approximately 1/3 of the trials; two for eye tracking problems; one because they previously participated in a closely related experiment.

Stimulus Materials

384 pairs of object photographs or detailed, colored drawings served as stimuli (Koutstaal, 2006; Koutstaal, et al., 2001). The items within a pair were related to each other because they were both exemplars of the same category and shared a common verbal label (e.g., wrench, dog, tree). However, the two pictures were perceptually distinct exemplars of the category and, at a minimum, differed in terms of color or orientation. On each trial of the recognition test, two pairs were used to generate the array. One of the pairs was used to produce the two related items and one of the pairs was used to produce the third item (in which case only one of the exemplars in the pair was presented). Thus, with four conditions, there were eight possible assignments of any particular pair. The object pairs were counterbalanced across all eight possible assignments. Then, exemplars serving as targets were further counterbalanced, for a total of 16 iterations of the counterbalancing. This process was repeated to produce a sample size of 32. Therefore, each stimulus occurred in each condition equally often across participants.

Eye Tracking

Eye tracking data were collected using an EyeLink 1000 MRI compatible eye tracking system (SR Research Ltd, Mississauga, Ontario, Canada). The pupil and corneal reflection were identified using automated thresholding. The pupil was modeled as an ellipse. For calibration, 9 fixation targets in an equally spaced grid covering the entire stimulus display were presented in a random order. The accuracy of the calibration was validated by presenting an additional 8 fixation targets at new locations on the display as well as the center of the display. The average measured error across the validation points was typically less than .5° for each participant. Validation was conducted before each block; calibration was rerun as necessary. Monocular data were collected at 1000 Hz. Saccades and fixations were detected online by the EyeLink software using the default thresholds (Stampe, 1993).

Procedure

Participants were told that their memory would be tested later and were presented with 144 objects (500 ms duration, 1500 ms ISI) and indicated whether the object could fit into a 13-inch box in the real world by a button press. A box measuring approximately 13 inches was presented. The participant was given a self-paced break halfway through the study session. Then the participant was placed in an MRI scanner. The occurrence of similar foils was clearly explained to the participant. The test was divided into four blocks; each began with 15 s of fixation and ended with 10 s of fixation and contained 12 trials of each condition. Each trial lasted 5 s. With the constraint that the two related items were next to each other, there were four possible arrangements of the pictures; each occurred equally often within each block. 24 fixation trials, also lasting 5 s, were randomly intermixed within each block.

Results

Accuracy

The accuracy data are shown in Figure 2. Within the baseline target condition, the hit rate was reasonably high (.73, SEM = .03, green bar). Within the baseline foil condition, the correct rejection rate was reasonably high (.69, SEM = .03, blue bar). Therefore, participants performed well.

Figure 2.

Figure 2

Open in a new tab

Accuracy data from Experiment 1. The color of each bar indicates the participant’s choice. In the examples provided in Figure 1, the scissors, bulldozers, and beavers are examples of paired unrelated foils. The apple and diskette are examples of single unrelated foils. Error bars show SEM.

In the single related item condition, participants incorrectly selected the related item at a high rate (.41, SEM = .02, yellow bar). This false alarm rate was approximately four times larger than the baseline false alarm rate to items not presented in a pair, such as the apple in Figure 1 [.10, SEM = .01, orange bar, t(31) = 17.10, p < .001]. As we expected, there was a robust gist-based false recognition effect in the single related item condition.

In the target and related item condition, the false alarm rate to related foils dropped substantially (.13, SEM = .01, yellow bar), corresponding to a 67% reduction in the rate of gist-based false recognition [t(31) = 12.24, p < .001]. Nonetheless, within the target and related item condition, false alarm rates to related foils were higher than false alarm rates to unrelated foils [.05, SEM = .01, orange bar, t(31) = 8.06, p < .001]. Thus, gist-based false recognition was dramatically reduced but not entirely eliminated.

Reaction Time

Reaction time data are presented in Table 1. Reaction times were obtained by calculating the median reaction time for each participant and then averaging across participants. Two findings are noteworthy. First, gist-based false recognition in the single related item condition was associated with longer reaction times (2029 ms, SEM = 53) than veridical recognition in the baseline target condition [1849 ms, SEM = 35, t(31) = 4.22, p < .001]. Second, reaction times associated with hits were longer in the target and related item condition (2210 ms, SEM = 75) than in the baseline target condition [1849 ms, SEM = 35, t(31) = 6.19, p < .001].

Table 1.

Reaction Times (ms) in Experiment 1

Baseline Target Baseline Foil Single Related Item Target & Related Item
Target 1849 (35) NA NA 2210 (75)
Related Foil NA NA 2029 (53) Low N
New Low N 2469 (78) 2605 (90) Low N
Open in a new tab

Note. We exclude incorrect responses (with the exception of false alarms to related foils in the single related item condition) because these occurred infrequently and are associated with high estimation error and missing data values for certain participants. SEM in parentheses.

Eye Tracking

To verify that participants systematically compared the target and related foil in the target and related foil condition, we calculated the number of saccades between similar pictures per trial. These data are presented in Figure 3. There were more saccades between related pictures in the target and related item condition (2.69, SEM = 0.13, green bar) than the baseline target condition [1.47, SEM = 0.09, green bar, t(31) = 17.80, p < .001].

Figure 3.

Figure 3

Open in a new tab

Eye tracking data from Experiment 1. The color of each bar indicates the participant’s choice. We exclude incorrect responses (with the exception of false alarms to related foils in the single related item condition) because these occurred infrequently and are associated with high estimation error and missing data values for certain participants. Error bars show SEM.

Discussion

Under the same encoding conditions that produced high rates of gist-based false recognition in the single related item condition, participants overwhelmingly chose the correct target in favor of its related foil when given the option to do so. This finding indicates that the details distinguishing the target from its related foil were, to a large extent, still retained in memory. Hence, gist-based false recognition is not produced by a loss of detail per se. Rather, gist-based false recognition is attributable, at least in part, to a failure to retrieve stored details.

One explanation for our data is that the retrieval failure contributing to gist-based false recognition resulted primarily from a suboptimal deployment of attention: in the single related foil condition, the participant failed to attend to the relevant details; in the target and related foil condition, the participant’s attention was drawn to the features that are relevant to the discrimination. However, another critical property distinguishes these conditions: in the target and related item condition, the originally encoded picture is presented at retrieval. In line with the encoding specificity principle (Tulving & Thomson, 1973), we would expect that this reinstatement of the originally encoded information led to increased access to the memory trace. There are thus two distinct mechanisms that could have contributed to the reduction in false recognition observed in Experiment 1: (i) attention to detail and (ii) reinstatement of detail. In Experiment 2, we attempted to tease apart these two mechanisms.

EXPERIMENT 2

Method

In order to distinguish between the effects of attention and reinstatement, we repeated Experiment 1 with the inclusion of a critical new condition. In the two related items condition (Figure 4), the participant was presented with two items, both of which were related to the same study item. The third item in the array was not systematically related to any of the studied items. Both of the related items should seem familiar and the participant should therefore systematically compare the two related items. Thus, in this condition the participant’s attention is drawn to the features that, on average, distinguish exemplars of the category. Critically, however, the originally encoded photograph is not presented.

Figure 4.

Figure 4

Open in a new tab

Recognition test conditions in Experiment 2. The task is the same as Experiment 1 (cf. Figure 1), except for the inclusion of the two related items condition. Stimuli were presented in color.

The methods were identical to those of Experiment 1 except as noted. 30 participants (13 male; ages 18–27, mean 21) were run. Data from two participants were replaced: one due to a technical problem during data acquisition and one because the participant’s responses on the Edinburgh Handedness Inventory indicated that the participant was left handed. A new set of 400 object triplets was generated. The counterbalancing method used in Experiment 1 was extended to five conditions and three exemplars. With five conditions there were 10 possible assignments of each triplet. Counterbalancing across three exemplars led to 30 iterations of the counterbalancing. Participants studied 160 objects. The test was divided into five blocks, each contained eight trials of each condition and 40 fixation trials.

A parallel behavioral experiment was run outside of the MRI environment to collect confidence ratings. The procedure was identical to the main experiment except that a confidence scale appeared following each recognition response. Participants responded without a deadline on a four-point scale, with 1 labeled “very uncertain”, 2 labeled “somewhat uncertain”, 3 labeled “somewhat certain”, and 4 labeled “very certain”. Eye tracking data were not collected. 30 participants (9 male; ages 18–27, mean 20) were run. Data from one participant was replaced due to poor performance (baseline false alarms greater than baseline hits).

Results

The results of Experiment 2 closely replicated Experiment 1. Below, we focus on the novel findings of Experiment 2.

Accuracy

The accuracy data are presented in Figure 5. The critical question concerns the gist-based false recognition rate in the two related items condition, where participants attended to the relevant details but did not benefit from reinstatement of the originally encoded photograph. There was no evidence of a reduction in false recognition to related foils in this condition. In fact, the false alarm rate to related foils was larger in the two related items condition (.46, SEM = .02, yellow bar) than in the single related item condition [.36, SEM = .02, yellow bar, t(29) = 4.29, p < .001]. This result could be partially attributable to differences in baseline false alarm rates to foils occurring in pairs and foils not occurring in pairs. Even within the baseline foil condition, the false alarm rate to foils occurring in pairs was larger (.12, SEM = .02, blue bar) than to foils not occurring in pairs [.08, SEM = .01, orange bar, t(29) = 2.44, p < .05]. Nonetheless, when the false alarm rates to related items were corrected by subtracting the appropriate baseline false alarm rates, there was a significant difference in false alarm rates to related items in the two related items condition (.33, SEM = .02) and the single related item condition [.27, SEM = .02, t(29) = 2.30, p < .05], indicating that differences in baseline false alarm rates do not fully account for this finding.

Figure 5.

Figure 5

Open in a new tab

Accuracy data from Experiment 2. The color of each bar indicates the participant’s choice. In the examples provided in Figure 4, the piggybanks, cats, and pretzels are examples of paired unrelated foils. The basketball, accordion, and cow are examples of single unrelated foils. Error bars show SEM.

Reaction Time

Reaction time data are reported in Table 2. Reaction times associated with false alarms to related items in the two-related items condition were longer (2642 ms, SEM = 99) than in the single related item condition [2293 ms, SEM = 74, t(29) = 5.33, p < .001].

Table 2.

Reaction Times (ms) in Experiment 2

Baseline Target Baseline Foil Single Related Item Two Related Items Target & Related Item
Target 1936 (65) NA NA NA 2127 (103)
Related Foil NA NA 2293 (74) 2642 (99) Low N
New Low N 2512 (101) 2724 (114) 2718 (100) Low N
Open in a new tab

Note. We exclude incorrect responses (with the exception of false alarms to related foils in the single related item condition and the two related items condition) because these occurred infrequently and are associated with high estimation error and missing data values for certain participants. SEM in parentheses.

Eye Tracking

The average number of saccades between related pictures is shown in Figure 6. This measure was larger in the two related items condition (2.96, SEM = 0.10, yellow bar) than in the single related item condition [1.46, SEM = 0.09, yellow bar, t(29) = 15.10, p < .001]. These data confirm that participants were directly comparing the two related items in the two related items condition.

Figure 6.

Figure 6

Open in a new tab

Eye tracking data from Experiment 2. The color of each bar indicates the participant’s choice. We exclude incorrect responses (with the exception of false alarms to related foils in the single related item condition and the two related items condition) because these occurred infrequently and are associated with high estimation error and missing data values for certain participants. Error bars show SEM.

Confidence

The accuracy data (Table 3) and reaction time data (Table 4) from the confidence experiment closely replicated those of the main experiment. The confidence data are reported in Figure 7. Confidence in gist-based false alarms in the single related item condition was less (2.76, SEM = 0.07) than in veridical memories during the baseline target condition [3.45, SEM = 0.06, t(29) = 9.47, p < .001]. The effects of the attentional manipulation are more relevant. There was a small but reliable decrease in confidence associated with hits in the target and related item condition (3.28, SEM = 0.06, green bar) relative to the baseline target condition [3.45, SEM = 0.06, green bar, t(29) = 4.03, p < .001]. Similarly, there was a small but reliable decrease in confidence associated with false alarms to related foils in the two related items condition (2.47, SEM = 0.08, yellow bar) relative to the single related item condition [2.76, SEM = 0.07, yellow bar, t(29) = 4.80, p < .001]. Although rates of gist-based false recognition were not reduced in the two related items condition, this condition was associated with a modest reduction in confidence.

Table 3.

Accuracy in the Confidence Experiment.

Baseline Target Baseline Foil Single Related Item Two Related Items Target & Related Item
Target .73 (.03) NA NA NA .66 (.03)
Related Foil NA NA .31 (.02) .37 (.02) .08 (.01)
Paired Unrelated Foil .06 (.01) .13 (.02) .11 (.02) NA NA
Single Unrelated Foil NA .08 (.01) NA .05 (.01) .03 (.01)
New .19 (.02) .76 (.04) .55 (.04) .55 (.03) .21 (.02)
Open in a new tab
Table 4.

Reaction times (ms) in the Confidence Experiment.

Baseline Target Baseline Foil Single Related Item Two Related Items Target & Related Item
Target 2169 (50) NA NA NA 2408 (74)
Related Foil NA NA 2508 (74) 3003 (86) Low N
New Low N 2619 (82) 2769 (83) 2881 (78) Low N
Open in a new tab
Figure 7.

Figure 7

Open in a new tab

Confidence data from Experiment 2, collected in a separate group of 30 participants run outside the MRI environment. The color of each bar indicates the participant’s choice. We exclude incorrect responses (with the exception of false alarms to related foils in the single related item condition and the two related items condition) because these occurred infrequently and are associated with high estimation error and missing data values for certain participants. Error bars show SEM.

Discussion

Experiment 2 differentiated between two mechanisms that could have contributed to the reduction in rates of gist-based false recognition observed in Experiment 1: (i) attention to detail and (ii) reinstatement of detail. In the two related items condition, participants actively attended to the details that, on average, distinguished between exemplars of the category. Nonetheless, there was no evidence of a reduction in gist-based false recognition rates. In contrast, in the target and related item condition — where participants had the added benefit of having the originally encoded photograph presented again at retrieval — there was a dramatic reduction in rates of gist-based false recognition, replicating Experiment 1. These results suggest that the outcome of Experiment 1 depended primarily on reinstatement of the studied picture and that increased attention to detail is not sufficient to reduce rates of gist-based false recognition.

One caveat is that the target and related item condition and the two related items condition may have differed in how effectively they drew the participant’s attention to specific diagnostic features. For instance, if the target is a red car and both related exemplars are blue cars, then the exemplars will not differ in terms of color in the two related items condition and the participant’s attention will not be directed towards color. Nonetheless, the two related items condition does draw the participant’s attention to the features that, on average, distinguish exemplars of a category. Therefore, if attention to detail were sufficient to reduce gist-based false recognition, then we would expect some reduction in gist-based false recognition, but we saw no evidence of such a reduction.

It is critical to note that the aforementioned caveat has no bearing on the broader and more significant conclusion that gist-based false recognition results from retrieval failure.

GENERAL DISCUSSION

We investigated the hypothesis that gist-based false recognition does not result from a loss of detail from memory per se, but rather from inadequate utilization of stored detail during retrieval. Experiment 1 investigated the possibility that people do not attend to the relevant details. In the critical condition of Experiment 1, there was a dramatic reduction in gist-based false recognition. Under the same encoding conditions that produced large rates of gist-based false recognition in the single related item condition, participants overwhelmingly chose the correct target rather than the related foil when given the option to do so. This finding indicates that the detail distinguishing targets and related foils was not lost and could be retrieved under appropriate conditions. Experiment 2 investigated the role of two distinct mechanisms that could have contributed to the initial results: (i) attention to detail and (ii) reinstatement of detail. The results of the second experiment suggest that the outcome of the first experiment was driven by reinstatement, not attention. These results suggest that retrieval failure is a major factor contributing to gist-based false recognition. The details are, to a large extent, still stored in memory. A critical factor determining whether false recognition will occur is whether the details can be accessed during retrieval.

Theoretical approaches to gist-based false recognition have implicitly assumed that gist-based false recognition occurs primarily because the original details have been lost from memory, either because they were not encoded initially or because the memory trace degrades over time. For instance, the Constructive Memory Framework proposed that failures of pattern separation, a process that occurs during encoding, are a major cause of gist-based false recognition (Schacter, et al., 1998). Fuzzy Trace Theory proposed that verbatim details are lost more rapidly than gist information, thus leading to gist-based false recognition (Brainerd & Reyna, 2002; Reyna & Brainerd, 1995). The current findings suggest that failures of encoding and storage play a smaller role in generating gist-based false recognition than acknowledged by existing theories, and that failures that occur during retrieval play a larger role than previously acknowledged.

Our findings are consistent with laboratory experiments on eyewitness identification of suspects in lineups. When the culprit is not present in the line up, rates of mistaken identification are typically high. However, when the culprit is present, participants often choose the culprit correctly (Wells & Olson, 2003). Although research on mistaken eyewitness identification has generally proceeded independently of research on gist-based false recognition, the present results raise the possibility that there are strong parallels between these two types of memory errors and that retrieval failure, in particular, is a common mechanism contributing to both phenomena.

The experiments reported here present interesting parallels to studies investigating the effects of misleading post-event information on memory. Building on the classic work of Elizabeth Loftus and her colleagues (Loftus, 1979, 2003; Loftus, Miller, & Burns, 1978), McCloskey and Zaragoza (1985) presented participants with slides depicting a man holding a hammer and subsequently presented them with a misleading narrative describing the man as holding a screwdriver. On a subsequent forced choice test, participants who had to choose whether the slides contained a hammer or a screwdriver performed poorly when they had been presented with misleading information, replicating the standard findings of Loftus and her colleagues. In a novel version of the test, participants had to choose whether the slides contained a hammer or a wrench; these participants were not impaired by the misleading information. McCloskey and Zaragoza (1985) concluded that the memory for the original event was not impaired by the misleading information (see also Zaragoza, Jamis, & McCloskey, 1987). Rather, screwdriver is an obvious choice for any participant that remembers that screwdriver was in the narrative but that cannot — for reasons unrelated to the misleading information — remember that a hammer was presented in the original slides. Thus, the misinformation effect depends heavily on how memory is probed. Similarly, we have found that the occurrence of gist-based false recognition depends heavily on the way that memory is tested. Although targeting very different mechanisms of memory distortion, both lines of research converge on the critical importance of retrieval processes in the manifestation of memory distortions.

These results offer an interesting contrast to findings obtained using the well-known Deese-Roediger McDerrmott (DRM) paradigm (Deese, 1959; Roediger & Mcdermott, 1995). In this paradigm, participants study a series of associatively related words that are all forward associates of a critical word that was not studied. The standard finding is that people falsely recognize the critical non-studied word very frequently. Schacter, Israel, & Racine (1999) modified the paradigm by presenting a picture with each studied word that depicted what the word named (see also Israel & Schacter, 1997). Participants that studied the pictures showed substantially lowered false recognition rates, relative to participants that didn’t study the pictures. Schacter et al. (1999) concluded that participants in the picture condition expected to recollect distinctive pictorial content and demanded access to that content in order to classify a word as “old”, thereby allowing them to reject related lures words that did not elicit the expected distinctive information; Schacter et al. (1999) termed this process the distinctiveness heuristic (see also Dodson & Schacter, 2002; Gallo, Weiss, & Schacter, 2004). In contrast to the present findings, work on the distinctiveness heuristic suggests that under some circumstances, people can strategically attend to perceptual information to reduce gist-based false recognition. A critical difference is that, when using the distinctiveness heuristic, retrieval of any pictorial information is sufficient to classify an item as “old”. The distinctiveness heuristic requires that people retrieve some sort of distinctive content, but it does not require that the content is highly detailed. In the current experiment, participants must retrieve enough detail to distinguish a target and a related foil. The difficulty of reconstructing this specific visual detail in the absence of the target object contributes to the robustness of gist-based false recognition.

Given these considerations, it will be important to determine whether our results depend on the use of visually rich and detailed stimuli such as colored pictures or whether they also generalize to gist-based false recognition of words and other stimuli. In the special circumstance in which words are presented in visually distinctive fonts (Arndt & Reder, 2003), it seems likely that similar results would be obtained, because participants’ attention is drawn to the visual characteristics that distinguish one word from another. False recognition of synonyms presented in a constant font would be a more marked departure from the current paradigm. It is possible that reinstatement of the originally studied word would enable access to orthographic, phonological, or semantic details stored in episodic memory that distinguish between the target and its synonym. Alternatively, it is possible that loss of detail plays a larger role in gist-based false recognition of words than it does in gist-based false recognition of pictures. Future studies should examine this important and unresolved issue.

More generally, our findings are consistent with the widely acknowledged fact that false memories are affected by the circumstances at retrieval. With encoding and retention held constant, false memories can be reduced by factors at retrieval, particularly manipulations that encourage participants to monitor contextual details, as has been emphasized in discussion of the source monitoring framework (Dodson & Schacter, 2002; Johnson, 1997; Johnson, et al., 1993; Mather, Henkel, & Johnson, 1997). The current results suggest that, in addition to these previously demonstrated retrieval factors, the failure to retrieve detailed information that is still available in memory is a major contributor to gist-based false recognition.

Our conclusions also have much in common with Fuzzy Trace Theory (Brainerd & Reyna, 2002; Reyna & Brainerd, 1995). According to that model, people encode two independent types of traces: gist traces are semantic abstractions and tend to produce false recognition (as well as true recognition); verbatim traces contain specific details and support true recognition. It is proposed that verbatim traces are forgotten more rapidly than gist traces, thereby producing such phenomena as gist-based false recognition. However, the model also allows for cue-based retrieval effects. It is postulated that presentation of a target tends to elicit retrieval of a verbatim trace whereas presentation of a related foil tends to elicit retrieval of a gist trace. This notion of differential access to detail depending on the cue information presented to the participant is consistent with our interpretation of the present results. However, rather than focusing on the distinct properties of verbatim and gist traces, we emphasize the role of retrieval cues in providing access to information that is available in memory (Tulving & Pearlstone, 1966; Tulving & Thompson, 1973) and that can be used under appropriate conditions to avoid gist-based false recognition.

Research on recognition memory has been heavily influenced by the proposal that there are two distinct processes contributing to recognition: recollection is retrieval of specific contextual details, such as where you were or what you were thinking when an item was encountered; familiarity is a general impression that an item has been encountered recently in the absence of specific contextual details (Jacoby, 1991; Mandler, 1980; Tulving, 1985; Yonelinas, 2002). The current results clearly indicate that, in the target and related item condition, participants retrieved some type of information that distinguished the target from the related distracter. However, this retrieval could take the form of familiarity, recollection, or some combination of the two. For instance, it is possible that the degree of perceptual match between a test item and a studied item affected familiarity (Jacoby & Dallas, 1981). If so, a signal detection model could be used to explain the present findings (see Shiffrin & Steyvers, 1997 for a discussion of signal detection models with multiple test items). For instance, participants could perform a recognition decision on each item individually and, if more than one item surpassed the criterion, choose the item with the highest familiarity. This theoretical approach is consistent with our conclusion that the details distinguishing the target and related distracter are often still available in memory. It is also consistent with our conclusion that the ability to utilize the stored detail to avoid gist-based false recognition is heavily dependent on the nature of the retrieval cues provided during the memory test and, in particular, on reinstatement of the originally studied item. Future experiments that examine the role of retrieval failure in generating gist-based false recognition should attempt to disentangle the contributions of recollection and familiarity.

Although these results clearly suggest that retrieval failure is a major factor in gist-based false recognition, it is probably not the only factor. Gist-based false recognition was dramatically reduced in the target and related items condition, but it was not completely eliminated. Even these residual gist-based false memories could be attributable, in part, to retrieval failure: it is in principle possible that the provision of still more cue information, such as reinstatement of the precise study context (Smith & Vela, 2001) or the temporal ordering of studied items (Jacoby, 1972), could reduce the rate of gist-based false alarms further. Nonetheless, it is plausible that the impressive degree of detail stored in memory is not without bound. Our contention is not that loss of detail from memory plays no role whatsoever, but rather that it plays a smaller role than has been previously assumed. In contrast, retrieval failure plays a larger role in generating gist-based false recognition than has been acknowledged to date.

Our results raise the prospect that memory distortions may be avoided by improving the accessibility of information that is still stored in memory. Thus, a pressing avenue for future research is to determine the extent to which similar retrieval failure mechanisms apply to other laboratory demonstrations of false memory, such as imagination inflation (Garry, Manning, Loftus, & Sherman, 1996; Goff & Roediger, 1998), DRM false recognition (Gallo, 2006; Miller & Wolford, 1999; Roediger & Mcdermott, 1995; Weinstein, McDermott, & Chan, 2010), and memory conjunction errors (Reinitz, Lammers, & Cochran, 1992). Further investigation of the role of retrieval failure in producing false memories is likely to help us understand more fully how a system capable of impressive levels of retention, and which in daily life often serves us well, is also capable of error and distortion (Brainerd & Reyna, 2002; Reyna & Brainerd, 1995; Schacter, 1999).

Highlight.

  • People often falsely recognize items that are similar to previously encountered items.

  • A widely held view is that this error occurs because the details fade rapidly from our memory.

  • Participants overwhelmingly chose the correct target rather than its related foil when given the option to do so.

  • The details needed for accurate recognition are, to a large extent, still stored in memory.

  • A critical factor determining whether false recognition will occur is whether these details can be accessed during retrieval.

Acknowledgments

Thanks to Ross Mair, Tammy Moran, Caroline West Miguel Cutiongco, David Dornblaser, Rebecca Hersher, and Allison Hyland for assistance, Marcus Johnson for assistance with installation of the eye tracker and programming and Elissa Aminoff and Wilma Koutstaal for providing stimuli Supported by NRSA AG034699to SAG, and National Institute on Aging AG08441 and National Institute of Health MH060941to DLS

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Anisfeld M, Knapp M. Association, synonymity, and directionality in false recognition. J Exp Psychol. 1968;77(2):171–179. doi: 10.1037/h0025782. [DOI] [PubMed] [Google Scholar]
  2. Arndt J, Reder LM. The effect of distinctive visual information on false recognition. Journal of Memory and Language. 2003;48(1):1–15. [Google Scholar]
  3. Bartlett FC. Remembering; a study in experimental and social psychology. Cambridge [Eng.]: The University press; 1932. [Google Scholar]
  4. Brady TF, Konkle T, Alvarez GA, Oliva A. Visual long-term memory has a massive storage capacity for object details. Proceedings of the National Academy of Sciences of the United States of America. 2008;105(38):14325–14329. doi: 10.1073/pnas.0803390105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Brainerd CJ, Reyna VF. Fuzzy-trace theory and false memory. Current Directions in Psychological Science. 2002;11(5):164–169. [Google Scholar]
  6. Deese J. On the prediction of occurrence of particular verbal intrusions in immediate recall. J Exp Psychol. 1959;58(1):17–22. doi: 10.1037/h0046671. [DOI] [PubMed] [Google Scholar]
  7. Dodson CS, Schacter DL. When false recognition meets metacognition: The distinctiveness heuristic. Journal of Memory and Language. 2002;46(4):782–803. [Google Scholar]
  8. Gallo DA. Associative illusions of memory: false memory research in DRM and related tasks. New York: Psychology Press; 2006. [Google Scholar]
  9. Gallo DA, Weiss JA, Schacter DL. Reducing false recognition with criterial recollection tests: Distinctiveness heuristic versus criterion shifts. Journal of Memory and Language. 2004;51(3):473–493. [Google Scholar]
  10. Garry M, Manning CG, Loftus EF, Sherman SJ. Imagination inflation: Imagining a childhood event inflates confidence that it occurred. Psychonomic Bulletin & Review. 1996;3(2):208–214. doi: 10.3758/BF03212420. [DOI] [PubMed] [Google Scholar]
  11. Goff LM, Roediger HL. Imagination inflation for action events: Repeated imaginings lead to illusory recollections. Memory & Cognition. 1998;26(1):20–33. doi: 10.3758/bf03211367. [DOI] [PubMed] [Google Scholar]
  12. Israel L, Schacter DL. Pictorial encoding reduces false recognition of semantic associates. Psychonomic Bulletin & Review. 1997;4(4):577–581. [Google Scholar]
  13. Jacoby LL. Effects of Organization on Recognition Memory. Journal of Experimental Psychology. 1972;92(3):325. [Google Scholar]
  14. Jacoby LL. A process dissociation framework: Separating automatic from intentional uses of memory. Journal of Memory and Language. 1991;30(5):513–541. [Google Scholar]
  15. Jacoby LL, Dallas M. On the relationship between autobiographical memory and perceptual learning. J Exp Psychol Gen. 1981;110(3):306–340. doi: 10.1037//0096-3445.110.3.306. [DOI] [PubMed] [Google Scholar]
  16. Johnson MK. Source monitoring and memory distortion. Philosophical Transactions of the Royal Society B-Biological Sciences. 1997;352(1362):1733–1745. doi: 10.1098/rstb.1997.0156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Johnson MK, Hashtroudi S, Lindsay DS. Source Monitoring. Psychological Bulletin. 1993;114(1):3–28. doi: 10.1037/0033-2909.114.1.3. [DOI] [PubMed] [Google Scholar]
  18. Konkle T, Brady TF, Alvarez GA, Oliva A. Conceptual Distinctiveness Supports Detailed Visual Long-Term Memory for Real-World Objects. Journal of Experimental Psychology-General. 2010;139(3):558–578. doi: 10.1037/a0019165. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Koutstaal W. Flexible remembering. Psychon Bull Rev. 2006;13(1):84–91. doi: 10.3758/bf03193817. [DOI] [PubMed] [Google Scholar]
  20. Koutstaal W, Schacter DL. Gist-based false recognition of pictures in older and younger adults. Journal of Memory and Language. 1997;37(4):555–583. [Google Scholar]
  21. Koutstaal W, Schacter DL, Verfaellie M, Brenner C, Jackson EM. Perceptually based false recognition of novel objects in amnesia: Effects of category size and similarity to category prototypes. Cognitive Neuropsychology. 1999;16(3–5):317–341. [Google Scholar]
  22. Koutstaal W, Wagner AD, Rotte M, Maril A, Buckner RL, Schacter DL. Perceptual specificity in visual object priming: functional magnetic resonance imaging evidence for a laterality difference in fusiform cortex. Neuropsychologia. 2001;39(2):184–199. doi: 10.1016/s0028-3932(00)00087-7. [DOI] [PubMed] [Google Scholar]
  23. Loftus EF. Malleability of Human-Memory. American Scientist. 1979;67(3):312–320. [PubMed] [Google Scholar]
  24. Loftus EF. Make-believe memories. American Psychologist. 2003;58(11):867–873. doi: 10.1037/0003-066X.58.11.867. [DOI] [PubMed] [Google Scholar]
  25. Loftus EF, Miller DG, Burns HJ. Semantic Integration of Verbal Information into a Visual Memory. Journal of Experimental Psychology-Human Learning and Memory. 1978;4(1):19–31. [PubMed] [Google Scholar]
  26. Mandler G. Recognizing - the Judgment of Previous Occurrence. Psychological Review. 1980;87(3):252–271. [Google Scholar]
  27. Mather M, Henkel LA, Johnson MK. Evaluating characteristics of false memories: remember/know judgments and memory characteristics questionnaire compared. Mem Cognit. 1997;25(6):826–837. doi: 10.3758/bf03211327. [DOI] [PubMed] [Google Scholar]
  28. McCloskey M, Zaragoza M. Misleading Postevent Information and Memory for Events - Arguments and Evidence against Memory Impairment Hypotheses. Journal of Experimental Psychology-General. 1985;114(1):1–16. doi: 10.1037//0096-3445.114.1.1. [DOI] [PubMed] [Google Scholar]
  29. Miller MB, Wolford GL. Theoretical commentary: The role of criterion shift in false memory. Psychological Review. 1999;106(2):398–405. [Google Scholar]
  30. Reinitz MT, Lammers WJ, Cochran BP. Memory-conjunction errors: miscombination of stored stimulus features can produce illusions of memory. Mem Cognit. 1992;20(1):1–11. doi: 10.3758/bf03208247. [DOI] [PubMed] [Google Scholar]
  31. Reyna VF, Brainerd CJ. Fuzzy-Trace Theory - an Interim Synthesis. Learning and Individual Differences. 1995;7(1):1–75. [Google Scholar]
  32. Roediger HL. Memory illusions. Journal of Memory and Language. 1996;35(2):76–100. [Google Scholar]
  33. Roediger HL, Mcdermott KB. Creating False Memories - Remembering Words Not Presented in Lists. Journal of Experimental Psychology-Learning Memory and Cognition. 1995;21(4):803–814. [Google Scholar]
  34. Schacter DL. The seven sins of memory - Insights from psychology and cognitive neuroscience. American Psychologist. 1999;54(3):182–203. doi: 10.1037//0003-066x.54.3.182. [DOI] [PubMed] [Google Scholar]
  35. Schacter DL. The seven sins of memory: how the mind forgets and remembers. Boston: Houghton Mifflin; 2001. [Google Scholar]
  36. Schacter DL, Israel L, Racine C. Suppressing false recognition in younger and older adults: The distinctiveness heuristic. Journal of Memory and Language. 1999;40(1):1–24. [Google Scholar]
  37. Schacter DL, Norman KA, Koutstaal W. The cognitive neuroscience of constructive memory. Annu Rev Psychol. 1998;49:289–318. doi: 10.1146/annurev.psych.49.1.289. [DOI] [PubMed] [Google Scholar]
  38. Shiffrin RM, Steyvers M. Model for recognition memory: REM - Retrieving effectively from memory. Psychonomic Bulletin & Review. 1997;4(2):145–166. doi: 10.3758/BF03209391. [DOI] [PubMed] [Google Scholar]
  39. Slotnick SD, Schacter DL. A sensory signature that distinguishes true from false memories. Nat Neurosci. 2004;7(6):664–672. doi: 10.1038/nn1252. [DOI] [PubMed] [Google Scholar]
  40. Smith SM, Vela E. Environmental context-dependent memory: A review and meta-analysis. Psychonomic Bulletin & Review. 2001;8(2):203–220. doi: 10.3758/bf03196157. [DOI] [PubMed] [Google Scholar]
  41. Stampe DM. Heuristic Filtering and Reliable Calibration Methods for Video-Based Pupil-Tracking Systems. Behavior Research Methods Instruments & Computers. 1993;25(2):137–142. [Google Scholar]
  42. Standing L. Learning 10,000 Pictures. Quarterly Journal of Experimental Psychology. 1973 May;25:207–222. doi: 10.1080/14640747308400340. [DOI] [PubMed] [Google Scholar]
  43. Tulving E. Memory and consciousness. Canadian Psychology. 1985;26(1):1–12. [Google Scholar]
  44. Tulving E, Thomson DM. Encoding Specificity and Retrieval Processes in Episodic Memory. Psychological Review. 1973;80(5):352–373. [Google Scholar]
  45. Weinstein Y, McDermott KB, Chan JCK. True and false memories in the DRM paradigm on a forced choice test. Memory. 2010;18(4):375–384. doi: 10.1080/09658211003685533. [DOI] [PubMed] [Google Scholar]
  46. Wells GL, Olson EA. Eyewitness testimony. Annual Review of Psychology. 2003;54:277–295. doi: 10.1146/annurev.psych.54.101601.145028. [DOI] [PubMed] [Google Scholar]
  47. Yonelinas AP. The nature of recollection and familiarity: A review of 30 years of research. Journal of Memory and Language. 2002;46(3):441–517. [Google Scholar]
  48. Zaragoza MS, Jamis M, McCloskey M. Misleading Postevent Information and Recall of the Original Event - Further Evidence against the Memory Impairment Hypothesis. Journal of Experimental Psychology-Learning Memory and Cognition. 1987;13(1):36–44. [Google Scholar]

RESOURCES