In a Stockholm emergency clinic for women who had been raped, researchers asked 298 women, within a month of the assault, what their bodies did while it happened. Seventy per cent reported significant tonic immobility: an involuntary paralysis in which the body will not move or cry out. Forty-eight per cent reported extreme immobility (Möller et al., 2017). Their bodies did what frightened bodies do. Then, some day, one of them sits across from someone whose job is to decide whether she's telling the truth. What does that person see?
This article discusses sexual assault. 1800RESPECT is on 1800 737 732, any time.
What her body does next
Six months on, the women who froze were 2.75 times more likely to have post-traumatic stress disorder and 3.42 times more likely to have severe depression. The researchers end their abstract with a sentence aimed at a courtroom: knowledge of this reaction "is important in legal matters" (Möller et al., 2017, p. 932).
The reason is worse than it first appears. A frozen body offers no resistance, so it gets read as consent (de la Torre Laso, 2024). Later, in the interview room, the same nervous system that locked her limbs can lock her voice. She looks at the floor, answers in fragments, loses the order of events and cannot say what happened first.
Every one of those behaviours is on the world's list of signs that someone is lying.
The world's list
The Global Deception Research Team asked people in 75 countries and 43 languages how you can tell when someone is lying. In the first survey, 2,320 people in 58 countries answered the open question. Gaze aversion led, named by 63.66%. In the second, run in 63 countries, 71.5% agreed that liars avoid eye contact. The authors call it a "pan-cultural stereotype" (Global Deception Research Team, 2006).
Now the evidence. Bella DePaulo and colleagues combined 1,338 estimates of 158 cues from 120 independent samples. Their abstract: liars are less forthcoming, tell less compelling stories and seem more tense, "however, many behaviors showed no discernible links, or only weak links, to deceit" (DePaulo et al., 2003, p. 74).
- Eye contact: d = 0.01 across 32 estimates.
- Gaze aversion: d = 0.03 across 6. This is the world's most trusted lie cue.
- Filled pauses ("um", "ah"): d = 0.00.
- Silent pauses: d = 0.01. Posture shifts: d = 0.05. Self-fidgeting: d = −0.01.
Then the reversal. Truth tellers make more spontaneous corrections (d = −0.29) and more often admit they cannot remember something (d = −0.42). In the authors' words, liars' stories "include fewer ordinary imperfections" (DePaulo et al., 2003, p. 74). The imperfect, self-correcting, "I'm not sure" account is the honest one. The courtroom hears it as the dishonest one.
People aren't good at this, and training doesn't fix it. The largest synthesis pooled 206 documents and 24,483 judges. People classified lies and truths correctly 54% of the time, where chance is 50%. They caught 47% of lies, worse than guessing (Bond & DePaulo, 2006). Investigators trained in behavioural cues do not discriminate truth from lies any better than untrained people; they shift towards calling statements deceptive, and grow more confident while doing it (Meissner & Kassin, 2002).
The stumble
Labs aren't courtrooms, so go to real ones. The Real-Life Trial corpus holds 121 transcripts from real court proceedings, with ground truth from verdicts and exonerations (Pérez-Rosas et al., 2015). Count the language in them and one marker stands out: truthful speakers stumble more. About five ums, restarts and corrections per 100 words, against about three for the liars. It's a small collection, and verdicts aren't always right. But a person who pauses, restarts and corrects themselves in the witness box is, in this data, more likely to be the one telling the truth.
Then there are confessions. Rizzelli and colleagues (2021) published word counts for 98 confessions presumed true and 37 proven false by DNA. False confessions run longer, and the words that outrun even that are the telling ones: "you", "that's", "it's". That's what happened. It's like you said. A false confession isn't a lie told by a liar. It's a script, supplied under pressure, recited by a person who has given up being believed.
Whose truth does it sound like?
An assessor's sense of how an honest person sounds is trained on the people around them. Train a computer lie detector on truthful and deceptive essays by US crowd workers (Pérez-Rosas & Mihalcea, 2014), then show it truthful essays by US and Indian writers. It calls about one in five American truths a lie, and more than half the Indian ones. The writers weren't lying more. The instrument measured distance from the people it was built on. Nobody has tested this on Australian speech. But think of an Aboriginal witness, or a witness speaking English as a second language, read against someone else's baseline.
Autistic people meet a version of it. In one study 1,410 observers watched video interviews with autistic and neurotypical adults: autistic adults telling the truth were judged more deceptive and less credible (Lim et al., 2022).
Then look at who sits across the table. Among 468 men in a Welsh prison, 84.2% reported at least one adverse childhood experience and 45.5% reported four or more (Ford et al., 2020). A method that reads trauma as deceit, used on people soaked in trauma, produces disbelief exactly where the system does most of its work.
"But the cues are just weak"
The best objection comes from Hartwig and Bond (2011): people's judgements track the cues that really differ, and lie detection fails mainly because those differences are small. That's fair, and most of these effects are small, and a stumble alone proves nothing either. Nobody should be convicted or cleared on an um. That's the point. If the cues are this weak, they shouldn't be doing the work they do.
What follows
People who are disbelieved while telling the truth learn. They practise steady eye contact. They rehearse. They cut the "I think" and the "um". Then they sound like a liar in the only way that matters, and they've been taught it by us.
The fix isn't training assessors to read faces better. Faces were never the evidence. A jury told to judge credibility should hear that demeanour performs near chance, as it hears the limits of any other forensic method. Non-accusatory interviewing such as PEACE works at scale (Gudjonsson & Pearse, 2011), and information-gathering methods cut false confessions without losing true ones (Meissner et al., 2014). None of this is exotic. England has done it for decades.
What you can do
When someone tells you something terrible happened to them, don't watch their face. Don't ask why they didn't scream or fight or leave. Now you know why. Let them get the order wrong and correct it. Let them say "I don't remember". You're not the trier of fact. You're the person they told.
If you're on a jury, remember 54%. Judge what's been said, what fits and what can be checked. Ignore the eyes.
And write to your attorney-general: ask whether juries in your state are told what the research says about demeanour, and whether police are still trained to read body language. Return to those 298 women in Stockholm. Seven in ten froze. Nobody should have to perform their own truth well enough to be believed.
Sources
Bond, C. F., Jr., & DePaulo, B. M. (2006). Accuracy of deception judgments. Personality and Social Psychology Review, 10(3), 214-234. https://doi.org/10.1207/s15327957pspr1003_2
de la Torre Laso, J. (2024). The reality of tonic immobility in victims of sexual violence: "I was paralyzed, I couldn't move". Trauma, Violence, & Abuse, 25(2), 1630-1637. https://doi.org/10.1177/15248380231191232
DePaulo, B. M., Lindsay, J. J., Malone, B. E., Muhlenbruck, L., Charlton, K., & Cooper, H. (2003). Cues to deception. Psychological Bulletin, 129(1), 74-118. https://doi.org/10.1037/0033-2909.129.1.74
Ford, K., Barton, E. R., Newbury, A., Hughes, K., Bezeczky, Z., Roderick, J., & Bellis, M. A. (2020). Adverse childhood experiences: A retrospective study to understand their associations with lifetime mental health diagnosis, self-harm or suicide attempt, and current low mental wellbeing in a male Welsh prison population. Health & Justice, 8(1), 13. https://doi.org/10.1186/s40352-020-00115-5
Global Deception Research Team. (2006). A world of lies. Journal of Cross-Cultural Psychology, 37(1), 60-74. https://doi.org/10.1177/0022022105282295
Gudjonsson, G. H., & Pearse, J. (2011). Suspect interviews and false confessions. Current Directions in Psychological Science, 20(1), 33-37. https://doi.org/10.1177/0963721410396824
Hartwig, M., & Bond, C. F., Jr. (2011). Why do lie-catchers fail? A lens model meta-analysis of human lie judgments. Psychological Bulletin, 137(4), 643-659. https://doi.org/10.1037/a0023589
Lim, A., Young, R. L., & Brewer, N. (2022). Autistic adults may be erroneously perceived as deceptive and lacking credibility. Journal of Autism and Developmental Disorders, 52(2), 490-507. https://doi.org/10.1007/s10803-021-04963-4
Meissner, C. A., & Kassin, S. M. (2002). "He's guilty!": Investigator bias in judgments of truth and deception. Law and Human Behavior, 26(5), 469-480. https://doi.org/10.1023/A:1020278620751
Meissner, C. A., Redlich, A. D., Michael, S. W., Evans, J. R., Camilletti, C. R., Bhatt, S., & Brandon, S. (2014). Accusatorial and information-gathering interrogation methods and their effects on true and false confessions: A meta-analytic review. Journal of Experimental Criminology, 10(4), 459-486. https://doi.org/10.1007/s11292-014-9207-6
Möller, A., Söndergaard, H. P., & Helström, L. (2017). Tonic immobility during sexual assault: A common reaction predicting post-traumatic stress disorder and severe depression. Acta Obstetricia et Gynecologica Scandinavica, 96(8), 932-938. https://doi.org/10.1111/aogs.13174
Pérez-Rosas, V., Abouelenien, M., Mihalcea, R., & Burzo, M. (2015). Deception detection using real-life trial data. Proceedings of the 2015 ACM on International Conference on Multimodal Interaction, 59-66. https://doi.org/10.1145/2818346.2820758
Pérez-Rosas, V., & Mihalcea, R. (2014). Cross-cultural deception detection. Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), 440-445. https://doi.org/10.3115/v1/P14-2072
Rizzelli, L., Kassin, S., & Gales, T. (2021). Language of criminal confessions. Wrongful Conviction Law Review, 2(3), 205-225. https://doi.org/10.29173/wclawr58