Is Memory Reliable?

by Lola Henry

chalk dust on chalkboard
Photo by Noita Digital on Unsplash

Humans extensively rely on memory in all aspects of life, from remembering everyday tasks to performing important societal functions, such as using the memories of a witness to determine guilt in the criminal justice system. An understanding in the reliability of memory in a range of applications is therefore critical. A distinction is made between reproductive memory, which involves recitation of rote information, and reconstructive memory, which involves the reconstruction of complex events. This paper conducts a review of reliability of reproductive and reconstructive memory, and the implications for their use in real world applications. Detailed reconstructive memory, such as that used in eyewitness testimony, is highly malleable which renders details unreliable. Reconstructive memory may be altered by emotional state during the event or while remembering it, the phrasing of questions, or questions about details not originally coded into memory. Therefore, witness accounts should not be used as primary evidence unless there is no alternate evidence or recollection of an event. Reproductive memory less malleable, however individuals are unable to distinguish between directly asserted and pragmatically
implied information in reproductive memory, which may result in unreliability. As a whole, the reliability of memory has a wide range of implications; from court procedure to advertising ethics to acceptable psychiatric methods the assumption of a reliable memory system is engrained in societal procedures.

Psychology, memory, reliability, testimony

The reliability of memory is often assumed. People generally accept that they will be able to recount events they have experienced accurately, so much so that historically many people have been convicted of crimes primarily on eyewitness testimony. While the reliability of memory, or lack thereof, may not derail day-to-day life, it may have a significant impact on the lives of individuals and society as a whole. Inaccurate recollections of performance may affect which schools or jobs one applies to, a well-meaning witness may contribute to the incarceration of an innocent, and a psychiatrist may mistakenly plant memories of a phantom trauma. Understanding the reliability of memory is therefore critical to address a wide range of both public and individual interests. Much of the research regarding memory reliability suggests that memory is highly malleable; thus, we should not rely on information from memory in situations where detailed accuracy is needed.

While memories can be classified into particular types, memory generally refers to the ability to recall information that an individual has previously been made aware of. Information in this sense does not merely refer to factual information but any stimulus input consciously perceived. For example, people may remember sensory information such as smell and spatial input in addition to abstract factual information such as the capital of a country. Roediger and McDermott (1995) make a distinction between reproductive memory, which consists of rote information and is unlikely to result in errors, and reconstructive memory, which involves an active process of filling in missing elements while remembering and is significantly more prone to errors. Loftus and Palmer (1974) propose for complex occurrences that information from the initial perception of the occurrence and later from supplementary information over time become integrated into memory in such a way that the individual cannot disentangle from which source the information came. As a result, an individual cannot determine what is remembered from the experience and what was learned about the event afterward. Similarly, Harris (1977) finds that individuals remember directly asserted and pragmatically implied information in the same way and are unable to distinguish in which way the information remembered was presented, which may have significant implications for advertising ethics. Studies have also shown that increasing the salience of gender stereotypes influences self-reporting of prior school marks in line with what would be stereotypically expected (Chatard et al., 2007). For example, female students reminded of the stereotypical male dominance in STEM subjects recalled having lower grades in these subjects than they actually obtained. This indicates that even basic, clearly quantified information about ourselves may be remembered inaccurately with minimal suggestion. As a whole, remembering details is an unreliable enterprise given the mechanisms that may alter the memory of detailed events.

One of the most well-known studies in memory reliability to date is the Loftus and Palmer (1974) eyewitness testimony study. To examine leading questions, which are frequently used in court trials and police questioning practices, Loftus and Palmer (1974) had participants watch footage of a staged car crash and answer questions about the video they had seen immediately after and one week after viewing. Immediately after viewing and giving an account of what they had seen, groups of participants were asked variations of the question, “About how fast were the cars going when they hit each other?” For different groups, the word “hit” was substituted with “smashed,” “collided,” “bumped,” and “contacted.” The group that encountered the most intense word—“smashed”—had a mean speed estimate that was nine miles per hour faster than that of the least intense word—“contacted”—with mean speed estimates of 40.8 miles per hour and 31.8 miles per hour, respectively. This finding demonstrates that even the wording of questions can affect the recollected event. A week later, participants who had been asked about the speed with the word “smashed” were over twice as likely to report remembering seeing broken glass as those questioned with the word “hit,” despite no broken glass appearing in the video. Such a result demonstrates that the influence of leading questions not only affects the individual’s answer when they must infer information based on their recollection (such as speed) but also can affect their explicit memory of the event they witnessed. This study demonstrates that an individual’s memory of complex events can be manipulated, rendering complex-event memory and eyewitness testimony unreliable in light of the influence of interview tactics and question wording. Police and courtroom interactions often involve leading questions, which means these questions may alter an eyewitness’s memory of an event in such a way that they are unable to discern what was implied and what was seen.

The powerful effect of questioning on memory has serious implications for the criminal justice system. The Innocence Project reports that eyewitness misidentification contributed to approximately 70% of wrongful convictions overturned by post-conviction DNA evidence in the United States (“Eyewitness Misidentification,” 2019). Additionally, the Loftus and Palmer study (1974) demonstrates how easily people may be influenced to recall an event as more violent than it was. Thus even in cases where eyewitnesses correctly identify a suspect, they may be unknowingly influenced to recall the crime as more severe or more violent than it was if questioned a certain way. Relying on eyewitness testimony in court not only runs the risk of incarcerating innocent individuals but also risks inflating the severity of a crime and the sentencing unless extreme care is taken to ensure the memories remain unadulterated.

Photo by Tingey Injury Law Firm on Unsplash

Complex events are not the only form of memory that can be easily manipulated. In research comparing students’ beliefs about gender stereotypes relating to school performance (e.g., “men are generally better at mathematics,” and “women are generally better at arts”), the more students expressed belief in gender stereotypes, the more likely they were to inaccurately report recalled school marks in a way consistent with gender stereotypes three months later (Chatard et al., 2007). This finding shows that even beliefs not directly related to the task at hand can interfere with the recollection of noncomplex factual information. Chatard et al. (2007) also found that manipulating the saliency of relevant gender stereotypes prior to score reporting led to significant differences in the accuracy of scores reported between high and low saliency groups. Accordingly, the mental state at the time of recollection can influence our recollection of factual material, as well. Although this study noted some limitations (e.g., the length of time between the remembered event and the recall request, as well as the relevance of the remembered event to the current lives of the participants), it was the first to compare reported marks to those actually obtained in order to investigate the effects of self-stereotype biases on factual recollection. Ultimately, we can reasonably say that recollection, even of simple factual information, is not necessarily reliable, particularly if the information pertains to the self. Such an assertion may be of particular importance to those in marginalized communities, as inaccurate reflections influenced by self-stereotypes may damage self-esteem and mental health. Marginalized individuals who are more aware of negative stereotypes about themselves may even avoid pursuing more prestigious academic programs or career paths due to such a harmful effect. Underestimating performance in this way may be a contributing factor to the gender and race gaps observable in many esteemed career fields in the United States.

Even the memory of things that individuals are often exposed to and may be fairly confident in is not necessarily reliable (Blake et al., 2015).  According to Blake et al. (2015), only one in eighty-five participants were able to correctly recall the Apple logo, and half of all participants failed to identify the logo correctly in a set of similar designs. These failures came in spite of the fact that the Apple logo is widely considered one of the most familiar logos in the United States. Here, the inability to accurately remember something so prominent in everyday life may be the result of “inattentional amnesia,” which suggests that the availability of the logo due to frequent exposure paired with its simplicity may lead people to stop attending to the details and, therefore, to encode only an approximate memory of the logo suitable for recognition in context (Blake et al., 2015). This study also found that participants tended to rate their confidence in their memory of the Apple logo significantly higher before as opposed to after completing the tasks involved with the study (Blake et al., 2015). This overconfidence in perceived memory also renders memory unreliable in the sense that people are likely to overestimate how accurate their memory is.

Memory is generally unreliable for highly accurate recollections. While memory appears fairly reliable for what Blake et al. (2015) term a “gist-based” recollection, if something needs to be recollected with complete accuracy, such as in the event of reporting marks or giving eyewitness testimony, memory is not reliable. Similarly, our confidence in the accuracy of our memories also appears to be unreliable, with the literature reporting a multitude of factors that can affect the accuracy of recollection while the individual continues to believe their recollection is accurate (Chatard, Guimond, & Selimbegovic, 2007; Loftus & Palmer, 1974; Roediger & McDermott, 1995). The unreliability of memory is also not limited to complex information or events, but affects simple recollection of personal information. We should, therefore, not rely on memory for information in which perfect accuracy is required, such as in criminal proceedings. If memory must be used for a scenario requiring accuracy, measures must be taken to avoid altering the memory, and information recorded from memory should be flagged accordingly. The details of memory are, in general, malleable. The reliability of memory is often taken for granted, and as a result a substantial part of our lives involve recollecting information whether it be for ourselves or to serve a societal purpose, such as providing eyewitness testimony. As we examine the unreliability of memory it is important that we not only critically evaluate societal processes that rely on memory, but that we critically examine and attempt to mitigate the societal mechanisms which may result in unreliable memories as well.

pile of photographs
Photo by Jon Tyson on Unsplash

Further Reading

Eisen, M. L., Goodman, G. S., Qin, J., Davis, S., & Crayton, J. (2007). Maltreated children’s memory: Accuracy, suggestibility, and psychopathology. Developmental Psychology, 43(6), 1275–1294.

Patihis, L., & Younes Burton, H. J. (2015). False memories in therapy and hypnosis before 1980. Psychology of Consciousness: Theory, Research, and Practice, 2(2), 153–169.

Wixted, J. T., Mickes, L., & Fisher, R. P. (2018). Rethinking the Reliability of Eyewitness Memory. Perspectives on Psychological Science, 13(3), 324–335.


Blake, A. B., Nazarian, M., & Castel, A. D. (2015). Rapid communication: The Apple of the mind’s eye: Everyday attention, metamemory, and reconstructive memory for the Apple logo. Quarterly Journal of Experimental Psychology, 68(5), 858–865.

Chatard, A., Guimond, S., & Selimbegovic, L. (2007). “How good are you in math?” The effect of gender stereotypes on students’ recollection of their school marks. Journal of Experimental Social Psychology, 43(6), 1017–1024.

Harris, R. J. (1977). Comprehension of pragmatic implications in advertising. Journal of Applied Psychology, 62(5), 603–608.

Loftus, E. F., & Palmer, J. C. (1974). Reconstruction of automobile destruction: An example of the interaction between language and memory. Journal of Verbal Learning and Verbal Behavior, 13(5), 585–589.

The Innocence Project. (2019, April 10). Eyewitness Misidentification. Retrieved September 29, 2022, from

Roediger, H. L., & McDermott, K. B. (1995). Creating false memories: Remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(4), 803–814.

Citation Style: APA