Skip to content
PaperFren

Do we still see faces 'as a whole' when they're tilted up or down?

Open paper intelligence

Turning faces upside down hurt recognition for all side-turned views and for moderate up/down tilts, but not for extreme up/down tilts, suggesting holistic face processing stops working at those unusual angles.

Source

The face inversion effect following pitch and yaw rotations: investigating the boundaries of holistic processing

Favelle SK, Palmisano S · Frontiers in psychology · 2012

doi.org/10.3389/fpsyg.2012.00563Read the full paper ↗8 citationscc by

Study at a glance

Design
Human experiment — Within-subjects sequential same/different face matching: a full-face study image followed by a test face rotated 0, 15, 45 or 75 degrees in yaw (sideways) or pitch (up/down), with both faces upright or both inverted.
N
N=20 · 20 undergraduate volunteers, each completing 360 trials.
Population
Undergraduate students at the University of Wollongong, Australia, unfamiliar with the face stimuli
Outcome
Matching sensitivity (d′) and reaction time for correct 'same' responses; the face inversion effect (upright minus inverted) at each viewpoint

Structured fields used in claim comparison tables when every cited study has a complete layer.

What they did

Twenty undergraduates saw a front-on photo of a man's face, then a second photo, and decided whether it was the same person. The second face could be rotated sideways (yaw) or tilted up or down (pitch) by 15, 45 or 75 degrees, and in separate blocks both faces were shown upside down. The logic was that if a view is processed holistically when upright, inverting it should disrupt recognition.

What they found

Recognition was best for front-on faces and got worse as rotation increased. Inversion reduced recognition sensitivity at every sideways view and at up/down tilts of 15 and 45 degrees, but there was no inversion effect at all at 75-degree tilts from above or below. Because performance at 75 degrees was still above floor (d′ around 0.5), the missing inversion effect was unlikely to be due to people simply failing the task.

The limits

What it doesn't show

The sample was small and the stimuli were only nine Caucasian male faces with hair removed, so the results may not extend to other face sets. The inversion effect is an indirect marker of holistic processing; the study did not use more direct tests such as the composite-face task. The authors cannot tell whether extreme pitch views are processed holistically but inefficiently, processed as faces but piece by piece, or treated like ordinary objects.

Key terms

Face inversion effect
The finding that turning faces upside down impairs recognition much more than it does for other objects.
Holistic processing
Perceiving a face as an integrated whole rather than as a collection of separate features.
Yaw and pitch
Yaw is turning the head side to side toward profile; pitch is tilting it up or down.
d′ (d-prime)
A signal-detection measure of how well someone discriminates 'same' from 'different', independent of their bias to say one or the other.
Face template
A proposed generic mental representation of an upright face that incoming images are matched against to trigger holistic processing.

Flashcards

1 / 10

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.

Open paper intelligence

Quiz yourself

1 / 5

What does 'pitch' rotation of a face mean?

Common questions

Why is inversion a useful test of holistic processing?

An inverted face has exactly the same low-level features as the upright one, so any drop in performance must come from higher-level, face-specific processing such as holistic integration.

Why might steep up/down views not be processed holistically?

The authors suggest these views, rarely seen and losing much facial information, may not match our experience-based face template, so holistic processing is never triggered and inversion has nothing to disrupt.

Did recognition still work at 75-degree tilts?

Yes, above chance, which is why the absence of an inversion effect there is meaningful rather than a floor effect.

More on Face processing