Developing a system for measuring movements of the human face
FAST vs FACS : The need for a more comprehensive facial coding system
In my cross-cultural studies researching display rules, I developed and published a tool for measuring facial movements, the Facial Affect Scoring Technique (FAST). A year after it was published, an anthropologist, Wade Seaford, showed me a facial movement on his own face that was not included in FAST. I was devastated by Wade’s demonstration, not knowing how much else, how many other facial movements we had missed in creating FAST. The only way to have a comprehensive system, to include everything the face can and sometimes does do, would be to base it on how the muscles worked to change facial appearance, the anatomy of facial movement. And that is what we did, albeit reluctantly.
I had no choice. I knew that if I was to develop a comprehensive system of facial measurement I would have to learn facial anatomy. I ended up creating a functional anatomy, illustrating in text, photographs, and film, how muscular actions change facial appearance. A topic largely heretofore ignored because there is no medical intervention that required it.
I was cautioned that the endeavor to catalog all facial movements would be too complicated. I was cautioned I would get lost and very likely fail. Wally Friesen was also reluctant, but once I started, Wally became thoroughly engaged and was a true partner in developing the facial action coding system.
The motivation behind creating FACS
My motivation for taking on this daunting task was twofold: I knew we would need a very precise and sensitive facial measurement tool to spot clues to deceit in the face, and I wanted to make it possible for any scientist to extract the information Silvan Tomkins was able to see in expressions. Also, I had to provide a substitute for FAST, since I now knew it was incomplete. Just how incomplete I could only discover by developing a truly comprehensive measurement system based on the anatomy of facial movement. I had a five year grant from NIMH to continue my study of depressive patients, and I thought I could justify spending some of that time creating a facial measurement tool, which could then be used to evaluate the depressive patients. I didn’t realize when I started that it would take more than five years to complete, and that I would run out of grant money before I finished.
The anatomy of facial expressions
The standard anatomy textbooks were not of much help. They showed where muscle strands were, but not how the muscles generated changes in facial appearance, the expressions. It was a dead anatomy, constructed from dissections of dead people. We needed a functional anatomy, but it didn’t exist, so we had to create it.
French neurologist Duchenne de Boulogne’s book on the mechanics of facial movement was of great use. It had been published in 1862, never translated from the French, and long out of print. Darwin included some of Duchenne’s photographs and findings in his own book. I was to later find some of the correspondence between the two of them. Duchenne worked with some patients who had no pain sensations in their face, so it did not hurt when he electrically stimulated muscle areas on their face, photographing the resulting movement that occurred. So, for example, he put the electrode on the bony area high in the cheekbone, which generated a smiling appearance. This proved that the muscle extending from the zygomatic arch to the lip corners was responsible for a smiling appearance. By good luck, one of the very few surviving copies of Duchenne’s book was in the rare book section of my university library. Duchenne was helpful on the single muscular contractions and the changes in appearance when there were contractions of two muscles simultaneously, but he did not explore all of the two muscle contractions, nor how the simultaneous action of three or more muscles would change appearance, except for one or two exceptions. There were some other limitations in his work, but it was of great help.
Duchenne’s work was instrumental in distinguishing the differences between a felt enjoyment smile and other types of false or social smiles. Duchenne’s observations about smiling have broader implications for distinguishing voluntary from involuntary expressions. Some muscular action is difficult to perform voluntarily, its absence is a mark that the facial expression is not generated by an emotion. Such difficult to perform actions mark what I have called reliable facial expression; in Emotions Revealed I have described the reliable sign for fear, anger, surprise, and sadness.
Building a catalog of facial expression
Building upon Duchenne’s descriptions for more than a year, Wally and I sat in front of video and still camera as we tried to make each muscle contract, one by one. It wasn’t hard for me to do, for I have always had exceptional control of my facial muscles. My mother often told me not to make all the crazy facial expressions with which I would try to entertain her, warning me they might freeze on my face. She didn’t live long enough to find out how that ability helped me in one of my most useful scientific studies.
Defining Action Units
For every muscle contraction, we wrote a precise description of how it changed facial appearance, highlighting what the video recordings depicted. We identified a little more than 40 action units, or AUs, and then examined what occurred when two of the AUs occurred simultaneously. We cataloged more than 300 AU two-way combinations, and then took on all the combinations of three AUs, and so on, until we stopped at combinations of six AUs.
There were a few cases in which we could not be certain which muscle we were voluntarily contracting, so I had a needle inserted into my face to electrically stimulate one or another muscle. It was painful, but we didn’t have to do it very often.
Similar studies corroborating our findings
When we were about halfway through cataloging these various facial combinations, I heard that there was a Swedish anatomist who had generated a functional anatomy of facial expression. I visited professor Hjorstjo in the anatomy department of the university of Lund, and found that he had indeed just published a book based on using the same method we had adopted, voluntarily contracting single muscles in some combinations and photographing the changes. His book, which was in English, used drawings of the face, based on his photographs. While useful in corroborating our findings to date, it was incomplete and not intended to provide a measurement tool. By the end of four years, we had completed what we intended to be a self-instructional manual, including photographs and filmed examples of each AU and many of the combinations.
The publication and application of FACS
FACS, the Facial Action Coding System, was published in 1978, and thousands of scientists and graduate students have used FACS in their research. A volume has been published reprinting some of their published scientific articles using FACS. Animation studios, such as Pixar, Disney, and Industrial Light and Magic, have also studied FACS to assist in depicting realistic human expression.
To accomplish a complete FACS measurement of one minute of facial behavior (which means identifying each action unit, exactly when it began to appear, when it reached its apex, how long the apex was held on the face, when it began to decline, and when it disappeared) is very slow, precise work. Rarely does just one AU appear. Instead, three to five AUs may appear in overlapping time, creating the impression of an “expression”. Usually it takes 50 to 60 minutes to score one minute. Learning FACS is also a slow process requiring 75 to 100 hours, and then a Final Test, which we provide to determine if you have learned it accurately. Nevertheless, in the decades since FACS was published in 1978, hundreds of articles by various scientists have been published. A selection of them appear in a volume I co-edited with Erika Rosenberg, What the Face Reveals, currently in its third edition.