8 Object Recognition Notes

OBJECT RECOGNITION

1. Two visual systems

   1. Dorsal: vision for (“where” function )

   2. Ventral: vision for identification (“what” function)

• Evidence: receptive fields + monkey + neuroimaging + neuropsychological

2. Agnosia

   1. Appreceptive agnosia object recognition problems only when limited stimulus information is present (e.g., when the object is presented from an unusual viewpoint).

   2. Integrative agnosia ability to integrate features of objects into a coherent whole

   3. Associative agnosia recognize the meaning if visually presented objects (patients have normal visual representations, but cannot use them to recognize objects).

   4. Prosopagnosia deficit in facial recognition that cannot be attributed to deterioration in intellectual functions

3. Face recognition

~representing faces at the neural level: Grandmother cells and Ensemble coding

4. Mind reading

1. Two visual systems

a. the dorsal (occipitoparietal) pathway is specialized for spatial perception (for determining “where” an object is)

b. the ventral (occipitotemporal) pathway is specialized for object identification (for determining “what” we’re looking at)

Evidence:

1. From examination of receptive fields

The physiological properties of neurons within the temporal and parietal lobes are distinct:

a. In P, 60% of the neurons have receptive fields that exclude the fovea, and respond to a variety of stimuli – large objects, small objects confined to region in space

b. In T cortex, receptive fields always encompasses the fovea (good at subjecting stimuli to high level fine analysis of the object)

2. From lesion studies with monkeys

Pohl et al (1973) found that animals with lesions to the temporal lobe have great difficulty in discriminated between shapes (“what” discrimination) – for example, they made many errors when learning that a cylinder is associated with a reward when paired with another object. However, they could determine where and object is in relation to other objects. They found the opposite for monkeys with lesions to parietal cortex. Monkeys had to learn to associate with location (when cylinder was in a specific location reward)

Cylinder + another object = reward

3. Neuroimaging

Kohler et al (1995) – used PET

In the object matching task, participants had to identify whether the objects on the screen were the same/different.

in the spatial matching task ----‘’------ presented at the same location. found extensive activation in the dorsal stream In the comparison of object matching vs. spatial matching, they found activation in the ventral stream during object matching, and evidence that these 2 functions are dissociated

4. Neuropsychological

P cortex lesions lead to hemispatial neglect (inability to locate objects in space) and optic ataxia (inability to guide acions according to visual information – eg, reach for objects)

T cortex lesions lead to visual agnosia, which is a selective deficit in recognizing objects. There are different types of agnosia, depending on the specific area of damage. Object recognition may fail for a variety of reasons and different categories of object might also be affected.

2. Agnosia

a. Apperceptive agnosia is characterized by object recognition problems only when limited stimulus information is present (e.g., when the object is presented from an unusual viewpoint).

Thus, they have a deficit in ‘object constancy’ = the ability to recognize an object despite different retinal input (the viewpoint, illumination, surrounding objects etc. lead to changes in retinal imput). The Muller-Lyer illusion shows how object constancy can become hijacked sometimes: we use our knowledge about the cues in the world to inform our object recognition system (rather than match retinal input to a template in the brain). One theory is that the illusion might represent common ways in which we represent edges of buildings (one might represent the inside of a building, and the other – the outside). So the difference in the length might be related to the fact that we are used to representing these type of cues in different 3D situations (we might see the one that seems longer closer to us, and the other one further away).

Henson et al (2002) - fMRI adaptation reveals a region in left fusiform cortex that represents objects in a viewpoint interdependent manner. Method: fMRI adaptation, based on neural adaptation (if a neuron is presented with an input many times, it reduces its firing). If this brain region represents an object in a constant way regardless of the viewpoint that object is seen from, then neurons in that region should reduce its responding to that object, even if the object I presented from different views (the retinal input changes). So neurons in the left fusiform cortex perfroma higher-level function - hold an abstract representation of that object, not just template.

b. Integrative agnosia is an impairment in the ability to integrate features of objects into a coherent whole.

E.g., patient CK was able to copy a figure composed of two diamonds and a circle quite accurately. However, unlike healthy subjects, CK did not draw each object in its entirety, but rather followed the outer boundary, even if doing so meant switching back and forth between objects. This suggests an inability to integrate feature parts (the lines) into a coherent whole (the shapes).

fMRI evidence suggests that there is a specific region in the Lateral Occipital Cortex that performs the ability to integrate features into whole shapes. For example, Kanwisher et al. (1997) showed participants familiar, novel (features that combine into new shapes) and scrambled objects (disconnected features). So, only familiar and novel objects require integrating feature parts into whole shapes. They found greater activation in the Lateral Occipital Cortex for familiar novel pictures in for the familiar pictures and novel pictures.

[CK reproduced pictures only by focusing on the outer boundaries]

c. Associative agnosia is an impairment in the ability...

Buy the full version of these notes or essay plans and more in our Cognition and Emotion Notes.