How to tell if someone is lying

Not well. **Bond & DePaulo (2006)** analysed **206 studies** involving over 24,000 participants and found average lie-detection accuracy sits at roughly **54%** — just above chance. Trained professionals — police, customs officers, judges — perform no better than untrained observers in controlled experiments. The unsettling implication is that your gut feeling about whether someone is lying is a weak signal at best and a confident illusion at worst. This isn't a reason to abandon judgment; it's a reason to treat any single cue with real scepticism.

No — this is one of the most durable myths in lie detection. The research, reviewed extensively by **Aldert Vrij**, finds no reliable link between gaze aversion and deception. Liars do not systematically look away more than truth-tellers; in fact, some people who know the stereotype actively maintain eye contact _longer_ when lying. What eye contact does signal is **emotional state**, not truth value — and those are different things. Reading someone's honesty from their gaze alone will produce more errors than insights.

A cluster is two or more cues appearing together, rather than in isolation. **Houston, Floyd & Carnicero (Spy the Lie, 2012)** argue that no single signal — a micro-expression, a gesture, a vocal change — carries enough weight alone. What matters is seeing **multiple indicators** in response to the same question or topic, especially when they appear quickly and involuntarily. A single fidget proves nothing; a sudden voice shift, an incomplete answer, and a change in eye contact on the same topic is meaningfully different. Train yourself to notice coincidences, not individual ticks.

Repeating the same question gives a liar the chance to **rehearse and entrench** a consistent version of events. **Houston et al. (Spy the Lie, 2012)** found that varied, unexpected follow-up questions — 'Tell me more about what you were doing before that', 'Walk me back to the start' — disrupt a prepared story more effectively. A truthful account is drawn from memory and remains stable when retold non-linearly; a fabricated one is a _script_, and scripts break under unexpected angles. In personal relationships, varied questions also land as curiosity rather than interrogation, which matters.

Quite a bit, but subtly. **James Pennebaker (The Secret Life of Pronouns, 2011)** found that liars use fewer **first-person singular pronouns** — 'I', 'me', 'my' — than truth-tellers. This appears to be a distancing mechanism: speakers psychologically remove themselves from what they're describing. Beyond pronouns, people who are concealing something often over-explain details that weren't asked about, answer a slightly different question than the one posed, or use unexpectedly formal or lawyerly language. None of these is conclusive alone; combined, they're worth noting.

No. This is the central, uncomfortable conclusion from decades of deception research, summarised by **Vrij's** systematic reviews: there is no single behavioural cue that reliably distinguishes liars from truth-tellers across people and contexts. _Pinocchio's nose does not exist._ Micro-expressions, body language, nervous fidgeting, gaze — all are influenced by anxiety, personality, culture, and context in ways that swamp any lying-specific signal. Anyone selling you a single foolproof tell is selling you something that doesn't hold up in the data.

Baseline is how a particular person behaves when they are **not under pressure** — their ordinary speech rhythm, eye contact pattern, gesture frequency, and verbal style. Change from baseline is far more informative than any absolute behaviour. A person who naturally rambles becoming suddenly terse, or a calm speaker whose voice pitch rises sharply on a specific topic — these deviations are worth noticing. You can only read deviation if you know the normal. This is why lie detection in close relationships, where you know someone's baseline well, is more plausible than detecting deception in strangers.

Yes — though not in the tidy way pop-psychology suggests. **Amy Cuddy (Presence, 2015)** argues that the body expresses emotional states that we haven't consciously decided to display, and that these leakages are real even when someone is trying to conceal them. The practical implication isn't 'crossed arms means lying'; it's that noticing a **mismatch between what someone says and how their body is behaving** in that moment is worth a follow-up question. Congruence — words, tone, and body pointing the same direction — is genuinely informative. Incongruence is a signal to slow down, not an accusation. See also our piece on [reading body language](/en/blog/how-to-read-body-language) for the full picture.

With a lot more humility than interrogation-style tactics suggest. In a personal relationship, the goal isn't to catch someone; it's to understand what's actually happening. That means creating space for the other person to tell the truth — reducing the social cost of honesty — rather than engineering a confession. Ask open, non-accusatory questions. Describe your observation, not your conclusion: 'I noticed you seemed to hesitate when I asked about Thursday' is better than 'You're lying about Thursday.' Our guide on [how to build trust](/en/blog/how-to-build-trust) covers the conditions that make honesty easier in relationships.

Start by examining your own interpretation. **Confirmation bias** means that once you suspect deception, you'll notice cues that fit and discount those that don't. Run the alternative hypothesis: what if they're telling the truth and I'm misreading anxiety or discomfort as guilt? If doubt persists, the most productive path in a personal relationship is a direct, non-hostile conversation about what you observed and why it concerns you — not a covert attempt to catch them out. Deception in close relationships often touches [manipulation and trust](/en/blog/gaslighting-and-manipulation); if patterns are chronic, that framing is more useful than any single lie-detection technique.