Behaviour Analysis: What is Facial Action Coding System (FACS)?
This year I gained certifications in Facial Action Coding System (FACS) and Six Channel Analysis System (SCAnS™), two scientific systems that can be useful in analysing human behaviour. I’m currently working on gaining CatFACS certification. If you’re wondering whether CatFACS is a coding system for cat behaviour, you guessed right, that’s exactly what it is! This article focuses on explaining FACS, while separate articles focus on CatFACS and SCAnS™.
What is Facial Action Coding System (FACS)?
Facial Action Coding System is a method for scientifically measuring Action Units (AUs), unique anatomically based facial movements involving one or more muscles. It’s considered the gold standard for facial movement analysis. Action units create distinct changes in the appearance of facial areas and work together (or on their own) to create facial expressions. Facial Action Coding System does not measure expressions. Instead, it focuses on the presence and intensity of independent action units.
Primarily used as a research tool for scientific investigation, FACS has led to fascinating research and discoveries. Research findings include facial movement patterns connected to emotion, deception, and medical and psychological disorders, for example, differences between suicidal and non-suicidal depressed patients.
Behaviour analysts look for the presence (or lack of) action units when investigating behaviour due to their connection to emotion. We also look for specific combinations of action units that form facial expressions of emotion to guide understanding of an individual’s mental state. During deception detection analysis, it’s crucial to know whether emotional expression is aligned with verbal communication because a mismatch can be a red flag. However, it’s important to remember that FACS does not measure expression or ascribe emotional interpretation. It has though, contributed to our understanding of the relationship between specific action units (and action unit combinations) and emotion.
Facial Action Coding System (FACS) Development
Previous approaches focused on measuring underlying emotion; however, Ekman and Friesen considered this approach subjective and open to bias. That is, despite their proposal that specific action unit combinations represent emotional expression. Instead, they focused on the objective measurement of frequency and intensity of action units, the smallest constituent component of expression. This allows for research of facial movements unrelated to emotion, such as conversation punctuators, pain and depression.
In the 1970s, Ekman and Friesen developed FACS, building on the work of Swedish anatomist, Hjortsjö. Facial Action Coding System was substantially updated in 2002. The development of FACS created a standardised system with a common descriptive language, providing a more objective methodology for measuring facial movement. Its use is wide and varied and includes fields such as consumer research, behavioural science, computer graphics, animation, forensic science, neuroscience and social and clinical research.
Facial Action Coding System has been modified into BabyFACS to facilitate objective research of infants since facial morphology differs between adults and infants. If you’re an animal lover like me, you’ll be excited to know that FACS has been modified into eight (to date) animal-based research systems, including ChimpFACS, OrangFACS, EquiFACS, CatFACS and DogFACS. These systems allow for comparison across species in communicative and emotional contexts, offering valuable insight into the differences and similarities between humans and animals, addressing interesting questions, such as, ‘What makes us uniquely human?’.
How Facial Action Coding System (FACS) Works
Trained FACS coders identify and measure action units, occurring independently or in combinations, usually for scientific research. They study human faces in still images (with a neutral comparison) or on videos. Identified action units are rated on a five-point intensity scale from A to E (trace to maximum). The lower face is coded first (16 potential action units), then the upper face (14 potential action units). There are also miscellaneous action units and action descriptors to look for, as well as head (8) and eye position (6) action units, which are sometimes important in research.
Although FACS is not concerned with emotion, researchers can infer emotion in a second system, for example, the Emotion Facial Action Coding System (EMFACS) or Facial Action Coding System Affect Interpretation Dictionary (FACSAID), measuring action units and associated emotional expression.
Studying Facial Action Coding System (FACS)
The FACS training is available to purchase online from The Ekman Group’s website. The training is self-instructional, including a manual, videos and score checker. It is encouraged to study with others to compare facial movements and scoring. I studied alone, so it is possible. That said, I had a fairly robust prior understanding of the action units that are associated with emotion. I prefer to work alone, but, at times, I longed to compare and contrast action units within a group.
Part of the problem with studying FACS is the manual’s volume, academic style, terminology, repetition and complex instruction. These are not conducive to learning. Coming from an education and training background, I felt somewhat frustrated by this. It was incredibly tedious and, at times, I did wonder whether I would be able to pass the test or not. Perhaps the tediousness wouldn’t be as intense in a group setting–dependent on group dynamics.
Due to the tedious training and life/work commitments, I studied FACS on and off for several years. I had a couple of acquaintances who were also studying FACS on and off; however, our periods didn’t coincide, despite hopes to coordinate. Endorsed by The Ekman Group, Erika Rosenberg runs in-person and online FACS training. I planned to perhaps take this route if I failed, but it wasn’t necessary.
The complex system requires a significant time investment to learn, with approximately 100 hours of self-study and practice; I’m certain I invested significantly more hours. Therefore, training can take many months to complete and gain certification via the FACS test.
I moved through the manual, celebrating the end of each chapter. Since taking the FACS test and gaining certification, it makes more sense why it was designed that way. For it to work as a scientific method of measurement it needs to be robust. I wonder how many people start but don’t finish–I was almost one of them.
What I struggled with most was the intensity scoring. The manual describes intensities as trace, slight, marked, pronounced, severe, extreme and maximum. I especially struggled with marked and pronounced–there’s a fine line difference that can easily be influenced by mood. Ekman himself acknowledges that perceptions can change when coding, depending on emotional state or mood. Furthermore, individuals assign weighted meanings to words, which changes how we perceive them. I found myself asking my Husband several times, “Would you describe this movement as marked or pronounced?”, having changed my opinion since the previous day.
The Facial Action Coding System (FACS) Test
The test consists of 34 short videos displaying individuals (head and shoulders) speaking or listening. You score the action units present, as well as their intensity. To pass, you need to gain a reliability score of .70 or higher. This can be tricky as the video quality isn’t clear. That said, out-of-test behaviour analysis is often based on unclear footage. For example, blurred police body cam footage is sometimes all you have to work with. You save each video score as you move through the test. It’s recommended to check all answers before the final submission. I certainly recommend this because I changed all of my answers during the checking stage.
I don’t know how long you have to complete the test. I suspect you have it for as long as it takes because there was an expiry date showing, if I recall correctly, 1976! Obviously, that wasn’t correct since it would have expired when I was three years old. I figured it would have been correct if it was important and if it did actually expire. Don’t take my word for it though, it’s an assumption.
I also don’t know what happens if you fail the test–whether you have to pay again and get a new set of videos, or whether you keep adjusting your scores. These were all questions I searched for answers to, convinced I would fail. I couldn’t find answers and my enquiry to The Ekman Group’s website directed me to ask Erika Rosenberg future questions (without answering my current question). I thought this was odd and didn’t want to bother Erika, so I settled for unanswered questions.
I contacted Erika after passing the test as I had expected to receive EMFACS (EmotionFACS) instructions and an update to the manual. Erika kindly supplied the EMFACS instructions on seeing my certificate and explained that there was no update to the manual and that what I had read about an update was outdated information. I couldn’t actually recall where I had read that.
Drawbacks of Facial Action Coding System (FACS)
While FACS is a comprehensive system for measuring visible facial movement, it omits subtle changes in muscle tonus and other facial phenomena such as tears, sweating and blushing. Plus, focusing on the face provides data from only one communication channel.
Using FACS alongside other methodologies could provide more valuable data and insight; for example, the Six Channel Analysis (SCAnS) methodology identifies data from six channels of communication: voice, linguistic content, interactional style, facial movements, body movements and psychophysiology.
In addition to the significant time investment for training, FACS coding is time-consuming and can be prone to human error, bias and subjectivity. Typically, it takes over two hours to code one minute of video, rendering it unsuitable in settings requiring real-time analysis, for example, clinical settings where detecting action units associated with pain can improve the quality of care. However, all thorough behaviour analysis, no matter which system is used, is time-consuming. That is why I stipulate, on my behaviour analysis services pages, that it takes approximately one hour of analysis per one minute of video footage.
Automating Facial Action Coding System (FACS)
You might be asking, ‘Why don’t we automate FACS?’. Automating FACS would remove the human requirement—the need to invest time in training and coding, reliability and subjectivity. However, computer scientist and digital activist Joy Buolamwini posits that current artificial intelligence (AI) systems used to measure the face are flawed due to the embedding of unconscious human bias during development.
Datasets used to train AI contained more images and videos of males and whites, leading to algorithms that perform better on male faces than female faces and significantly better on lighter-skinned faces, with skewed datasets producing skewed results. The human element of bias remains, with technology highly susceptible to bias.
Wrapping Up Facial Action Coding System (FACS)
Despite being the gold standard of facial movement measurement, offering a comprehensive, unobtrusive and more objective methodology, problems with FACS exist, including time-intensive training and coding, reliability, subjectivity and bias.
While automating FACS may be some time away, manually applied FACS has proved a valuable research tool for over four decades. We can look forward to human coders assisting in research to uncover many more exciting discoveries.
If you’re still reading, you’re possibly interested in becoming a FACS coder. I hope I haven’t discouraged you with my honest experience. I’m thrilled I passed FACS certification and I’d do it all again if I needed to. I encourage you to persist in attaining certification too. Let me know how you go and whether you need any help. If you can answer any of the questions that for me remained unanswered, please leave a comment.
