Case File 062 · Investigation and Psychology

Can You Actually Tell When Someone Is Lying?

What decades of research say about how accurately people can tell lies from truths, and why behaviour alone is a weak guide.

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

Usually not very well. The largest review of the research, a 2006 meta-analysis by Charles Bond and Bella DePaulo covering 206 documents and 24,483 judges, found that people who try to tell lies from truths in real time, with no special aids or training, are right about 54% of the time. Guessing would give 50%. People were better at recognising truths (61%) than lies (47%), partly because they tend to assume others are honest. A later meta-analysis by the same authors found that differences between individual judges are tiny, so there is little evidence of "human lie detectors". The main reason is not that people watch the wrong things. It is that the behaviour of liars and truth tellers differs only a little. Accuracy improves most when an interview draws out useful differences and when statements can be checked against evidence.

Key takeaways

On average, people judge lies and truths correctly about 54% of the time, only slightly better than chance.
People are better at spotting truths than lies because they lean towards believing others.
Differences in lie-detection ability between individuals appear to be very small.
The main limit is that behavioural differences between liars and truth tellers are weak.
Research suggests that checking statements against evidence can help more than watching body language.

What does "54% accuracy" actually mean?

In a typical deception study, researchers know the truth. Some participants tell the truth about something and others lie, and observers judge each statement. Accuracy is the share of correct judgements.

Bond and DePaulo pooled results from hundreds of these studies. Their headline figure was 54% correct overall, made up of 61% of truths correctly judged as truthful and 47% of lies correctly judged as lies. Two further findings from the same review are worth knowing:

People regard their conversation partners as honest. This "truth bias" helps explain why truths are recognised more often than lies.
People were more accurate judging lies they could hear than lies they could only see. Watching the face and body alone was the weaker option.

The authors also note that the overall ability to tell lies from truths is not trivial when compared with other effects in social psychology. It is real, but small. In practical terms, a 54% hit rate is a poor basis for deciding whether a specific person is lying.

Are some people natural lie detectors?

The idea of a gifted lie catcher is popular in fiction, but it has little support in the research summaries. In a 2008 meta-analysis of 247 samples, Bond and DePaulo corrected for random measurement error and found that differences in the ability to detect lies are minute. Judges ranged no more widely than chance alone would predict.

What people did differ in was bias: how inclined they were to see others' statements as truthful. The review also found that the people being judged differ more than the judges do. Some people simply come across as more believable, whether they are lying or telling the truth, and the outcome of a judgement depended more on that credibility than on any other individual difference.

That has an uncomfortable implication. An honest person who happens to seem nervous or awkward can be judged as deceptive, while a confident liar can be believed.

Why are people so bad at spotting lies?

A common explanation is that people rely on myths, such as "liars look away". A 2011 set of meta-analyses by Maria Hartwig and Charles Bond tested that explanation using a lens model, which compares the cues people use with the cues that actually differ between liars and truth tellers.

Their conclusion was that people rarely rely on the wrong cues. The cues people use to judge deception overlap substantially with the cues that differ between liars and truth tellers. The problem is that those differences are weak. Even a perfect observer of behaviour would have little to go on.

This fits a wider review in the Annual Review of Psychology by Aldert Vrij, Maria Hartwig and Pär Anders Granhag (2019). It concluded that the nonverbal cues to deceit found so far are faint and unreliable, and that people are mediocre lie catchers when they pay attention to behaviour. The article Why There Is No Single Reliable Sign of Deception looks at the cue research in more detail.

Do trained professionals do better?

Training in "reading" body language has a weak scientific basis, for the reasons above. Research points in a different direction: professionals tend to do better when the interview is designed to produce useful information, rather than when they simply watch more closely.

One example is research on the Strategic Use of Evidence (SUE) approach. In a 2006 study published in Law and Human Behavior, police trainees interviewed mock suspects who were either lying or telling the truth about a staged event. Trainees who had been taught to use the available evidence strategically reached 85.4% accuracy, compared with 56.1% for untrained trainees. The trained interviewers were better at bringing out inconsistencies between what liars said and what the evidence showed. The article How Investigators Find Inconsistencies in a Story covers this kind of work at a high level.

Another line of research, the cognitive approach to lie detection, focuses on questioning methods that are harder for liars than for truth tellers. It has produced promising results in laboratory studies, but other researchers have argued that the size of the benefit has been overstated. See What Is Cognitive Load in an Interrogation? for both sides.

These are experimental findings with mock suspects. They show what can help under controlled conditions, not that any method can reliably identify a liar in a real case.

What are the limits of this research?

Deception studies have to know the truth in order to score accuracy, so most take place in controlled settings where participants lie or tell the truth on instruction. That makes accuracy measurable, but it also means real-life stakes, relationships and context are often missing.

The research also cannot be turned into a rule for everyday life. A shaky voice, a pause or a change of story can have many explanations: stress, embarrassment, fear of not being believed, fatigue, or an ordinary gap in memory. Truthful accounts can be inconsistent too, as explained in Why Truthful Accounts Can Still Be Inconsistent.

So can you tell when someone is lying?

Sometimes, but not reliably from behaviour alone. The fairest summary of the research is:

On average, people perform only slightly above chance.
Nobody has been shown to have a special gift for reading lies.
In experiments, accuracy has improved when interviewers used evidence to test what people said.
Any single impression, however strong, is a reason to ask more questions, not a conclusion.

That is also why investigators are warned against deciding early that someone is lying. A strong first impression can shape the rest of an investigation, a problem covered in How Confirmation Bias Can Affect an Investigation.

How does this apply to Murder Book?

Murder Book: The Art of Interrogation is an upcoming narrative interrogation game from Gungrounds, a studio based in Zagreb, Croatia. It is inspired by real police footage and psychological interviews. Players examine unreliable stories, behavioural changes, narrative gaps, avoidance and contradictions. The press kit names behaviours such as Fabrication, Avoidance, Narrative Gaps and Emotional Leakage.

The 54% finding is a useful way to approach a game like this. Spotting a behaviour is not the same as catching a lie. A suspect who avoids a question may be hiding something, or may be frightened, confused or protecting someone for reasons unrelated to the case. In Murder Book, a behaviour is best read as a reason to look closer: compare it with the rest of the account, check it against the evidence and ask what else could explain it. Uncertainty is part of the challenge, not a flaw in it.

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