Is It Time to Leave the Shopping Mall Behind? Measurement Flaws, Plausibility, and External Validity of False Memory Research

This commentary discusses the recently published article by Andrews and Brewin (2024) that reanalyzed data collected by Murphy et al. (2023) to replicate the well-known “lost in the mall” study first published by Loftus and Pickrell (1995). We begin by outlining initial and more recent findings that brought the “lost in the mall” paradigm to the forefront of false memory research before considering the thought-provoking results of the reanalysis by Andrews and Brewin (2024). We then highlight some of the implications of the reanalysis for child sexual abuse investigations, and more broadly, for the reliability and validity of psychological research that relies on researchers' coding and interpretation of information provided by participants about the content of their memories. We ask whether the definition and measurement of false memories within laboratory experiments can be meaningfully applied to real-life debates concerning justice for alleged victims and perpetrators of sexual abuse.

In the 1970s Elizabeth Loftus and her team conducted a series of highly influential experiments demonstrating that misleading information received after a personal experience can lead people to make mistakes when they later try to describe what happened (Loftus and Palmer 1974; Loftus 1975). After establishing the impact of misinformation on memory for personal experiences, an innovative research paradigm was designed to demonstrate that memories of entire events that never occurred could be implanted in people's minds with relative ease. Loftus and Pickrell (1995) misled 24 adult participants to believe that their family members provided descriptions of four true past events, but unbeknownst to the participants, one of the supposed true events, being “lost in the mall”, was made up by the researchers. After participants were told that they had been lost in the mall many years earlier they were then asked to recall what they could remember in writing and verbally and rate the clarity of their memories. The results showed that a quarter of the participants were successfully induced to claim that they remembered the false event, although their average clarity ratings for the false memory were substantially lower than scores assigned to true events. (1) The “lost in the mall” study resulted in a “veritable explosion of cognitive research on the topic of false memory” (Pezdek and Lam 2007), (2), and led to the establishment of a new view of human memory as being particularly fragile and easily manipulated.

However, while most memory researchers accept that false memory implantation is possible, the proportion of people who can be induced to develop false memories has been the subject of fierce debate (Wade et al. 2002). Scrutiny of false memory implantation experiments identified two main challenges concerning the definition of false memories: (1) differentiating between false beliefs and false memories, and (2) differentiating between flawed memories and false memories. The first challenge stems from the difficulty of determining whether participants in memory implantation studies genuinely remember the false event or simply believe the researchers' assertion that the false event had occurred. Failure to differentiate the former (false memories) from the latter (false beliefs) could easily inflate the rate of false memories reported in memory implantation studies (Wade et al. 2002). The second issue results from the potential confusion between false memories, defined as entire false events that have been implanted into memory, and flawed memories, referring to incorrect details in otherwise true memories (Pezdek and Lam 2007). This is a particular concern for versions of the memory implantation paradigm that rely on relatively common experiences, such as being lost in the mall. In these studies, there is a non-negligible probability that some of the participants have truly experienced situations like the false events suggested, leading again to inflated rates of false memories reported by researchers.

The debate about defining and categorizing false memories has been reignited by the recent publication of a much larger scale replication of the original “lost in the mall” study by Murphy et al. (2023). Aiming to address prior criticism of the research paradigm, the researchers introduced a novel coding scheme that explicitly differentiates complete false memories (substantial remembering of all central details relating to the false memory), and partial false memories (partial recall of the false memory). Results showed a slightly higher rate of successful memory implantation than the 25% reported in the original “lost in the mall” study (Loftus and Pickrell 1995), with Murphy et al. (2023) classifying 8% of participants as having full false memories, and a further 27% as having partial false memories. However, consistent with the findings of the original study, participants' clarity ratings of the false memories were relatively low, and consistently lower than those of true memories. Importantly, fewer than half (14%) of the participants judged by researchers as having false memories self-reported remembering the event. In addition to publishing their findings, Murphy et al. (2023) made their data file and raw data public allowing for the possibility for other researchers to re-analyse their data.

Andrews and Brewin's (2024) reanalysis of Murphy et al. (2023) data explicitly aimed to address the aforementioned criticisms of the “lost in the mall” paradigm: (1) that researcher-identified “false memories” may not reflect genuine remembering, and (2) that researcher-identified “false memories” may actually be distortions of participants' true memories. To investigate these concerns, the authors devised a more systematic coding approach that relies on counting the number of core details participants report about the suggested event and assessing the clarity of each of those details on a scale from “no mention” to “explicit recall”. The researchers also attempted to identify potentially true experiences by coding for mentions of being lost in different circumstances to the fake event, being lost on more than one occasion, or experiences similar to the target event that did not actually involve being lost.

The findings of the reanalysis cast doubt on claims that 25%–35% of people can be “led to remember entire events that never actually happened to them” (Loftus and Pickrell 1995, 725). The re-coded data indicated that on average, participants judged by Murphy et al.'s (2023) research team as having a false memory explicitly recalled only 1.47 of the 6 core details. Even the participants judged to have full false memories tended to recall fewer than half of the core details, and 20% did not explicitly recall the most fundamental detail of actually “being lost”. As noted by Murphy et al. (2023), the rate of participants who self-reported remembering the target event (14%) was substantially lower than the rate of participants judged by the researchers as having developed a false memory (35%). Andrews and Brewin (2024) showed that participants' own criteria for remembering was related to the clarity of core details; participants who self-reported having developed a false memory explicitly mentioned significantly more core details than those who did not believe they remembered the event.

The results of the re-analysis also validate prior concerns that some participants judged by researchers as having developed false memories may be referring to potentially true experiences. According to Andrews and Brewin (2024), 31% of participants produced descriptions of past experiences that were similar to the fake event but distinguished by key differences in core details, such as being lost in a different shopping location or being abandoned rather than being lost. The impact of these potentially true experiences on false memory rates was not negligible, as they were present in the accounts of 50% of those judged by Murphy et al. (2023) as having full false memories and 52% of those with partial false memories.

Based on the results of their reanalysis, Andrews and Brewin (2024) concluded that previous studies using the “lost in the mall” paradigm have substantially overestimated the proportion of people who have developed false memories. The authors suggest three steps to improve the methodology of memory implantation studies: the exclusion of participants with potentially true experiences of the target event, the use of core details as minimum criteria for false memories, and the consideration of self-report measures alongside researcher identification of false memories. Using a step-by-step exclusion approach, Andrews and Brewin (2024) demonstrated that applying these methodological improvements to the data of Murphy et al. (2023) resulted in a drastically lowered false memory rate of only 4%. In their recent commentary, Wade et al. (2025) question the validity of this figure, suggesting that the authors' criteria exclude genuine false memories constructed from a combination of suggested details and memory traces from other sources. Nonetheless, Andrews and Brewin's (2024) argument that entirely false memories are more infrequent than memory implantation research would lead us to believe has implications both for the real-world application of the concept of false memories, and for the use of researcher-coded data in the field of memory research.

In the 30 years since the publication of the original “lost in the mall” study, the results of false memory research have been applied far beyond “recovered memory” cases, with the apparent ease of memory implantation and high rate of participants who develop false memories leading to a widespread view that false memories of child sexual abuse are common and have likely resulted in an unknown number of miscarriages of justice (Blizard and Shaw 2019; Crook and McEwen 2019). Concerns about miscarriages of justice have also been expressed by Wade et al. (2025) commentary on Andrews and Brewin (2024), stating that “false memory rates in the lab might underestimate those in real cases, where factors are present that research has shown can exaggerate the likelihood that false memories are formed” (3). We argue that when it comes to claims of child sexual abuse, the opposite is true, with memory implantation experiments giving the impression that accusations based on false memories are more common than they really are. This inflated false memory rate occurs partly because lab studies rely on a specific set of highly suggestive techniques to induce false memories, and partly because lab studies fail to account for factors that reduce the likelihood of false allegations in real cases.

To evaluate the claim that laboratory research provides a conservative estimate of the frequency of false memories in real cases, it is helpful to break down the numerous methods of suggestion and deceit that the “lost in the mall” paradigm and other memory implantation designs rely on to convince participants that they experienced a fake event. Firstly, the researchers provide the participant with the core details of the event in order to “remind them” of what happened, including the main action, the time and location of the event, the participant's emotional reaction, and the resolution of the crisis (Loftus and Ketcham 1994; Loftus & Pickrell; Murphy et al. 2023). These researcher-provided elements provide a coherent narrative framework that serves as a script or schema, making it easy for the participants to “fill in” the details even if they have not personally experienced them. Secondly, the participant is led to believe that these core details were provided by a trusted family member who was present when the event occurred. Given that the experimental paradigm involves no stakes for the participant, the relative, or anyone else, there is no reason for participants to suspect that their relatives would mislead them. Moreover, the description of the false event is presented after the participant has read the summaries of true memories, eliminating potential doubts about the veracity of the accounts. Not all memory implantation studies involve these “tricks” of suggestion, but those that alter the proven formula tend to add a different element of deception, such as showing participants fake photographs or materials related to the target event (e.g., Braun et al. 2002; Wade et al. 2002). Furthermore, false memory studies also rely on repeated recall attempts (three including the booklet and consecutive interviews) to maximize the likelihood that participants will acquiesce to suggestion. Although adults and children can recall events they truly experienced accurately across multiple interviews, a wealth of research has demonstrated an increase in erroneous details resulting from a combination of suggestive techniques and repeated recall occasions (La Rooy et al. 2009). Thus, despite some claims to the opposite, implanting false memories is no simple matter and requires the use of a specific set of highly suggestive techniques under laboratory conditions.

Perhaps even more importantly, real cases involve factors that research has shown can reduce the likelihood of false allegations of child sexual abuse, including the implausibility of the event, children's reluctance to disclose abuse, and the presence of procedural safeguards to prevent miscarriages of justice based on false memories. Andrews and Brewin (2024) highlight that one criticism of the “lost in the mall” paradigm is that being “lost” is a common or plausible event. Even if we have not experienced being lost ourselves, most of us have a schema for it; that is being lost is a theme of many books, television shows and other media, so it is easy to imagine what it would be like to be lost. Most children, however, without the exposure to extreme suggestion, do not have a schema for child sexual abuse. Pezdek and Hodge (1999) conducted a study where they looked at younger (5–7 years) and older (9–12 years) children's susceptibility to accept a false memory for plausible (i.e., being lost in the mall) versus implausible events (i.e., receiving a rectal enema). Most children in this study did not report remembering either false event, but those who did were far more likely to recall the plausible event than the implausible event. The study's conclusion that false memories are not likely to be implanted for less plausible events is consistent with research showing very low rates of acquiescence to false suggestions of genital touch during a real medical examination, even among the youngest children (Saywitz et al. 1991). This is important as we consider how the “lost in the mall” paradigm has been, in our opinion, inappropriately applied to child sexual abuse cases in the courtroom. In the “lost in the mall” paradigm, children are tested about a plausible memory event that they are told their parents said was true, and that is embedded in other true events.

Beyond the implausibility of sexual abuse narratives for children who have not experienced abuse, we also know that children who have experienced abuse are reluctant to disclose to adults (Lyon et al. n.d.), implying that children are far less likely to report false narratives of sexual abuse than false narratives pertaining to innocuous events. Motivational barriers to disclosing sexual abuse include children's concerns about their parents' reactions and the perceived negative consequences of the allegations for the child and the family (Lemaigre et al. 2017). Furthermore, children who were abused by a family member or groomed by a trusted adult often express feelings of love and care toward the perpetrator and are reluctant to report the abuse due to concerns about the consequences the abuser may face (Christensen et al. 2015; Lemaigre et al. 2017). Finally, many victims feel shame, guilt, and self-blame about their abuse (Alaggia et al. 2019; Goodman-Brown et al. 2003; Hershkowitz et al. 2007), further increasing their reluctance to disclose. Indeed, these barriers to disclosure are so strong that an estimated 50% of substantiated victims initially deny abuse when questioned (Lyon et al. n.d.), suggesting that false denials of abuse likely pose a much greater obstacle to justice than allegations based on false memories.

In addition to internal and external factors that reduce the likelihood of real victims making allegations based on false memories of sexual abuse, Andrews and Brewin (2024) note the presence of procedural safeguards against the impact of false memories in real cases, such as the extra scrutiny of jury trials in adversarial legal systems. Perhaps anticipating this criticism, Murphy et al. (2023) extended the original memory implantation design with a mock jury experiment in which 1024 lay “jurors” were asked to read participants' descriptions of the “lost in the mall” event and provide a yes/no judgment regarding whether they reflect genuine memories. Mock jurors believed that memories of the fake event were real even more frequently (39%) than the researchers (35%), demonstrating that lay observers may find it difficult to distinguish between true and false memories. The authors explain their decision not to warn mock jurors that some memories may be false by stating their aim to “mirror the experience of real jurors listening to a witness describe events from their past” (822). However, we disagree with the claim that the researchers created a realistic trial scenario, as real-life cases typically involve opening and closing statements, cross-examination of witnesses, and specific jury instructions around credibility and the burden of proof. Real jurors are often told by defense attorneys in the opening statement that witnesses may lie, and indeed the goal of the defense throughout the trial is to cast doubt on the truth of prosecution's evidence. Furthermore, before a case even reaches the courtroom, the credibility of the child's disclosure is scrutinized many times by many different professionals. Having examined 500 reported cases of child sexual abuse in the United States, Block et al. (2023) found that only 53% of cases were investigated and as few as 17% progressed to court, demonstrating the rigorous criteria cases must meet to move forward in the criminal legal system. Thus, we are inclined to agree with Andrews and Brewin's (2024) conclusion that memory implantation studies overestimate the proportion of participants who develop false memories that observers would judge genuine in a legal context.

In conclusion, we argue that the external validity of laboratory research on false memory implantation is too low to meaningfully inform real investigations involving allegations of child sexual abuse. Although, as demonstrated, laboratory research relies on multiple highly suggestive techniques to induce false memories, Wade et al. (2025) are correct in pointing out that real cases may involve additional suggestive influences that are not present in the laboratory. However, the impact of these is likely outweighed by the presence of factors that reduce the likelihood of false allegations, and procedural safeguards that mitigate the risk of false memories resulting in miscarriages of justice.

In addition to the implications for real-life sexual abuse cases, the results of Andrews and Brewin's (2024) re-analysis also raise questions about the reliability and validity of coding approaches used widely within some areas of psychological research. If re-coding the same dataset with a different coding approach leads to wildly different conclusions with regard to the main hypotheses of a study, how can we trust the results of any research relying on researcher-coded data?

Firstly, researchers relying on manually coded raw data must ensure that their coding is reliable, meaning that their coding guide contains clear rules that the coders follow objectively and accurately. In psychology, reliability is generally assessed through measuring inter-rater agreement among multiple coders, most commonly by calculating Cohen's Kappa. Although there are variations in the field, it is generally accepted that Kappa values at or above 0.8 (sometimes 0.7) reflect almost perfect agreement between coders, signifying a reliable coding approach. Both Murphy et al. (2023) and Andrews and Brewin (2024) fall short of this established standard, with inter-rater agreement as low as k = 0.60 (Murphy et al. 2023) and k = 0.49 (Andrews and Brewin 2024). These figures reflect at best moderate agreement between coders and are approaching the lower limit of what could be considered “reliable” coding. Given the low inter-rater agreement figures, it is questionable whether the results of Murphy et al. (2023) and Andrews and Brewin (2024) could be replicated even if a research team used the exact same coding guide as the original studies.

Low reliability figures limit the strength of conclusions one can draw from quantitative analysis, so it is unfortunate that neither Murphy et al. (2023) nor Andrews and Brewin (2024) highlight the relatively low inter-rater agreement achieved by coders (k = 0.60 and 0.49 at the lowest, respectively) when discussing their findings. Difficulties with achieving reliability in coding may be indicative of more pervasive problems with the study design, such as challenges with operationalizing vague concepts like “partial memory” (Murphy et al. 2023). When objective definitions of psychological concepts prove elusive, an alternative approach is to deconstruct the variables in question into smaller, more easily circumscribed components. Despite Andrews and Brewin's (2024) detail-focused coding guide aiming to capitalize on this approach, there was still substantial inter-rater disagreement with regards to two out of the six core details identified by the authors. One potential contributor to low interrater reliability is human error, which might play a significant role in research designs relying on manual coding of large amounts of complex data even when the coding categories are clearly defined. In this respect, the development of machine-assisted coding approaches is a promising avenue for increasing the reliability of research studies, as machine learning models have been found to outperform manual coders in accuracy when coding interview transcripts (Szojka et al. 2025). Although training machine models still requires an initial dataset of reliably coded data, the trained model can then be applied to new datasets that rely on the same coding categories. This method has the advantage of providing a standardized coding approach that can be used across studies addressing the same research question, as is the case for the numerous direct and quasi-replications of the “lost in the mall” experiment.

However, as Andrews and Brewin (2024) demonstrate, questions about the results of the “lost in the mall” study and its replications go beyond the reliability of the coding guide and concern its validity; the extent to which researcher-defined false memories reflect genuine remembering on the part of the participant. Even if coders in Murphy et al. (2023) achieved perfect reliability, re-analysing the data with a new operational definition of false memories may produce different results. The Andrews and Brewin (2024) re-analysis is not the first study to raise questions about the validity of the definition and measurement of false memories in memory implantation designs. In 2015, a team of researchers criticized the astonishing findings of a study that reported successful induction of false memories of committing a crime in 70% of participants (Shaw and Porter 2015), arguing that the false memory rate was inflated by the authors' failure to differentiate between false beliefs and false memories (Wade et al. 2018). Wade et al. (2018) recoded the data of Shaw and Porter (2015) using two separate coding schemes that distinguish false beliefs from false memories (Lindsay et al. 2004; Scoboria et al. 2017) and obtained a more conservative false memory rate of 26%–30%. Highlighting the stark difference in results between the two analyses, the authors suggest that eschewing established coding approaches in favor of new definitions leads to imprecision that “fuels skepticism of memory research and detracts from the understanding of real-world behavior” (Wade et al. 2018, 474).

Andrews and Brewin's (2024) re-analysis of the data collected by Murphy et al. (2023) is based on the premise that the relatively well-established approach of categorizing partial and full false memories itself lacks validity, necessitating the introduction of a new conceptualisation of false memories. In line with Wade et al. (2018) suggestions, Andrews and Brewin (2024) ensured that their approach is clearly positioned in relation to previous research by (1) providing a comprehensive description of how the data was coded, (2) explaining their motivation for developing an alternative coding approach, and (3) reporting their findings alongside the results obtained with a different coding guide by Murphy et al. (2023). While these steps certainly contribute to making the dialog about the definition and measurement of false memories in the field more transparent, they ultimately cannot answer the question of which of the many definitions of false memories is correct. To determine whether the concept of false memories suggested by Andrews and Brewin (2024) improves on previous definitions used by Murphy et al. (2023) or indeed Loftus and Pickrell (1995), researchers need to examine the world beyond controlled experimental conditions and investigate the meaning and usefulness of the concept of false memories in real cases.

In conclusion, if the question is whether the findings of Loftus and Pickrell's (1995) “lost in the mall” study, one of the most influential and surprising experiments in the history of memory research, can be replicated, the answer has to be a confident “yes”. Over a period of 30 years, a multitude of studies, including a meta-analysis (Scoboria et al. 2017) and a direct replication (Murphy et al. 2023), confirmed that under strictly controlled experimental conditions, a substantial minority of participants can be misled to report details of fake events as if they remembered them. The contribution of Andrews and Brewin (2024) is to move the debate away from the reliability of the false memory phenomenon and challenge its validity. Is it meaningful for researchers to state that an individual has developed a false memory if the participant herself does not think she remembers it? Would a researcher-identified false memory constitute credible evidence at court? Renewed interest in these questions supports our view that it remains inappropriate to interpret false memory rates reported by laboratory studies using a memory implantation paradigm as evidence that a substantial proportion of real-life allegations of child sexual abuse are based on false memories. Even if further replications of the “lost in the mall” study were able to introduce reliable and valid methods of measuring false memories, the paradigm itself will still fail to account for the real-life context of child abuse investigations. We argue that the aim of preventing miscarriages of justice is better served by observational studies and new, ecologically valid research paradigms than by the continued deconstruction and reconstruction of a 30-year-old experiment.

Zsofia A. Szojka: writing – original draft, writing – review and editing, conceptualization. Stephanie Block: writing – review and editing, conceptualization. David La Rooy: writing – review and editing, conceptualization.

An ethics approval was not required, as this study is based exclusively on published research.

The authors declare no conflicts of interest.