Darwin, Deception, and Facial Expressions

Paul Ekman has long been fascinated by Charles Darwin’s theories on emotions. He recommends Darwin’s book, The Expression of the Emotions in Man and Animals (first published in 1872), which introduces the concept of universal facial expressions, as “the one psychology book you should read if you only have time for one.” In 1998, this book was reissued with Ekman’s commentary, analyzing Darwin’s ideas from a modern perspective. The article “Darwin, Deception, and Facial Expressions” serves as a continuation of these commentaries.

While Darwin did not specifically study facial expressions as a means of detecting lies, he recognized the potential for such analysis. In his conclusion, Darwin wrote, “They [facial expressions] reveal the thoughts and intentions of others more truly than do words, which may be falsified.” He also proposed two ideas directly relevant to the principles of lie detection, which Ekman explores in this article.

The “Face-Body Leakage” Hypothesis

The first idea, which Ekman calls the “face-body leakage” hypothesis, is that people, according to Darwin, can control their body movements better than their facial expressions. Therefore, signs of deviation from a baseline—so-called “leakages” that may indicate deception—are more visible on the face than on the body.

At first glance, Darwin’s hypothesis that the face gives more signals than the body seems logical. After all, it’s easier to control hand and leg movements than facial expressions. However, Ekman notes that most people do not pay attention to controlling their body movements. If this is the case, the body becomes a rich source of “leakages” or “deception cues,” which contradicts Darwin’s assumption.

Experimental data revealed a more complex picture. In one study, participants watched videos of women who were either lying or telling the truth about positive emotions experienced while watching a nature film. Half of the women had actually watched the nature film, while the other half had watched a graphic surgery video and had to lie about their feelings. Observers were shown either just the women’s faces or just their bodies (with sound off) and asked to judge whether the behavior was trustworthy. They were also shown baseline examples of each subject. Lie detection was more accurate when observers watched the body rather than the face. When asked which aspects of their behavior they tried to control while lying, almost all the women mentioned their faces, and very few mentioned their body movements.

Still, in some cases, the face remains a much more important source of information than the body—specifically, in the case of microexpressions. Microexpressions are facial expressions so brief that they are difficult for the untrained eye to detect. Ekman’s team first discovered microexpressions while reviewing videos of psychiatric patients who were lying about their condition during clinical interviews. Microexpressions can reveal a person’s true emotions.

Trained experts in emotions and microexpressions were shown the facial videos from the experiment described above. Their accuracy in detecting truthfulness was 80% or higher. These experts based their judgments on microexpressions and inconsistencies in emotional expressions (Ekman emphasizes the importance of the natural dynamics of facial reactions).

After expert analysis, the videos were further evaluated: facial expressions were coded using the Facial Action Coding System (FACS), the frequency of illustrative gestures was measured, and voice pitch was analyzed. When participants tried to hide negative emotions (fear, sadness, disgust), they masked them with a social smile, involving only the zygomatic major muscle. When telling the truth about positive emotions, they displayed a genuine (Duchenne) smile, involving both the zygomatic major and the orbicularis oculi muscles. Only a quarter of subjects showed microexpressions. The number of illustrative gestures decreased during lying, but not significantly. A third of the subjects displayed a shoulder-shrug gesture indicating uncertainty. Voice pitch increased when lying.

Summing up, Ekman writes that Darwin was correct in suggesting that body movements can be controlled, but most people do not do so, making the body a valuable source of cues for detecting deception. Ekman concludes that multiple parameters—gestures, voice pitch, and more—should be analyzed together.

The “Inhibition Hypothesis”

Ekman refers to Darwin’s second idea as the “inhibition hypothesis.” Darwin suggested that it is very difficult to suppress involuntary muscle contractions, especially when it is hard to voluntarily produce those contractions. In other words, if we cannot voluntarily produce a certain facial expression, we will also have difficulty suppressing it when it occurs spontaneously.

From a neurophysiological perspective, facial muscle activity works as follows: the facial nerve nucleus, which controls facial muscle contractions, receives input from many brain regions. The motor cortex sends impulses for voluntary movements, while subcortical structures send impulses for involuntary movements. In some neurological disorders affecting the pathways from the cortex, a person may be unable to smile on command but can smile spontaneously at a joke. The opposite occurs with subcortical damage.

The exact mechanism for suppressing facial expressions is not fully understood, but to analyze the inhibition hypothesis, Ekman used facial expressions that are difficult to produce on command. Fewer than 25% of people can perform these expressions voluntarily. Ekman calls these signals “reliable” and notes that repeated analysis of facial expressions shows that these signals are indeed not inhibited. Although this hypothesis has never been quantitatively tested, repeated empirical observations support Darwin’s view.

Building on Darwin’s ideas, Ekman discusses criteria for distinguishing between facial expressions produced voluntarily and those triggered by emotion. As an example, he cites the Duchenne smile, first described by French neurologist and psychiatrist Guillaume Duchenne. The Duchenne smile involves both the zygomatic major and orbicularis oculi muscles (see illustration). Most people cannot produce this smile on command, and those who can often cannot do so symmetrically on both sides of the face, although they can maintain the contraction once it starts. The key sign of a fake Duchenne smile is an uneven, asymmetrical onset.

Darwin described the Duchenne smile in his book and noted the connection between smile analysis and lie detection, a point Ekman emphasizes. He points out that many lie detection experiments fail to distinguish between genuine (Duchenne) and social smiles—if only the zygomatic major muscle is considered, it is impossible to tell whether a smile is masking negative emotions or simply expressing joy.

Seven Criteria for Distinguishing Voluntary and Involuntary Facial Expressions

1. Morphology – The presence of “reliable” signals increases the likelihood that the expression is involuntary.
2. Symmetry – Asymmetrical expressions may indicate conscious control.
3. Duration – Very short (<0.5 seconds) or very long (>5 seconds) expressions are more likely voluntary (microexpressions are an exception).
4. Onset Speed – Voluntary expressions begin more quickly and abruptly.
5. Apex Overlap – Overlapping peaks (apexes) of different parts of the expression are characteristic of spontaneous expressions.
6. Ballistic Trajectory – The absence of “jerky” movements in the expression indicates spontaneity.
7. Congruence – Spontaneous expressions match not only the meaning but also the timing of speech.

In conclusion, Ekman writes that there are no behavioral signals that can be fully trusted or completely distrusted—there are no absolute signs of truth or deception. It is necessary to consider everything: voice, facial expressions, microexpressions, gestures, and posture.

Ekman quotes Darwin, who wrote about how easily words can be falsified, highlighting the need to analyze speech for pauses, changes in emphasis, speech errors, indirect or distancing language, slips of the tongue, and logical inconsistencies.