What a face does not say

Charles Darwin published in 1872 a work on the expression of the emotions in man and animals, in which he argues that certain expressions are inherited. The thesis lay dormant for nearly a century, then revived at the end of the 1960s with the work of Paul Ekman and Wallace Friesen, notably observations conducted in Papua New Guinea among people little exposed to Western images.

From this work came the list of the six so-called basic emotions: anger, disgust, fear, joy, sadness, surprise. It has irrigated textbooks, training courses and, later, software.

The classic experiments present a face and offer a list of words. The format makes the task enormously easier, since it is enough to eliminate. When that choice is replaced by a free response, recognition rates fall, including in Western populations.

Work conducted in 2014 among the Himba of Namibia replaced the labels with a free sorting of photographs. The groupings obtained did not correspond to Western categories and followed the action depicted rather than the supposed emotion.

A series of experiments conducted by Hillel Aviezer showed that the same face changes perceived emotion depending on the body it is placed on. The most striking dates from 2012: photographs of tennis players taken at the moment of the decisive point, victory or defeat, are indistinguishable when only the face is shown. It is the bodily cues that allow a decision.

In other words, at maximum intensity, a face does not even say the sign of what is happening. This result contradicts the most solid intuition we have about reading others.

Lisa Feldman Barrett and four co-authors reviewed several hundred studies and concluded that there is a double variability: the same emotion is expressed through diverse facial configurations, and the same configuration accompanies different states. The correspondence exists, but it is too loose to ground an inference about a particular person.

Defenders of the classical position reply that nobody ever maintained that an emotion was always expressed the same way, and that the review targets a caricatured thesis. The disagreement therefore bears as much on what had been asserted as on the data, which explains why it lasts.

The usable conclusion is not that nothing can be seen on a face. It is that a face signals the existence of something, rarely its nature, and never independently of the context, the body and what one already knows about the person.

The best readers do not guess better than others: they integrate more information and, above all, they check. Perceiving a signal and interpreting it correctly are two distinct operations, and the second often requires asking a question.