Getting Your Head Around the Action Coding System Manual
When I first sat down with the Action Coding System Manual Ekman and tried to actually code a video, I had no idea how much of a gap existed between reading the descriptions and making the call on screen. The manual itself is dense, methodical, and occasionally frustrating — not because it lacks clarity, but because the subject matter demands you hold dozens of categories in your head simultaneously. Let me walk through what this system actually is, how to use it, and where people routinely trip up.
What the Action Coding System Actually Codes
The Action Coding System, developed by Paul Ekman and Wallace Friesen at the University of California, San Francisco, is designed to categorize and quantify human behavior across multiple channels. The manual covers several coding systems — the Facial Action Coding System (FACS) for facial movements, the Body Action and Movement Coding System (BAMCS), and various supplementary systems for vocalics, proxemics, and other nonverbal channels. FACS itself is the most widely used component. It breaks down every observable facial movement into discrete "Action Units" — specific muscle contractions. There are over 40 individual action units, and they combine in ways that produce recognizable expressions like happiness, sadness, anger, fear, surprise, and disgust. But the system doesn't just label emotions. It records what physically moves on the face, leaving interpretation largely to whoever is reading the code.
How the Manual Is Structured
The Action Coding System Manual Ekman spans multiple volumes depending on which subsystem you're working with. The core FACS manual contains detailed photographs and diagrams for every action unit, showing the onset, apex, and offset phases. Each unit gets a page or two of description, along with variations across different skin tones and face shapes. The manual also includes extensive coding instructions — how to handle partial movements, asymmetrical expressions, rapid movements that blur between units, and expressions that change within a single second. There are sections on inter-coder reliability, training protocols, and the specific criteria you use to distinguish similar-looking units from one another. Below the main text, you'll find coding forms, reference cards you can clip out and keep at your desk, and appendices with common coding dilemmas and how experienced coders resolved them.
Get the Full Details
Running Through an Actual Coding Session
Here's what happens when you actually sit down with footage and a stopwatch. You load a clip — maybe ten to thirty seconds of a person talking, reacting, or sitting still. You pause, note the timestamp, and start watching for movements that correspond to action units. Each detected unit gets logged with its onset time, peak time, and offset time. The manual instructs you to code in real time or frame by frame depending on the research question. For microexpressions — movements lasting a fraction of a second — you'll need frame-by-frame analysis. Most standard expressions can be captured in real time with practice. I spent about six weeks training before I felt confident coding basic expressions independently. That involved watching hundreds of example clips, coding alongside experienced trainers, and comparing my codes against theirs until the disagreement rate dropped below acceptable thresholds. The manual recommends this kind of supervised practice, though the exact timeframe depends on how much time you can commit daily.
Where People Get Stuck
The most common problem I encounter involves distinguishing between Action Unit 1 (inner brow raiser) and Action Unit 4 (brow lowerer) when they occur close together. These two units can create a similar visual effect — the brows move upward in AU1 and then immediately downward in AU4, sometimes within a half-second window. When that happens fast enough, your eye sees one continuous brow movement instead of two distinct ones. The workaround is to slow the footage to 0.25x speed and watch for the momentary pause or reversal between the two movements. The manual mentions this briefly, but it doesn't really drive home how frequently it happens in natural conversation. I coded a participant once who did AU1-4 pairs repeatedly while listening to difficult questions, and I kept missing the transitions until I started treating them as separate events rather than trying to process them in real time. Another persistent issue involves asymmetrical expressions. One side of the face may show a clear AU while the other shows nothing or something different. The manual tells you to code both sides separately, but that doubles your workload and makes real-time coding nearly impossible for complex expressions. Frame-by-frame is the only reliable approach here.
About Downloading and Using the Manual
The Action Coding System Manual Ekman is published by the University of Washington's Focus Foundation and can be ordered through their website or academic distributors. It's not inexpensive — a complete set runs into the hundreds of dollars — and it's copyrighted material, so I can't provide a direct download link to the full manual. What I can tell you is that universities with psychology or communication departments often have copies in their libraries, and graduate students typically get access through their programs. There are also certified training courses offered periodically, usually online or at conferences. These courses include access to practice materials and supervised coding sessions, which the manual itself references repeatedly as essential for developing accuracy.
What the Manual Doesn't Cover Well
One limitation worth noting: the manual's reference photographs are predominantly of light-skinned participants. While later editions added more diverse skin tones, coders working with darker skin often report that certain action units — particularly those involving subtle eyebrow or cheek movements — are harder to detect using only the manual's examples. I've found that supplementing the manual with ethnically diverse reference videos helps, but there's no official guidance on this in the text itself. Another gap involves dynamic, conversational expressions. The manual excels at teaching you how to code discrete, isolated expressions. It's less helpful when you're trying to code rapid expression changes that happen during natural speech — which is actually the most common scenario in applied research. The strategies in the manual assume a degree of control over the stimulus that doesn't always exist in real-world data.
Alternatives Worth Considering
If the Action Coding System Manual Ekman feels too cumbersome for your project, there are simpler frameworks. The Dimensional Model of Emotion, for example, reduces everything to valence and arousal scores. It's faster to code and requires less training, though it sacrifices the granular detail that FACS provides. Automated coding software like FaceReader or Affectiva offers real-time detection without manual coding, but these tools have their own accuracy issues, particularly with certain demographics and in uncontrolled lighting conditions. For most serious research, though, the manual remains the standard. It takes time to learn, it's expensive to acquire, and it demands practice to achieve acceptable reliability. But once you can read a face through its action units, you have a level of precision that alternative systems simply can't match.
Getting Started
If you're committing to this, here's the practical path. Order the manual. Watch the training videos that accompany it if available. Practice coding simple expressions first — happiness, sadness, anger — until you can identify the core action units without hesitation. Then move to more complex combinations. Try coding with a partner and compare notes frequently. Aim for at least 80% agreement with a trained coder before you consider yourself ready for independent work. The Action Coding System Manual Ekman isn't quick to master. But it's the closest thing we have to a shared language for describing what faces do, and that makes the effort worthwhile for anyone doing serious work in this area.
