Questions: Cortical Face Processing

Question 1

A researcher takes a face and scrambles the spatial layout of its features — the eyes, nose, and mouth are all clearly recognizable as individual parts, but their positions relative to each other are rearranged. Compared to normal faces, recognition of these scrambled faces should:

• A. Improve slightly, because the sharpness of individual features is no longer obscured by distracting context
• B. Remain the same, since the face parts are all intact and the brain processes features independently
• C. Degrade substantially, because face recognition depends on configural information — the spatial relationships among features — not just the features themselves
• D. Degrade only for inverted faces, not for upright ones, since inversion disrupts only feature processing

Answer: (C)

Face recognition is configural: the brain encodes the spatial relationships among features (the distance between eyes, the relative position of nose and mouth) as an integrated whole, not as a bag of independent parts. Disrupting these relationships — even while leaving each feature perfectly intact — impairs recognition dramatically. This is the central evidence for holistic face processing. Option B represents the classic feature-based misconception that configural processing directly refutes.

Question 2

Expert bird-watchers show elevated FFA activation when viewing images of birds. The most theoretically parsimonious interpretation of this finding is:

• A. The FFA evolved two separate modules — one for faces and one for birds — and bird-watchers happen to have both active
• B. The FFA is an area for expert-level individuation of visually homogeneous object categories; faces simply happen to be the category every human practices to expertise from infancy
• C. Bird-watching training permanently rewires face-specific neurons to respond to birds instead
• D. The FFA is a general visual area with no specialization, and all complex object categories activate it equally

Answer: (B)

The expertise hypothesis proposes that the FFA is not a face module per se, but a region that becomes recruited for fine-grained within-category discrimination of any sufficiently practiced, visually homogeneous category. Faces are universal because every normally developing human becomes an expert face recognizer from birth. Bird-watchers, car experts, and chess players show analogous effects for their domains. This interpretation bridges face specialization with general perceptual learning rather than requiring a purpose-built 'face module.'

Question 3

Turning a face upside down impairs recognition more severely than turning a similarly complex non-face object upside down.

• T. True
• F. False

Answer: True

This is the classic 'face inversion effect' — one of the strongest pieces of evidence for configural face processing. Objects can be recognized upside-down fairly well because object recognition relies on feature analysis that works in any orientation. Face recognition degrades dramatically when inverted because configural processing — reading the spatial relationships among features — requires the face to be in its canonical upright orientation. Configural information is not simply recovered by mentally rotating the image.

Question 4

Prosopagnosia — the inability to recognize individual faces following brain damage — is typically accompanied by equally severe deficits in recognizing other complex visual objects such as cars or tools.

• T. True
• F. False

Answer: False

The hallmark of prosopagnosia is a selective dissociation: face recognition is severely impaired while recognition of other object categories remains largely intact. A prosopagnosic patient may be unable to recognize their own face in a mirror yet easily identify a hammer or a cup. This dissociation is the primary neural evidence for specialized face-processing machinery in the ventral stream (especially the FFA). If face and object recognition relied on identical mechanisms, selective damage to only one would be impossible.

Question 5

Why does the Thatcher effect demonstrate that face recognition is configural rather than feature-based?

Model answer: The Thatcher effect shows that locally rotating the eyes and mouth within an inverted face looks grotesque when viewed upright but goes almost undetected when the whole face is inverted. If the brain processed faces feature-by-feature, the local rotations would be equally detectable in both orientations. Instead, the brain is sensitive to violations of the spatial relationships among features (configural information) only in the upright orientation — the canonical one where configural processing is engaged. When the face is already inverted, configural processing is disabled, so the local distortions go unnoticed.

The Thatcher effect is powerful because it cleanly separates feature-level and configural processing. The individual features (eyes, mouth) are individually normal when the face is inverted — each looks upright — so feature detectors are not triggered. It is only when the whole face is upright that the configural violation (eyes and mouth oriented opposite to the face) becomes horribly salient. This shows that configural processing, not feature detection, is doing the work of normal face recognition.