Case File 073 · Investigation and Psychology

Why Memory Is Not a Perfect Recording

Why human memory rebuilds events rather than replaying them, and what that means for honest witnesses and investigations.

By Murder Book Editorial TeamPublished 16 September 2026Updated 16 September 2026

Memory is not a perfect recording because the brain does not store an event and play it back unchanged. Psychologists describe episodic memory, our memory for personal experiences, as reconstructive: each time we remember, we rebuild the event from stored fragments, general knowledge, expectations and information we have picked up since. Harvard psychologist Daniel Schacter and his co-author Addis describe it as "fundamentally constructive, rather than reproductive". This usually works well enough for daily life, but it means memories can fade, blend together or absorb new details without the person noticing. Elizabeth Loftus's research showed that even the wording of a question can change what people report remembering. For investigations, the lesson is that an honest witness can be sincerely wrong, and a changed account is not automatically a lie.

Key takeaways

Remembering is an act of reconstruction, not playback, so memories can change each time they are recalled.
In Loftus and Palmer's 1974 study, asking how fast cars were going when they "smashed" rather than "hit" each other raised speed estimates and later reports of broken glass that was never shown.
Schacter's "seven sins of memory" group everyday memory failures into forgetting, distortion and intrusive memories.
Vivid, emotional memories can feel certain while still changing over time.
Memory is often reliable, but its accuracy depends heavily on how and when it is tested.

What does it mean that memory is reconstructive?

A video file is fixed once it is saved. Human memory is not. When you remember an event, your brain pulls together pieces of the original experience and fills the gaps with what usually happens, what makes sense and what you have learned since.

Schacter and Addis (2007) reviewed evidence that episodic memory is prone to "various kinds of errors and illusions" because it is constructive. They argue this is not simply a design flaw. The same system that recombines pieces of past experience also lets us imagine and plan for future situations, and brain imaging research shows considerable overlap between remembering the past and imagining the future.

The practical result is that a memory can feel whole and genuine even when parts of it were added later.

What did Loftus and Palmer's car crash study show?

One of the best-known demonstrations comes from Elizabeth Loftus and John Palmer's 1974 paper, "Reconstruction of automobile destruction".

In the first experiment, participants watched films of traffic accidents and then answered questions about them. The key question asked how fast the cars were going when they "smashed", "collided", "bumped", "hit" or "contacted" each other. Only one word changed, but it made a difference to the answers:

Verb in the question	Mean speed estimate
Smashed	40.8 mph
Collided	39.3 mph
Bumped	38.1 mph
Hit	34.0 mph
Contacted	31.8 mph

In the second experiment, participants watched a film of a multiple-car accident. Some were asked about the cars "smashing" into each other, some about them "hitting" each other, and some were not asked about speed at all. A week later, everyone was asked whether they had seen any broken glass. There was no broken glass in the film. Sixteen of the 50 participants in the "smashed" group said yes, compared with 7 of the 50 in the "hit" group.

The second experiment matters because it suggests the question did not just change how people answered. It appeared to change what they later remembered.

Can people remember things that never happened?

Yes. The misinformation effect, which Loftus has studied for decades, describes how misleading information presented after an event can impair or change memory for that event. In a 2005 review of 30 years of this research, Loftus noted that warnings about misinformation help only under limited circumstances, and that some people are more susceptible than others.

The same review describes a newer line of research showing that people can come to believe falsely that they experienced "rich complex events that never, in fact, occurred". In an earlier paper, Loftus and Hoffman (1989) argued that misinformation research tells us something important about "the creation of new memories".

A false memory is not a lie. The person reporting it believes it. That difference is explored in What Is the Difference Between a Lie and a False Memory?

What are Schacter's seven sins of memory?

In a 1999 article in American Psychologist, Daniel Schacter proposed a way to classify the ways memory goes wrong. He described human memory as often reliable but also fallible, and grouped its failures into seven "sins":

Forgetting: transience, absent-mindedness and blocking.
Distortion: misattribution, suggestibility and bias.
Intrusive memory: persistence, meaning recollections that are difficult to forget.

Schacter argued that although the sins may look like flaws in the design of memory, they are better understood as by-products of features that are otherwise adaptive. He reviewed evidence for each one from cognitive, social and clinical psychology and from cognitive neuroscience.

The distortion sins are the most directly relevant to witness statements. Misattribution and suggestibility are covered in What Makes a Statement Unreliable?

Why do vivid memories feel so certain?

Many people have "flashbulb" memories of where they were when they heard shocking news. These memories feel vivid and permanent. Research suggests the feeling is misleading.

On 12 September 2001, Talarico and Rubin asked 54 Duke University students to record how they had first heard about the terrorist attacks of the previous day, along with an everyday event from around the same time. The students were retested 1, 6 or 32 weeks later. The consistency of both kinds of memory declined over time in a similar way. The difference was that vividness and belief in accuracy stayed high only for the flashbulb memories.

The authors concluded that flashbulb memories are special not in their accuracy but "only in their perceived accuracy". Emotion made people more confident, not more consistent.

Does this mean memory cannot be trusted?

No. Reconstructive does not mean random, and Schacter's own framing starts from the point that memory is "often reliable".

The key is that reliability depends on conditions. Research on eyewitness identification shows this clearly. Wixted and Wells (2017) reviewed evidence that, under "pristine" testing conditions, such as an initial, uncontaminated memory test with a fair line-up and an immediate confidence statement, high-confidence identifications by adults are remarkably accurate. When those conditions are not met, even a confident identification can be seriously compromised.

In other words, the first uncontaminated test of a memory is especially valuable. Every later conversation, question or news report is a chance for the memory to change.

What does this mean for investigations?

Because memory can be shaped by how it is retrieved, investigative guidance focuses on protecting it:

Ask open questions first. The U.S. National Institute of Justice's eyewitness guide recommends open-ended questions and advises avoiding leading questions such as "Was the car red?".
Limit outside influence. The same guide advises separating witnesses and discouraging them from discussing the incident or following media coverage of it.
Interview promptly and record carefully. The UK Achieving Best Evidence guidance notes that the potential for memory contamination increases with delay.
Avoid repeated focused questioning. The same guidance warns that repeating the same focused questions over time could lead to unreliable or inconsistent responses in some witnesses. How Repeated Questioning Can Affect Memory covers this in more detail.

It also means an investigator should expect honest accounts to contain gaps and small changes. Why Truthful Accounts Can Still Be Inconsistent explains why inconsistency on its own is weak evidence of deception.

How does Murder Book reflect unreliable memory?

Murder Book: The Art of Interrogation is an upcoming narrative interrogation game from Gungrounds, an independent studio in Zagreb, Croatia. It is built around unreliable stories, and its press kit lists Narrative Gaps and Fabrication among the behaviours players examine.

Memory research is a useful reminder when playing a game like this. A gap or a changed detail is a reason to look closer and ask more, not a verdict. Someone can misremember without lying, and someone can sound certain while being wrong. In Murder Book, behaviour is treated as a reason to look closer, not as proof. Each case is designed for a focused 5–10-minute session.

Murder Book is in development for PC (Steam) and mobile, with release currently planned for the end of Q2 2027.