Blink rate and deception: what the research shows

← All articles · By the RedFlag team · September 4, 2026

Blinking is one of the few behavioral signals with a genuinely interesting research record — not because it reveals lies (it does not), but because it tracks mental effort in a measurable, replicated way. Here is what the evidence says, and what it takes to measure blink rate honestly in a video.

The baseline: wildly individual

A resting adult blinks somewhere around 15–20 times per minute, but the range across individuals is enormous — some people blink 4 times a minute, others 40. Blink rate also shifts with dry eyes, screens, contact lenses, lighting and medication. This is the first honest conclusion: any absolute blinks-per-minute threshold is meaningless. Only a comparison against the same person's own baseline, in the same recording, can carry information.

Load suppresses blinking; release rebounds it

The replicated core finding is about cognitive load. When people concentrate — mental arithmetic, demanding recall — blink rate drops (Holland & Tarlow, 1972, 1975; confirmed with modern eye-tracking many times since). When the load lifts, blink rate rebounds, often overshooting the baseline.

Leal and Vrij (2008) applied this directly to deception: participants showed suppressed blinking while lying, followed by a compensatory rebound in the seconds after. In a rare high-stakes field study of real police interviews, Mann, Vrij and Bull (2002) found suspects blinked less and paused more when lying — the opposite of the nervous-fidgeting stereotype.

Measuring it in real video is the hard part

Laboratory studies use eye-trackers. In an ordinary YouTube video you have a compressed, sometimes small, sometimes badly lit face. Honest measurement needs at least:

• A minimum face size. Eyelid movement cannot be resolved on a face that spans a few dozen pixels. RedFlag only records blink data when the face is at least 22% of the frame height; below that, the blink channel reports nothing rather than noise.
• Hysteresis in the detector. A blink is an eye-closure crossing a closing threshold and re-opening past a separate one — otherwise compression flicker gets counted.
• Normalization by observed face-time. A rate computed over minutes where the face was off-screen half the time is not a rate.
• The person-relative comparison — rate divided by that speaker's own median across the whole video, never an absolute cutoff.

How RedFlag uses blink rate

Blink elevation is one of three signals in RedFlag's verdict, at a weight of 0.25. It activates when the speaker's blink rate exceeds roughly 1.3× their own video-wide baseline over a 20-second window; unflagged stretches typically sit just below baseline. RedFlag is not a lie detector — a blink elevation is a deviation worth a second look, and the interpretation is yours. Full details: how RedFlag scores a video.

References

• Holland, M. K., & Tarlow, G. (1972). Blinking and mental load. Psychological Reports, 31(1), 119–127.
• Leal, S., & Vrij, A. (2008). Blinking during and after lying. Journal of Nonverbal Behavior, 32(4), 187–194.
• Mann, S., Vrij, A., & Bull, R. (2002). Suspects, lies, and videotape: an analysis of authentic high-stake liars. Law and Human Behavior, 26(3), 365–376.
• DePaulo, B. M., et al. (2003). Cues to deception. Psychological Bulletin, 129(1), 74–118.