Questions: The Ventral Stream and Object Recognition

Question 1

A patient with damage to the inferior temporal cortex can copy a drawing of a hammer accurately and describe it as 'a wooden handle with a heavy metal head.' However, they cannot name the object or explain its use. What does this dissociation indicate?

• A. The patient has lost basic visual sensation in that part of the visual field
• B. Object recognition and basic visual sensation rely on separable neural systems
• C. The patient's language areas are damaged, preventing naming
• D. The dorsal stream has compensated for the damaged ventral stream

Answer: (B)

The patient can copy the drawing (intact visual sensation and motor output) and describe its features (intact low-level processing), but cannot recognize it as a hammer. This dissociation — intact sensation, impaired recognition — is the signature of ventral stream damage. The ventral stream transforms visual features into object identity; the patient's basic visual system is intact, but the recognition machinery is offline.

Question 2

A neuron in inferior temporal cortex fires strongly when a monkey sees its trainer's face. The trainer moves 3 meters away, turns slightly to the side, and puts on glasses. The neuron fires at approximately the same rate. This property is called:

• A. Orientation tuning — the neuron prefers faces regardless of orientation
• B. Population coding — many neurons together represent the face
• C. Invariant object representation — selectivity is maintained across changes in size, position, and viewpoint
• D. Category selectivity — the neuron responds to any face, not just the trainer's

Answer: (C)

This is the defining property of IT cortex neurons: invariant representation. The neuron maintains its response to a specific object across transformations — changes in size, retinal position, viewing angle, and even partial occlusion. The visual input is completely different across these conditions, yet the neural response is stable. This invariance is what allows you to recognize your friend whether they are near or far, facing toward you or slightly away.

Question 3

V1 (primary visual cortex) neurons can recognize objects directly if enough of them pool their responses — invariance emerges from combining many V1 neurons together.

• T. True
• F. False

Answer: False

V1 neurons respond to simple local features like oriented edges at specific positions in the visual field. They are exquisitely sensitive to position and orientation, not invariant. Invariance emerges gradually through the hierarchy: V2/V3 encode intermediate complexity, V4 encodes surfaces and shapes, and IT cortex achieves invariant, category-level representations. The full hierarchy up through IT cortex is required — V1 pooling alone cannot produce it.

Question 4

Prosopagnosia — selective inability to recognize familiar faces — can occur in patients who retain intact object recognition for non-face categories.

• T. True
• F. False

Answer: True

Prosopagnosia is a documented dissociation: patients with bilateral damage to face-selective regions (especially the fusiform face area) cannot recognize familiar faces — including their own in a mirror — yet identify common objects normally. This double dissociation is strong evidence that IT cortex is organized into functionally distinct regions specialized for particular visual categories, rather than performing a single uniform recognition computation.

Question 5

Explain why visual agnosia — not blindness — is the expected consequence of inferior temporal cortex damage. What does this tell us about how the ventral stream is organized?

Model answer: IT cortex damage disrupts the final stage of object recognition but leaves earlier visual processing intact. A patient with IT damage still receives retinal input, processes edges and colors in V1/V2/V4, and has intact dorsal stream function for spatial vision. What they lose is the transformation of visual features into categorical identity. Agnosia rather than blindness is expected because the ventral stream is a processing hierarchy, and early stages (sensation) are anatomically separate from late stages (recognition).

The hierarchy and specialization within IT cortex explain why different agnosias can be selective: damage to face-selective regions produces prosopagnosia while object recognition is intact; damage to other IT sub-regions can produce selective deficits for tools, animals, or scenes. These selective dissociations are only possible because recognition is not a single monolithic computation but a distributed hierarchy with distinct functional zones.