Micro-Expressions: Could Micro-Expressions Be Quantified? Electromyography Gives Affirmative Evidence

0:00Welcome to this deep dive. Today, we're taking a granular look at the architecture of human emotion. Specifically, we're focusing on microexpressions. These are the brief, subtle facial movements that can, well, involuntarily disclose concealed feelings. They act as these tiny windows into a person's genuine emotional state. And for decades, our understanding of them has rested on mainly subjective visual observation. And that subjectivity is the problem, isn't it? That's what we're tackling today. Microexpressions, or MEs, are such vital signals in social interaction.

0:33Yet our reliance on manual coding, even on systems as robust as the facial action coding system, it just introduces this unavoidable inconsistency. So our mission today, guided by this source material, is to explore some groundbreaking research. Yes, research that seeks to replace that human guesswork with objective, rigorous physiological data. We're asking, can we define an ME not by what a person sees, but by what the body actually measures?

0:57The methodology here seems to be a significant step forward. It involves pairing facial electromyography, EMG, which measures muscle activity, with high-resolution video recordings. And they captured a huge data set. A substantial one, yes. 233 microexpressions, and then 147 of the more standard macroexpressions. And this was all from 35 participants. The findings are giving us a new, a two-dimensional physiological definition for these MEs.

1:25Okay, so before we dive into the data, let's ground ourselves in the basic definitions. We should. Traditionally, MEs are defined by their brevity, they are involuntary movements, and the key is they last less than 500 milliseconds. Half a second. That half-second boundary is the theoretical barrier. By contrast, you have macroexpressions, or MEs. They're longer, far more intense, and they typically fall within the realm of conscious control.

1:50They're what you think of as a normal facial expression. But the challenge has always been proving that distinction scientifically, especially when you get to intensity. I mean, the facial action coding system, FACS, it allows researchers to code every single muscle action unit or AU on the face. It's a gold standard. But when it comes to intensity, FACS asks human coders to, well, rate the strength of the expression on descriptive scales.

2:17That feels like subjective judgment. It is subjective judgment. Imagine asking two highly trained coders to rate a fleeting muscular twitch, one lasting three-tenths of a second. They have to rate it on a scale from trace to maximum. Their ratings could be all over the place. They might vary wildly. What one person calls a level-two movement, another might call a level-three. And this subjectivity means the data isn't replicable. It creates inconsistencies across studies.

2:44The situation just demanded an objective measure. And that's where the researchers identified electromyography, the EMG, as the optimal solution. It is the optimal tool. It's because of its temporal resolution. EMG measures the electromagnetic field potential that's generated every single time a muscle fiber contracts. And it can do that with millisecond precision. Millisecond precision. When you are trying to quantify an event that occurs in less time than it takes to blink, you need an instrument that reports the muscle firing directly.

3:12You're shifting the analysis from visual approximation to direct biological truth. So to get that level of measurement, the technical setup had to be specialized. The research team found that commercial EMG devices often have, well, limited channels or they're too complex to configure. Which would have compromised the study's flexibility, so they built their own. A custom eight-channel EMG module. That's right, based on a Texas Instruments ADS-1299 chip.

3:40They essentially customized their own instrument to ensure the highest possible fidelity for this facial data. And that precision extended to their sampling choices. It did. The typical frequencies for facial muscle signals, they range from about 20 to 500 hertz. The researchers selected a 1,000 hertz sampling rate. Okay, why not higher? Why not 2,000? Well, they reasoned that while 2,000 hertz would capture more data, it would also capture unnecessary noise.

4:05And that complexity could pollute the signal. They were seeking that perfect balance between high speed and clarity for this specific task. And then there's the critical issue of synchronization. How do you align a stream of millisecond-precise EMG data with video evidence that's only captured at 30 frames per second? Sounds complicated. It does, but their solution was elegantly simple. They integrated a digital tube display right into the video frame.

4:32This tube continually incremented its count, and it was captured by the 4K camera. So as the computer recorded each packet of EMG data, it updated that counter. It's like a time stamp. A precise frame-by-frame stamp. It allowed them to align the physiological signals with the visual evidence without any error. Turning to the anatomy of the data collection itself. Yeah. What about the electrode placement?

4:53They chose bipolar electrodes. These are superior because they're far less prone to interference or cross-talk between muscle groups. And they concentrated the placement on the left half of the face. The left half. Targeting seven key muscles. The ones for fundamental expressions. Like the frontalis and the forehead. The corrugator supracilii for the brow furrow. And the zegomatic is major for smiling. But why the left side specifically?

5:19Previous research has consistently suggested that the left side of the human face tends to exhibit stronger emotional expression. It often reflects a bias in the right hemisphere of the brain for processing emotion. And the right side was kept clear. Exactly. That was to facilitate the concurrent manual coding process by the human reviewers. Okay. So, with this robust data collection in place, the study delivered its core objective findings.

5:43The physiological evidence now quantitatively concerns that microexpressions have both shorter duration and, crucially, lower intensity. Yes. This moves the definition out of the realm of theory and into physics. And, to objectively quantify that intensity, the researchers introduced MVC percent. That's a percentage of maximum voluntary contraction. Explain that. What is MVC? Think of MVC as the maximum force a muscle can generate. If you were to say clench your jaw as hard as you possibly can, that's 100% MVC.

6:13It provides an objective baseline against which all other expressions can be measured. And, the numbers that resulted, they paint a vivid picture. For the macroexpressions, the typical visible emotional movements, they showed a mean MVC percent of 23.09%. A measurable conscious contraction. But, in striking comparison, the microexpressions registered a mean MVC percent of only 8.11%. Dramatically lower. This is the empirical proof of that low intensity.

6:43And, they narrowed it down further. They provided a 95% confidence interval, which places the typical intensity of a microexpressions between 7.01% and 9.21% of the maximum possible contraction. We're talking about muscle activity so faint you would probably be incapable of executing it consciously. That's a good way to put it, yes. Even if you attempted to fake a microexpressions, it would be difficult to produce something that weak.

7:06That makes the movement almost imperceptible. Now, what about duration? For duration, we can look at the integrated EMG, or IEMG. It's a measure that combines both the amplitude and the time of the muscle signal. For macroexpressions, the average IEMG score was over 440. For microexpressions? A mere 78. A huge difference. And, when they isolated just the duration, the 95% confidence range for MEs was found to be precisely 307 to 327 milliseconds.

7:35That interval is remarkably small. It confirms that half-second threshold everyone assumed. But here's the critical leap, the relationship between the two. Did they find a connection between duration and intensity? They certainly did. A linear regression analysis demonstrated compelling evidence. It showed that duration was a significant predictor for intensity, with a beta value of 0.492. Meaning? Meaning a shorter duration is intrinsically linked to a lower intensity.

8:01Prior views often focused only on duration. This research elevates the definition to a two-dimensional construct. Microexpressions are defined by their short duration and their low intensity. So the physics checks out. They're measurable, short, and weak. Let's shift to the behavioral and psychological dimension. The researchers explored the internal characteristics. Unawareness and uncontrollability. Why do these leaks happen? To test this, they put participants under cognitive strain.

8:30They viewed emotionally evocative videos, but were told to strictly maintain a neutral expression. The emotion suppression paradigm. Exactly. And they were asked to press a key if they became aware that an expression had leaked out. That key press was the measure of self-awareness. And how good are we at detecting our own emotional leakage under pressure? The results were quite telling. A staggering 93% of microexpressions were not reported by the participants.

8:5693%. They occurred entirely outside the threshold of conscious self-awareness. For perspective, 70% of the larger, longer macroexpressions were reported. This confirms the fundamental nature of the microexpression. It is a genuine, involuntary leakage that happens before we can consciously register it. Wait, just to let that sink in. 93% of these emotional leaks happened without the person even knowing. That statistic fundamentally changes how we think about self-monitoring.

9:25That is an astonishing degree of unawareness. It is indeed. And furthermore, the study examines self-control. After the viewing, participants assessed their own ability to control their facial expressions using a nine-point scale. And the analysis of this data yielded a paradox. Which was? The higher the self-control rate a participant reported, the more microexpressions appeared on their face. That seems backward. If I try harder to control my face, I should see fewer leaks, not more.

9:51What's the mechanism there? The proposed cognitive mechanism is that self-control, it successfully restricts the production of the larger, longer macroexpressions. But this act of intense, sustained suppression requires significant cognitive effort. It's a kind of mental overload. And in that state of focused suppression, the tiny, involuntary, brief muscle movements, the microexpressions, they slip through the cognitive net. They're spontaneous. And they're challenging to inhibit precisely because that conscious effort is focused elsewhere.

10:23So the very act of suppressing a big expression is what allows the tiny ones to escape. That strongly reinforces the idea that microexpressions are genuine indicators of concealed emotion. It justifies their use in high-stakes fields like deception detection, where people are trying their hardest to conceal something. This provides the physiological and cognitive evidence that validates a decades-old hypothesis. Now, a major byproduct of this methodical work is the creation of a new resource,

10:52the CASMAMS database. Tell us what makes this so important. CASMG is unique. It's the first microexpressions database to include the objective EMG signals right alongside the video recordings. It encompasses 380 total samples. For researchers, this is transformative. It meticulously provides keyframe timings, the specific action units involved, and emotion labels for every single sample, all backed by objective measurements. What's the practical utility of including that EMG data?

11:19Accuracy and detection. The study developed a baseline semi-automatic detection method based purely on the EMG data, and this method demonstrated superior performance compared to the traditional vision-only methods. In what way? In accurately localizing the ME intervals, pinpointing the exact start and end of the expression. It achieved an 84.7% average overlap rate with the ground truth. It solves the enormous difficulty that comes with manual, frame-by-frame coding.

11:48Even with these advances, the researchers acknowledged limitations. What were the key boundaries they identified? The main one was the physical setup. Since electrodes were only placed on the left half of the face, they had to exclude 21 samples where asymmetrical MEs occurred only on the right side. So some spontaneous data was missed. Inevitably, yes. Additionally, the macro expressions they collected, which averaged about 23% BAC,

12:11were lower in intensity than what you might see in unconstrained social interactions. Because they were in a lab trying to suppress their feelings. Corrupted control environment. So looking forward, how do they suggest overcoming these challenges? The source material emphasizes two paths. First, future work has to incorporate culturally diverse participant pools to enhance the universality of the conclusions. And second, to overcome the constraints of meeting a physical EMG device in the real world,

12:39the focus has to shift to multimodal models. What does that mean? It means models that will integrate the objective physiological data from EMG training with advanced computer vision techniques. This would allow for powerful non-contact recognition algorithms that you could deploy anywhere. So to synthesize the core knowledge from this deep dive, this meticulous research has provided the long-sought, precise physiological boundaries that define microexpressions.

13:06We have moved MEs from a theoretical idea to a quantifiable biological event. One characterized by a duration interval locked between 307 and 327 milliseconds and an intensity range of 7% to 9.2% MVC. And critically, these findings validate that MEs are unaware and uncontrollable leaks of genuine emotion. This objective confirmation of uncontrollability, well, it raises a profound question for you, the listener, to consider.

13:33If we continue to develop sophisticated, non-contact AI systems trained on this precise, objective EMG data, what might that future mean for our ability to maintain privacy over our internal emotional states? Since 93% of these leaks happen outside our own conscious awareness, AI detection could invade a realm of privacy we didn't even know we possessed. The CASMEMG database is now the proving ground for exactly that future technology.

13:58Thank you for joining us for this deep dive into the objective quantification of the human face.