Home / Insights / Valence, arousal and dominance

Valence, arousal and dominance (VAD), explained for advertising

Valence, arousal and dominance (VAD) are three dimensions used to describe any emotional state. Valence runs from negative to positive: is the feeling pleasant or unpleasant? Arousal runs from calm to activated: how much energy does it carry? Dominance runs from overwhelmed to in control: does the person feel on top of the situation? The model comes from Albert Mehrabian and James Russell's 1974 work on environmental psychology, where it was called PAD (pleasure, arousal, dominance), and Russell's 1980 circumplex model later arranged emotions on a circle of valence and arousal. In advertising, VAD is useful because it separates things a single emotion score blurs. A tense scene and a joyful scene can both be high in arousal, but only one is positive. A busy end frame can be pleasant yet leave viewers with low dominance, which often signals confusion. EchoDepth reports all three dimensions for every second of an ad.

Updated 17 October 2026 by Jonathan Prescott, Founder, Cavefish.

Where VAD comes from

In 1974 Albert Mehrabian and James Russell proposed that people's emotional reactions to their surroundings could be described on three dimensions: pleasure, arousal and dominance. They called it the PAD model, and it was built to explain how environments, from shops to offices, affect behaviour through the feelings they create.

In 1980 Russell published the circumplex model of affect, which arranged emotion words on a circle defined by two axes, pleasure (now usually called valence) and arousal. Excitement sits between high pleasure and high arousal; relaxation between high pleasure and low arousal. Read the paper. Dominance was dropped from the circle, but many applied fields, advertising research among them, still measure it because it captures something the other two miss.

VAD and PAD describe the same three dimensions. VAD is the more common name in emotion AI and affective computing.

The three dimensions, in an ad

Valence−1 to +1
Positive or negative. Valence rises when the joke lands or the product looks appealing, and falls when a scene irritates, confuses or puts people off.
Arousal0 to 1
Calm or activated. Arousal rises with surprise, laughter, tension or a sudden cut. It tells you how much energy a moment carries, not whether that energy is good.
Dominance−1 to +1
Overwhelmed or in control. Dominance often drops during fast, busy or information-heavy scenes, when viewers can't keep up. A pleasant end frame with low dominance may be one people don't fully take in.

Why arousal on its own misleads

Many attention and engagement measures are, in effect, arousal measures. They tell you something happened, not what. A horror trailer and a comedy sketch can produce the same arousal spike. So can a moment of delight and a moment of confusion.

Reading arousal with valence fixes this. High arousal with positive valence is excitement or delight. High arousal with negative valence is stress, tension or annoyance. The plane below shows the four combinations.

The valence and arousal plane A square divided into four quadrants. The horizontal axis is valence, from negative on the left to positive on the right. The vertical axis is arousal, from calm at the bottom to activated at the top. Top right is excited or delighted, bottom right is content or relaxed, top left is stressed or tense, and bottom left is bored or sad. NegativePositiveValence Arousal Excited, delightedContent, relaxed Stressed, tenseBored, sad The punchline landsA warm, slow brand film A confusing, rushed demoThe ad has lost them
Valence runs left to right and arousal bottom to top. The amber dot marks the kind of moment that lifts the Emotional Response Curve most: clearly positive and high in energy. Dominance, the third dimension, is not shown.

How EchoDepth measures VAD

EchoDepth does not ask viewers to rate their feelings on these scales. It measures them from facial movement. Consenting viewers watch the ad on their own webcam, software scores the Action Units of the Facial Action Coding System on every frame, and each dimension is calculated as a weighted combination of Action Units. For example, a cheek raise with a smile (AU6 with AU12) raises valence; a brow lowerer (AU4) lowers valence and dominance; raised brows (AU1 with AU2) raise arousal.

The full table of which Action Units raise or lower each dimension is in the methodology. Results are averaged across viewers for each second of the ad. For the evidence on what facial movements can and can't tell you, see is facial coding accurate?

How the Emotional Response Curve uses valence and arousal

EchoDepth's Emotional Response Curve combines valence and arousal for each second of an ad, averaged across viewers. Valence is weighted by arousal, so a strongly positive moment, such as a laugh, lifts the curve more than a mildly pleasant one, and a negative moment pulls it down. How the curve is calculated, with a worked example, is in the methodology.

Dominance is reported alongside the curve but is not part of it, because a dip in dominance can happen during both good and bad moments and is more useful read on its own.

The curve is separate from the Cavefish Emotional Index (CEI). The CEI does not use valence, arousal or dominance at all. It is a 0 to 100 score read from what viewers write or say about the ad after watching, averaged across five readings (Advocacy, Goodwill, Engagement, Drive and Assurance) and banded Strong (80 and above), Good (60 to 79), Watch (40 to 59) or Act (below 40). The curve shows what faces did in the moment; the CEI shows how people feel about the ad afterwards.

Questions

None in substance. PAD stands for pleasure, arousal and dominance, the names Mehrabian and Russell used in 1974. VAD uses valence instead of pleasure. Both describe the same three dimensions of emotional state.

There is no universal threshold. Valence is most useful read over time and compared between edits or with similar ads in the same market. Look for where it rises and falls rather than a single number.

Arousal shows how much energy a moment carries, which relates to whether people pay attention and remember it. But arousal alone doesn't say whether the reaction is positive or negative, so it should always be read with valence.

Yes, by coding facial movements (Action Units) and combining them. It works best averaged across many viewers, because a single face is not a reliable readout of feelings.

A drop in dominance often marks a busy or confusing scene, which can happen in both positive and negative moments. EchoDepth reports it separately so it can be read on its own.

References

  1. Mehrabian, A., & Russell, J. A. (1974). An Approach to Environmental Psychology. MIT Press.
  2. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39(6), 1161–1178. doi:10.1037/h0077714
  3. Höfling, T. T. A., & Alpers, G. W. (2023). Automatic facial coding predicts self-report of emotion, advertisement and brand effects elicited by video commercials. Frontiers in Neuroscience, 17. doi:10.3389/fnins.2023.1125983

See it on your own ad

EchoDepth pre-tests video ads on real viewers and reports the response second by second, with every score traced back to the Action Units behind it.

Test an adHow ad testing worksRead the methodology