Can your eyes give away a face you're lying about knowing?
When students denied recognising a lecturer's face, their eyes lingered longer on it than on truly unknown faces, and this held even when they deliberately tried to look at every face the same way.
Source
Eye see through you! Eye tracking unmasks concealed face recognition despite countermeasures
Study at a glance
- Design
- Human experiment — Sequential Concealed Information Test with eye tracking; within-subjects face type (familiar probe, familiar target, unfamiliar) crossed with between-subjects instruction (standard guilty vs fixed-sequence gaze countermeasure)
- N
- N=48 · 48 undergraduate psychology students, 24 in the standard guilty group and 24 in the countermeasures group; 1100 trials analysed
- Population
- Undergraduate psychology students at the University of Stirling who knew the lecturers shown for at least one semester
- Outcome
- Eye-movement markers of recognition (average and first fixation duration, number of fixations, face regions visited, proportion of fixations to inner features/eyes) expressed as Cohen's d and ROC area under the curve
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Students were shown photographs one at a time and had to say 'no' to one lecturer they actually knew (the lie), 'yes' to a second known lecturer, and 'no' to unfamiliar look-alike faces, while an eye tracker recorded their gaze. Half were simply told to conceal recognition and appear honest; the other half were also told to use a fixed eye-movement route over every face (forehead, ears, eyes, nose, mouth, chin). Each person did four blocks with new lecturers and new unfamiliar faces. The researchers compared gaze on the lied-about faces with gaze on unfamiliar faces using effect sizes and ROC classification.
What they found
In both groups, fixations were longer on the familiar face being denied, from the very first fixation; the effect was, if anything, larger when people used the countermeasure (d = 0.91) than when they did not (d = 0.66). Without the countermeasure, people looked more at the eyes of familiar faces, but this cue vanished when they followed the fixed route. A smaller share of fixations on inner features (eyes, nose, mouth) marked familiarity in both groups. Contrary to predictions, fixation counts did not reveal recognition, because in the standard condition people made only about two fixations per face before responding.
The limits
What it doesn't show
Roughly 30% of participants did not show the longer-fixation marker, and the authors cannot explain why, so the method would miss some liars. Participants knew in advance which face to lie about and faces were always centred after a central cross, which let them respond with very few fixations; different test formats might change the results. This was a low-stakes lab lie about lecturers with 24 people per group, so it does not show how accurate the test would be with real suspects or with newly seen faces from a crime.
Key terms
- Concealed Information Test (CIT)
- A lie-detection method that presents one crime-relevant item among many irrelevant ones and looks for a different response to the relevant item, rather than asking whether someone is lying.
- Countermeasure
- A deliberate strategy a guilty person uses to beat a detection test, here following the same fixed gaze path over every face.
- Average fixation duration
- The mean length of time the eyes stay still on one point, treated as an index of how deeply something is being processed.
- Cohen's d
- A standardised effect size expressing how many standard deviations apart two conditions are.
- ROC area under the curve (AUC)
- A measure of how well a marker separates two categories, where 0.5 is chance and 1 is perfect classification.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which marker detected concealed recognition in both the standard and countermeasure groups?
Common questions
Why did the countermeasure fail to hide recognition?
It removed some cues, such as extra looking at the eyes, but fixation duration still increased for familiar faces, possibly because recognition is fast and automatic and the effort of concealment adds its own processing load.
Why didn't the number of fixations reveal lying, as earlier studies found?
Participants made only one or two fixations before answering in the standard condition, leaving little room for count-based differences; the authors think knowing the target identities in advance made the task too easy.
Could this be used by police tomorrow?
Not yet: about 30% of people did not show the key marker, the lies were low-stakes, and the best test format for eye-based CITs is still unknown.
More on Face processing