Facial action coding

From How Emotions Are Made
Jump to: navigation, search

Chapter 1 endnote 11, from How Emotions are Made: The Secret Life of the Brain by Lisa Feldman Barrett.
Some context is:

So scientists also employ an alternative technique called facial action coding (FACS), in which trained observers laboriously classify a subject’s individual facial movements as they occur.

A facial movement, called a facial action, can be one muscle contracting or several different muscles contracting together.

In 1969, the Swedish anatomist Carl-Herman Hjortsjö developed a formal coding system for facial movements that later became “the Facial Action Coding System” (FACS) by Paul Ekman and his colleague Wallace Friesen in 1978.[1][2][3][4] Scientists have created FACS for babies,[5][6][7] chimpanzees, macaque monkeys, and even dogs.

FACS coders go through rigorous training to learn to identify individual facial action units when they are active, and they must meet certain standards of reliability before they are certified. According to Ekman and colleagues, particular combinations of facial action units are supposed to correspond to emotion-specific expressions, where each expression is created by contracting a specific set of facial muscles.

The one-to-one correspondence between particular facial actions and facial expressions is weakened by findings that different combinations of action units can produce similar looking expressions.[8] This is an example of degeneracy.


Another version of FACS, called “Emotion FACS” or EMFACS, produces results supporting the classical view.[9] There is no special training for EMFACS; anyone who is trained to use FACS can selectively apply the coding criteria to use EMFACS procedures. EMFACS coders decide whether a group of facial action units are active en masse (i.e., they identify the presence or absence of the putative expression), rather than detecting each action one at a time. (They encourage coders to engage in what scientists call "configural perception.") With the precision of FACS lost, EMFACS is less reliable[10] and potentially more prone to bias. It is essentially a fancy version of the basic emotion method.

Not surprisingly, when compared to studies using FACS coding, studies that use EMFACS are more likely to find support for the classical view. For example, of the hundreds of published studies using FACS and EMFACS to code facial muscle movements during emotion, a scholarly review published in 2008 listed only twenty-five that reported test subjects making spontaneous facial movements matching the configurations in the posed photos (23 of which used some version of FACS coding).[11] A close examination of those studies reveals an interesting pattern: Of the fourteen studies using FACS coding alone, only five found clear evidence for the classical view (that during episodes of emotion, test subjects spontaneously moved their faces in ways that matched the expected expressions); six studies did not find supporting evidence (and found, instead, that facial actions distinguished happiness), while three studies were not designed to test the question in the first place because the test subjects were only presented with pleasant stimuli like jokes and cartoons. Of the nine studies using EMFACS coding, all found some evidence supporting the classical view.

FACS and infants

Earlier in chapter 1, I note that FACS-coded adult faces do not match the posed photos of the basic emotion method. The same is true for infants. Noted developmental psychologist Harriet Oster explains:

"... the problem with Max and Affex "formulas" is not only that there's no evidence that infants differentially produce the Max-specified expressions of discrete negative emotions in response to predicted elicitors, but that expressions that fit the Max formulas differ from their presumed prototypes in terms of their underlying muscle actions. A Max anger expression is a cry face — same in adults as in infants. It lacks muscle actions seen in adult anger and involves actions seen in adult grief and express expressions. A crying infant (or adult) doesn't look angry."[12]

Notes on the Notes

  1. Hjortsjö, Carl-Herman. 1969. Man's face and mimic language. Sweden: Studentlitteratur Lund.
  2. Ekman, Paul, and Wallace V. Friesen. 1978. Facial Action Coding System. Palo Alto: Consulting Psychologist Press.
  3. Ekman, Paul, Wallace V. Friesen, and Phoebe Ellsworth. 1972. Emotion in the Human Face. New York: Pergamon.
  4. Ekman, Paul, Wallace V. Friesen, and Silvan S. Tomkins. 1972. "Facial affect scoring technique (FAST): A first validity study." Semiotica 3 (1): 37-58.
  5. Oster, Harriet, and Diana Rosenstein. 1996. "Baby FACS: Analyzing facial movement in infants." Unpublished manuscript, New York University.
  6. Also, see Izard, Carroll Ellis, מיה ויס, and Maya Weiss. 1979. The Maximally Discriminative Facial Movement Coding System. Newark, DE: University of Delaware, Instructional Resource Center.
  7. Izard, Carroll Ellis, Linda M. Dougherty, and Elizabeth Ann Hembree. 1983. A system for identifying affect expressions by holistic judgments. Unpublished manuscript, University of Delaware. 
  8. Tian, Ying-li, Takeo Kanade, and Jeffrey F. Cohn. 2001. "Recognizing Action Units for Facial Expression Analysis." IEEE Transactions on Pattern Analysis and Machine Intelligence 23(2): 1-19.
  9. Ekman, P., and W. V. Friesen. 1982. "Rationale and reliability for EMFACS coders." Unpublished. 
  10. Rosenberg, Erika. Frequently Asked Questions. erikarosenberg.com/faq.
  11. Matsumoto, David, Dacher Keltner, Michelle N. Shiota, Maureen O’Sullivan, and Mark Frank. 2008. "Facial expressions of emotion." In Handbook of Emotions, 3rd edition, edited by Michael Lewis, Jeannette M. Haviland-Jones, and Lisa Feldman Barrett, 211-234. New York: Guilford Press.
  12. Oster, Harriet, personal communication, February 22, 2013.