Case File 068 · Investigation and Psychology

Why Truthful Accounts Can Still Be Inconsistent

Why a truthful account can shift between tellings, what research says about details remembered later, and when inconsistency does and does not matter.

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

Truthful accounts can be inconsistent because human memory does not store events like a video file. People forget details, recall new ones later, fill gaps with what seems likely, and answer differently depending on how and when they are asked. Research on repeated recall shows that a witness can leave out something they mentioned before, or add a detail they did not mention the first time, and still be largely accurate. A study of children interviewed twice about a real, documented event found their reports were highly accurate but not very consistent. Inconsistency is therefore something to examine, not proof of lying. Some research does find that people whose accounts shift a great deal tend to be less accurate overall, so the honest answer is that it depends on the kind of inconsistency and the reason for it.

Key takeaways

Memory reconstructs events each time we recall them, so the same person can tell the same true story differently.
Details recalled for the first time in a later interview, known as reminiscence, are often nearly as accurate as details reported consistently.
In one study, experienced police detectives underestimated witness accuracy and treated new details as far less reliable than they turned out to be.
Outright contradictions deserve more caution than omissions or additions, but none of them is reliable proof of deception.
UK interview guidance says inconsistencies often come from misunderstanding or misremembering rather than deliberate fabrication.

Why is memory not perfectly consistent?

In a widely cited 1999 review, Harvard psychologist Daniel Schacter grouped memory's everyday failures into seven "sins": transience, absent-mindedness, blocking, misattribution, suggestibility, bias and persistence. The first three are types of forgetting. Misattribution, suggestibility and bias are types of distortion. Persistence refers to intrusive memories that are hard to forget.

Several of these can make a truthful account shift from one telling to the next:

Transience: details fade over time, so a later account may be thinner.
Blocking: a detail that is stored may not be retrievable at a given moment, then comes back later.
Misattribution: a real memory can be linked to the wrong time, place or person.
Suggestibility: information from questions, conversations or media can be absorbed into the memory.
Bias: current beliefs and knowledge can reshape how a past event is remembered.

Schacter argues that these are not design flaws so much as by-products of a memory system that is otherwise adaptive. A memory that keeps the gist and drops the clutter is useful in daily life, but it will not reproduce the same details every time. For a fuller explanation, see why memory is not a perfect recording.

What kinds of inconsistency are there?

Researchers and practitioners usually distinguish between several patterns. They do not carry the same weight.

Type	What it looks like	What research suggests
Omission	A detail from the first account is missing from the second	Common in accurate witnesses; forgetting and different questions both play a part
Reminiscence	A new detail appears in a later account	Often nearly as accurate as consistent details, though findings vary
Elaboration	The same point is described in more detail	Interview methods such as the cognitive interview are designed to draw out more detail
Contradiction	Two statements cannot both be true	Treated with more caution, but may reflect error rather than a lie

Are details remembered later less trustworthy?

A common assumption is that anything a witness adds later must be invented or borrowed from someone else. The research is more nuanced.

In a 2015 study published in PLOS ONE, Alana Krix and colleagues showed participants a film of a staged theft. The participants gave one statement straight away and another a week later. A separate group of experienced police detectives was asked to estimate how the witnesses would perform. The detectives consistently underestimated actual recall. They expected accuracy to drop sharply between the two statements, but it stayed stable. Most strikingly, they judged reminiscent details to be as inaccurate as direct contradictions, when in fact those new details were almost as accurate as the ones reported consistently.

A 2014 study by Gunn Astrid Baugerud and colleagues at the University of Oslo gave a rare real-world test. A researcher was present when children were removed from their families by child protective services and documented what happened. The children were interviewed one week and three months later. Their reports were highly accurate at both points, but consistency was fairly low: many new accurate details appeared in the second interview, and much accurate information from the first was left out. Older children were more consistent than younger ones. The authors conclude that low consistency does not imply inaccuracy.

Does inconsistency ever signal a less reliable witness?

Yes, sometimes, and it is important not to overstate the reassuring findings. In a 2016 paper, Sarah Stanley and Aaron Benjamin found that reminiscent memories in their experiments were less accurate than memories reported consistently. However, they were just as accurate as memories reported initially and then dropped, which suggests the key factor is inconsistency itself rather than the lateness of the detail. They also found that people with more inconsistent recall were less accurate overall, and noted that if this holds in more realistic settings, the legal assumption that inconsistent witnesses are less trustworthy may have some basis.

Taken together, the studies suggest a balanced reading. Inconsistency can be a sign that a particular memory is weaker, but a truthful witness can still be inconsistent, and new details are not automatically false.

How can the questions themselves create inconsistency?

Different questions draw out different memories. If a first interview never asked about the colour of a jacket, the jacket may appear only in the second. A more suggestive question can also change what is reported, as the classic research on open-ended versus leading questions shows.

The Ministry of Justice guidance Achieving Best Evidence in Criminal Proceedings warns that when an interview is split across sessions, repeating the same focused questions over time "could lead to unreliable or inconsistent responding in some witnesses". The problem of repeated questioning and memory is a topic of its own.

How does interview guidance treat inconsistencies?

The same guidance is direct about their likely causes. It says that when witnesses give misleading accounts, these "are often the result of misunderstandings or misremembering rather than deliberate fabrication."

It then sets out principles for when to ask a witness about an inconsistency:

only when the inconsistency is significant;
only after possible explanations have been considered;
only after the account has been fully explored;
in a planned, tactful and non-confrontational way;
remembering that the aim is an accurate account, not pressure to change it.

It also advises taking into account how vulnerable the witness may be to suggestion, compliance or acquiescence.

Why can't inconsistency be used as a lie detector?

Because truthful and deceptive accounts overlap. A 2006 meta-analysis by Charles Bond and Bella DePaulo, covering 206 documents and 24,483 judges, found that people correctly classify lies and truths only 54% of the time on average. Inconsistency is one more cue that feels meaningful but has innocent explanations: forgetting, stress, a different question, a misunderstanding, or simply a memory that returned later.

The guidance above reflects this: an inconsistency is a reason to explore and check an account against other information, not a conclusion in itself. That process is explored in how investigators find inconsistencies in a story and in the related question of the difference between a lie and a false memory.

How does this connect to Murder Book?

Murder Book: The Art of Interrogation is an upcoming narrative interrogation game from Gungrounds, an independent studio in Zagreb. Its cases focus on unreliable stories, narrative gaps, avoidance and contradictions, and the press kit lists behaviours such as Fabrication, Avoidance, Narrative Gaps and Emotional Leakage. Development screenshots also show an evidence board that links statements together.

The research above is a good lens for that kind of play. In real life, a gap or a changed detail can come from an honest memory as easily as from a lie, and behaviour on its own is not proof. A contradiction is the start of a question, not the end of one. Development materials also show a case select screen that includes a "not guilty" result, a reminder that uncertainty cuts both ways. For an overview of the game, read What Is Murder Book?

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