You can't read a mind from a face
Lesson 3 of 5 in this module · about 7 minutes · Romantic communication
The idea
A face or a body is a prompt for a question, not a readout, and feeling sure of a read is not the same as being right. Several of the studies below come from research on lie detection, not dating 1. They test how well people read a hidden state from how someone looks and sounds.
In a large meta-analysis, unaided real-time judgments of whether someone was lying averaged about 54% correct, only a little better than a coin flip 1. The average is above chance, so reading from demeanor is weak, not impossible. Judges and situations vary 1. One research team concluded that the behavioral cues to lying are themselves weak, and eye contact, for example, was only weakly related to being judged deceptive 2. That is their interpretation, and it concerns deception cues only 2. The same behavior has many causes: a glance down can come from a lie, a passing thought or a dropped pen.
More information did not always help. In the same meta-analysis, lies were detected better when heard than when seen, and in a mock-theft experiment deceivers seemed more credible to observers with full audio and video 3. That is an interrogation-style setting. It does not show that visual cues are useless for other judgments 3.
Facial movements carry some information about emotion. In a pooled review, people recognized emotions across cultures at better than chance levels, and better within their own cultural group 4. A critical 2019 review found that the same expression can mean different things across people, cultures and situations 4. The field is contested, and that review is one team’s view. So a face offers a prompt, not a statement of what someone feels.
You may have heard that trained observers catch “microexpressions”. In one lab study such expressions appeared in about a fifth of the participants and in 2% of expressions, and in one randomized test a training tool for them did not improve lie detection 5. Each finding is a single lab study, neither one about romance, so the fair summary is “little evidence”, not “proven false” 5.
Flirting is no easier. In one study, strangers watching short speed-dating clips judged who was interested better than chance, but close to chance for women’s interest. Accuracy varied a lot between the people watched, and a later study did not find the effect reliably 6. The record is mixed, and the observers were not the partners.
Confidence is the least reliable part. In two studies, people overrated how well they read others, and the least accurate overestimated the most 7. It rests on one paper and was not tested on dating 7.
Why it matters
A glance or a tight mouth can feel like a message. But in two studies, how well people rated their own skill at reading others tracked their actual accuracy only weakly or not at all 7.
What it looks like
Example
In a Friday meeting you joke, and Leona, 34, a colleague you have talked to a few times, looks down at the table and her mouth tightens. She looked down like that during the budget talk too. Irritated, shy, hiding a smile: each story feels certain in turn. After the meeting you ask, “Was that joke a miss?”
The same moment, handled the other way
Counterexample
Same meeting. By the end you have settled on one story, that she was irritated, and you stop joking with her and avoid her desk. You never asked. The same look has many causes, and a feeling of certainty is not evidence.
The one thing to remember
The rule
Describe what you saw, hold the meaning loosely, and check by asking when a read matters.
“She looked down and her mouth tightened” is something you saw. “She was irritated” is a story. Ask about the thing, not the feeling. When you feel certain, look for a second cue. In this course, a face is a prompt for a question, not an answer.
Try it now
Item 1 of this lesson’s set: “A trained eye catches microexpressions”.
In the real world (optional)
If you like, pause a film scene on one character’s face. Write what you see in neutral words, such as “mouth closed, eyebrows raised”, then the story that came to mind, and notice how fast it arrived.
What this doesn’t tell you
It does not say you can detect nothing. The studies found weak and variable accuracy from demeanor, and a feeling of certainty does not track it. Most of the evidence concerns lies, not romance, so it is a caution and not a measurement of how well you read the people you like.
Why do we believe this?
Each numbered mark in the text opens the evidence behind that sentence. Every claim in this course carries one of five labels.
- Well supported
- A meta-analysis or several independent studies agree, with no serious failure to replicate.
- Supported with caveats
- Several studies agree, with stated limits such as student samples or lab settings.
- Plausible
- One study, or indirect evidence from a nearby question.
- Speculative
- An idea or argument that hasn't been directly tested.
- Marketing claim
- Something a seller says about their product, without published evidence.
Evidence: Well supportedJudging in real time whether someone is lying from how they come across is only slightly better than a coin flip, averaging about 54% correct.
Why this label
A meta-analysis or several independent studies agree, with no serious failure to replicate. Lesson stance: describe what you observed and check it by asking, rather than concluding what a person is hiding or feeling. The meta-analytic mean is above chance, so the claim is that demeanour-based inference is weak, not that nothing can be detected.
What it does not show
- Mean across studies; individual judges and situations vary.
- Studies of real-time judgments with no aids or training; not courtship settings.
- Concerns lie versus truth, not the many other things nonverbal cues may convey.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Source
- Bond, C. F., DePaulo, B. M. (2006). Accuracy of Deception Judgments, Personality and Social Psychology Review.
Evidence: Supported with caveatsLie detection from behaviour is poor mainly because the behavioural cues to lying are weak, not because people attend to the wrong cues; for example, eye contact is only weakly related to being judged deceptive.
Why this label
Several studies agree, with stated limits such as student samples or lab settings. Warns against 'tells' lists (gaze aversion, fidgeting). Observing behaviour is still useful for what a person does, but the same behaviour has many causes, so it prompts a question rather than a verdict.
What it does not show
- One research team's lens-model synthesis of studies of deception judgments; the interpretation is theirs.
- Meta-analytic correlations are averages and do not describe any one person.
- Concerns deception cues; other cue-to-state links are not covered.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Source
- Hartwig, M., Bond, C. F. (2011). Why do lie-catchers fail? A lens model meta-analysis of human lie judgments, Psychological Bulletin.
Evidence: Supported with caveatsSeeing a person's face and body is not a better guide to truthfulness than hearing them and can mislead: in a meta-analysis, lies were detected more accurately when heard than when seen, and in one experiment deceivers were rated more credible when observers could both see and hear them.
Why this label
Several studies agree, with stated limits such as student samples or lab settings. Contradicts the intuition that more visual information means better reading. A confident, composed demeanour is read as credible whether or not it signals honesty, which is why lessons separate 'seems at ease' from 'is telling the truth'.
What it does not show
- Burgoon et al. is a single mock-theft experiment; the meta-analytic audible-over-visible result is an average across mixed designs.
- The meta-analysis abstract compares audible with visible lies; it does not show that adding vision to sound lowers accuracy in general.
- Interrogation-style settings with deception, not casual conversation.
- Does not show that visual cues are useless for non-deception judgments.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Sources
- Bond, C. F., DePaulo, B. M. (2006). Accuracy of Deception Judgments, Personality and Social Psychology Review.
- Burgoon, J. K., Blair, J. P., Strom, R. E. (2008). Cognitive Biases and Nonverbal Cue Availability in Detecting Deception, Human Communication Research.
Evidence: Supported with caveatsFacial movements carry some information about emotion above chance, but the same expression can mean different things across people, cultures and situations, so a face alone does not reliably show what someone feels.
Why this label
Several studies agree, with stated limits such as student samples or lab settings. Supports teaching 'notice the expression, hold the interpretation lightly'. Also marks the limit of the sceptical stance: recognition across cultures is above chance and better within one's own cultural group.
What it does not show
- Contested field: researchers disagree on how strongly universality should be claimed; the 2019 review is one team's critique.
- Elfenbein et al. pooled studies using posed, spontaneous and imitated expressions, and cross-cultural accuracy varied with study design; neither source describes everyday life.
- Population-level results; cannot say how readable any one person is.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Sources
- Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M., Pollak, S. D. (2019). Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements, Psychological Science in the Public Interest.
- Elfenbein, H. A., Ambady, N. (2002). On the universality and cultural specificity of emotion recognition: A meta-analysis, Psychological Bulletin.
Evidence: PlausibleMicroexpressions are not a dependable lie detector: in one lab study they appeared in about a fifth of participants and 2% of expressions (longer inconsistent expressions were more common), and a micro-expression training tool did not improve lie detection in one randomised test.
Why this label
One study, or indirect evidence from a nearby question. Counters the popular claim that a trained eye can catch fleeting expressions. Each finding is a single study, so lessons should say 'little evidence' rather than 'proven false'.
What it does not show
- Porter and ten Brinke used posed or simulated expressions in a lab; Jordan et al. tested one training tool (90 participants).
- Neither study speaks to romantic contexts.
- Other training evaluations may exist and were not reviewed here.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Sources
- Porter, S., ten Brinke, L. (2008). Reading Between the Lies: Identifying Concealed and Falsified Emotions in Universal Facial Expressions, Psychological Science.
- Jordan, S., Brimbal, L., Wallace, D. B., Kassin, S. M., Hartwig, M., Street, C. N. H. (2019). A test of the micro-expressions training tool: Does it improve lie detection?, Journal of Investigative Psychology and Offender Profiling.
Evidence: PlausibleStrangers watching short speed-dating clips could judge above chance who was interested in one study (near chance for female interest), accuracy varied a lot between the people watched, and a later study did not find it reliably.
Why this label
One study, or indirect evidence from a nearby question. Observers beat chance with clips as short as 10 seconds and were better for male than female interest. A second study is reported, from a search-result summary only, to show similar accuracy in three countries. Later, three studies with similar videos found observers could not reliably detect attraction, except when both daters were interested.
What it does not show
- Place 2009 and 2012 use the same stimulus videos, so they are not independent
- Observers, not the actual conversation partners, judged interest
- Speed dates are a short, structured setting
- A 2022 study did not reproduce the effect, so the record is mixed
- Abstract of Place 2012 seen only as summarised by a search result
Who it applies to
A pattern across groups of people, not a statement about any one person.
Sources
- Place, S. S., Todd, P. M., Penke, L., Asendorpf, J. B. (2009). The Ability to Judge the Romantic Interest of Others, Psychological Science.
- Place, S. S., Todd, P. M., Zhuang, J., Penke, L., Asendorpf, J. B. (2012). Judging romantic interest of others from thin slices is a cross-cultural ability, Evolution and Human Behavior.
- Samara, I., Roth, T. S., Nikolic, M., Prochazkova, E., Kret, M. E. (2022). Can third-party observers detect attraction in others based on subtle nonverbal cues?, Current Psychology.
Evidence: PlausiblePeople overrate how well they read others, and the least accurate overestimate most; self-rated mind-reading ability tracked actual accuracy only weakly or not at all.
Why this label
One study, or indirect evidence from a nearby question. Supports calibration drills with feedback instead of trusting a feeling of insight. Plausible rather than supported because it rests on one paper (two studies); the authors state that earlier work also found a weak or non-existent link, which was not checked here.
What it does not show
- Two studies: a video-based accuracy test and a negotiation exercise with MBA students (Study 2: 164).
- Self-estimates were relative-ability (percentile) judgments; not tested in dating or flirting.
- Narcissism result is correlational.
Who it applies to
A pattern across groups of people, not a statement about any one person.
Source
- Ames, D. R., Kammrath, L. K. (2004). Mind-Reading and Metacognition: Narcissism, not Actual Competence, Predicts Self-Estimated Ability, Journal of Nonverbal Behavior.
More practice
The rest of the set, Faces, certainty and what you can tell: six moments: 6 items, about 10 minutes. Decide what the evidence supports, and how well. Open the exercise set.
Go further
Outside talks, videos and pages on the topics this lesson cites. Nothing from them loads until you follow a link, and a link is not an endorsement: read the note under each one.
Jeffrey Hall (University of Kansas): Flirting hard to detect, researcher says
University summary of Hall's study of first conversations: partners were far less accurate at detecting flirting than at noticing its absence, and observers were no better.
Keep in mind. News release; student samples and one lab setting.
Vanessa Van Edwards: 28 Science-Backed Body Language Signs of Attraction
Catalogue of cues said to signal attraction (leaning in, mirroring, proximity, eye contact), with the useful advice to look for clusters of signals rather than one gesture.
Keep in mind. Mixes research with popular lists; the gendered framing and some involuntary-signal claims (pupils, blushing) are weakly sourced. Cues never authorise touch.
Vanessa Van Edwards: Cues by Vanessa Van Edwards: 9780593332191 | PenguinRandomHouse.com: Books
Book on reading and sending nonverbal cues for warmth and competence; the publisher page gives the official description and edition details.
Keep in mind. Self-help book; evidence quality varies by chapter. Link to the publisher only.