From Retina to the Brain 3

Question 1

Explain what happens when shifting between perceiving two images

Answer 1

By moving the card further away, all spatial frequencies shift to higher spatial frequencies. This shifts what word on the card is better visible than others.

Question 2

What does the perception of the watch in 3D using a sheet demonstrate?

Answer 2

Stereopsis from disparity differences between the two eyes ( slightly different in the POV of the watch)

Question 3

Explain the concept of disparity with reference to looking at point p

Answer 3

When you fixate on point P, you converge the two eyes . P will project on the fovea on each eye. Every other stimulus that falls within the same plane of depth (the horopter) will be projected at equal distance from the fovea in the two eyes. It has 0 disparity.

A stimulus that is more near in depth will produce unequal distance projections. This is called disparity (near disparity in this case, similar for far disparity.)

Question 4

What problem is posed by this theory? Describe it

Answer 4

The correspondence problem;
Each eye/ camera views three image primitives (dots). The problem then is, which dots in the left eye correspond to which dots in the right eye? the 9 dots represent all the possible matches that could be made, the black dots are the correct matches and the rest are incorrect, (referred to as either ‘false targets’ or ‘ghosts’)

Confronted by these 9 possible matches, we found ourselves are capable (in this instance) of making the three correct matches. The interesting thing about using the dot example is that no high level information or cues are presented to help the viewer in matching.

Question 5

What conclusion was then drawn by researchers regarding this problem?

Answer 5

Stereo matching matching is performed early in the human visual processes, it is assumed to be. a low level operation. Only after the two images are matched is any attempt made to understand what is actually being viewed.

Question 6

Name a practical example where failing of the correspondence problem may occur

Answer 6

When you are approaching a fence or other repetitive grating (could bump into it thinking that its further or grasp it thinking that its closer). Because of the repetitive nature of the grating, your brain makes these false matches.

Question 7

What is the result of when the correspondence problem can be solved?

Answer 7

3D image

Question 8

What results when the images are too different (as in squinting (cross eyes) or experimentally)? (3)

Answer 8

Fusion does not occur and the result is

• Double vision (diplopia)
• Suppression of one of the eyes
• binocular rivalry (mostly experimental

Question 9

In what two ways can suppression of one of the eyes occur?

Answer 9

Normal suppression, eye dominance
Pathological suppression (amblyopia, lazy eye)

Question 10

What is meant by Panum’s area?

Answer 10

Only objects that are not too far or too near relative to the horopter have small enough disparity to result in fusion (and hence depth perception). This region in space is called Penum’s area.

Question 11

how are objects that are farther than or closer than Panum’s area perceived? (2)

Answer 11

Results in either diplopia or suppression of their image in one of the eyes (usually the non-dominant eye)

Question 12

Give an example of a way you can demonstrate failed correspondence in yourself. Explain how to do it and what is going on

Answer 12

the ‘floating finger sausage.’ Hold your two index fingers about 5 inches in front of your eyes with their tips half an inch apart. Now look beyond them and not the weird result. Move your fingers out further and the retinal disparity- and the finger sausage- will shrink.

When the two images of the two eyes cannot be fused, normally one of them is suppressed (dominant eye test). Here the brain interprets the correspondence such that a new object emerges in near space relative to the horopter.

Question 13

What is meant by the term of strabismus?

Answer 13

Misalignment of the eyes

Question 14

What can cause strabismus? (2)

Answer 14

Congenial eye-muscle disorders, or due to cranial oculomotor nerve disorders (neurological disease)

Question 15

What do people with strabismus perceive?

Answer 15

No part of the two images will correspond, resulting in double vision, and eventually suppression of the input in one of the two eyes.

Question 16

What can be the result of strabismus during childhood? (0-4 years)

Answer 16

The result will be the permanent suppression of the input from one eye: amblyopia.

Question 17

What happens in the brain during amblyopia?

Answer 17

the ocular dominance receiving input from that eye become much thinner , which is why that eye has reduced vision.

Question 18

Name the four types of strabismus

Answer 18

Hypotropia (eye turns down)
Hypertropia (eye turns up)
Exotropia (eye turns out)
Esotropia (eye turns in)

Question 19

What occurs when the two images are completely different?

Answer 19

Binocular rivalry: While the images in the two eyes remain constant, the (conscious) percept spontaneously switches between the one and the other image, with different durations of dominance for each stimulus.

Question 20

When is the only time this usually happens?

Answer 20

Experimentally using devices to present two images to the two eyes

Question 21

Explain the neural mechanisms behind disparity, and when specifically it is activated.

Answer 21

If you record with a micro-electrode from a V1 neuron while an animal views orientated lines presented separately to the two eyes and vary the disparity, some neurons are selective for particular disparities. This neuron does not respond at all when a line is shown to one eye at a time. To get a response, the line must be presented simultaneously to both eyes, it must have the correct orientation, direction of motion and the correct binocular disparity. (e.g 1/2 degree (30’ of visual angle)

Question 22

In which visual areas are these disparities found in monkeys? (4)

Answer 22

V1, V2, V3/ V3A

Disparity in detected early in visual cortex

Question 23

In what ways can these disparity neurons be tuned? (4)

Answer 23

Tuned near
Tuned far
Tuned zero
Tuned inhibitory

(Look at graphs in docs)

Question 24

In what way can stereograms be used to test a certain type of disparity?

Answer 24

Using correlated and anticorrelate random dot stereograms, it can be tested whether neurons just compute disparity of contrast, or compute (perceived) depth.

Question 25

How are correlated dots perceived?

Answer 25

When the left and right eye see a stereogram with disparity in the two eyes, depth is perceived (a bulging dome). In this case the dots are correlated. i.e the matching dots are black in both eyes or white in both eyes.

Question 26

How is it perceived in the anticorrelated display?

Answer 26

The matching dots have opposite colours i.e matching dots are white in one eye and black in the other (and v.v.) in this case no depth is perceived. (fusion does still occur however)

Question 27

What did Cumming & Parker (1997) predict in a study regarding these anti-correlated dots?

Answer 27

Cumming & Parker 1997 recorded from neurons in monkey V1 using correlated and anti-correlated dots. Because V1 cells respond to contrast, they should not ‘care’ whether the matching stimulus is white in one eye, black in the other. Model simulation predicted reverse (but equally strong) disparity tuning for correlated and anti-correlated dots

Question 28

How did this prediction compare to their actual findings?

Answer 28

Indeed Cumming & Parker 1997 found that V1 neurons showed inverse tuning for anti-correlated versus correlated dots. The strength of the modulation (difference between optimal and least optimal disparity) was 0.52

Question 29

What implication does this have for how we perceive disparity?

Answer 29

V1 neurons also encode disparity for stimuli in which no depth is perceived

Question 30

How was this research on disparity taken a step further? Describe the study and the results

Answer 30

The next question was if there were neurons somewhere that only reacted to the disparity when depth was perceived.

In the experiment of Janssen et al., 2003, monkeys were trained to discriminate between concave or convex 3D shapes, while neurons in IT were recorded
•They could discriminate these with correlated, not with anti-correlated displays
•Monkeys also perceive depth in correlated, not in anti-correlated displays

An example neuron in inferotemporal cortex is tuned to convex (near disparity) depth, not to concave depth. It shows this tuning however only for the correlated displays, not for the anti-correlated displays. Therefore IT neurons encode perceived depth, not just contrast disparity.

Question 31

How is this highlight a common theme in the computations performed in the visual cortex?

Answer 31

They are comparable to component cells or cells detecting wavelengths (neurons responding to visual features that are preceding perception) and pattern cells or cells detecting perceived colour (neurons responding to visual features that are perceived.)

Question 32

How is hierarchal processing relevant to perceiving shapes?

Answer 32

Lower level areas typically responding to simpler shapes (e.g lines) while higher up levels typically respond to more complex shapes (e.g triangle, pentagon, face)

Also the receptive fields get bigger obviously

Question 33

What brain area is tuned for the orientation of illusory contours? (e.g four circles each with a quadrant gone and a square is perceived)

Answer 33

V2: Neurons tuned for the orientation of illusory contours

pretty similar to V1 otherwise, with slightly bigger receptive fields

Question 34

In addition to achromatopsia, what other effects do people with lesions to V4 suffer from?

Answer 34

Difficulty discriminating complex shapes. (They could distinguish between different shades in a circle, bars at different orientations or moving at different speeds but not more complex shapes such as varying spirals or the rotations of a contour ‘line’ through vertical bars)

Question 35

What area in the visual system is highly tuned to shapes regardless if they are familiar or not?

Answer 35

Visual Area LOC responds more strongly to shapes (regardless of whether they are familiar or not) than to scrambled objects > LOC recognises shapes

Question 36

What brain area responds to shapes even when defined by second cue orders? And what is meant by this?

Answer 36

Sill LOC, responds even when defined by second order cues, when not defined just by an outline like contour from motion stimuli (while not responding to motion itself!)

Question 37

What area in the visual cortex is best known for recognising shapes and why?

Answer 37

IT- inferotemporal cortex
It has neurons which seems to display object selectivity; e.g more detailed drawings of a hand will cause a greater activation in certain cells than a simpler depiction of a hand

The neurons in this cortex are often tuned to very complex shapes; to find out what would activate these cells, researchers would have to show it lots of natural shapes and then break it down to find out what the simplest shape was that would activate it.

Question 38

What cells in the visual system responds specifically to faces in monkeys?

Answer 38

“face cells” in monkey inferotemporal (IT) cortex

Question 39

To what extent are these cells tuned specifically to faces?

Answer 39

Responds to various faces (human and monkey). They respond selectively to faces and not to other objects or scrambled faces, and not when essential face parts are missing

Question 40

How may these face cells be tuned differently?

Answer 40

may be tuned to angle at which face is presenting itself

Question 41

What is meant by object constancy and viewpoint invariance?

Answer 41

We can recognise objects from any position, angle, and under all sorts of different shading and lighting conditions. In all these cases, the image on the retina (the distribution of light and contrast) is highly different. Object recognition is therefore highly viewpoint invariant. (regardless of the viewpoint)

Question 42

How is viewpoint constancy tested in different brain areas?

Answer 42

Using repetition priming; when an object is shown twice in fMRI, the response in the visual cortex is usually weaker the second time.

Objects can then be shown once then shown a second time 1) the same as the first time, 2) same viewpoint but different retinal image (further away), 3) from a different viewpoint (angle) or 4) with a different object identity from the same object class (e.g armchair and lawn chair)

Activity can then be measured to discern the extent to which the second response is weaker. If it is weaker even at these different second images, the area is assumed to have viewpoint constancy.

Question 43

What (v specific) areas of the brain did these studies show were activated for the same object but no invariance? (3)

Answer 43

Lateral occipital, fusiform and frontal gyri only shows RP for physically identical objects > no invariance

Question 44

What (v specific) areas of the brain did these studies show were activated for the same object at different sizes but no viewpoint invariance? (3)

Answer 44

Right anterior and posterior fusiform gyri (and parietal) for RP for identical objects of different size but not for different viewpoints > size invariant coding, no viewpoint invariant coding

Question 45

What (v specific) areas of the brain did these studies show were activated for the same object at different viewpoints? (2)

Answer 45

Left anterior and posterior fusiform gyri show RP for identical objects at different viewpoints > viewpoint invariant coding

Question 46

What (v specific) areas of the brain did these studies show were activated for objects within the same class? (2)

Answer 46

Left inferior frontal gyri show RP for objects within the same class (D) > exemplar invariant coding

(Rough model in copy)

Question 47

What is significant about the left inferior frontal gyrus being activated for object class?

Answer 47

It is also Broca’s area (involved in speech ) so may be more to do with think of the objects name

Question 48

How can these preferences for these processes exist in either hemisphere when they code for the contralateral visual field?

Answer 48

This notion of contralateral visual field seems to die off in higher visual areas

Question 49

What does the facts that visual neurons each have restricted receptive fields imply regarding the features they encode?

Answer 49

Most neurons also encode position

Question 50

Give a simple example of how ‘position’ is used in the visual system

Answer 50

The superior Colliculus converts a retinotopic map in the superficial layers ( ‘more shallow’) into a saccade map (in the deeper layers) so that stimuli are foveated. The neurons of the retinotopic mapping in the above layers are connected to the motor neurons of the saccade vector mapping in the deeper layers.

Question 51

What would happen if you stimulated these deeper layers?

Answer 51

You would make a saccadic eye movement

Question 52

How is position relative?

Answer 52

Position has to be encoded in different reference frames. (e.g the position of an object relative to your retina is different to the position of an object relative to your hand, foot, mouth). If you look to the left, an object can be right relative you your eyes yet left relative to your body’s reference frame.

Question 53

How could you test what from or reference a receptive field’s position sits?

Answer 53

Record from neuron and see how receptive field shifts (or not) when eyes are in another position. When RF shifts (on screen), RF is in eye centred frame of reference, when it stays, it is head/ body centred (frame of reference.)

Question 54

In what cortical areas is position transformed from retinal to other reference frames? (4)

Answer 54

VIP (Ventral interparietal cortex)
LIP (Lateral interparietal cortex)
area 7a
Area 5 (PRR) is a motor area that uses different reference frames for movements

Question 55

Do these neurons retain their reaction of a particular reticular field? Explain (2)

Answer 55

Yes, however while some VIP neurons change their responses as a function of eye position (‘gain modulation’). RF stays at the same position relative to fixation, yet strength of response differs. it is stronger at the area of the reference frame rather than the area of the receptive field relative to the fovea.

Other VIP neurons have their receptive fields in head/ body centred coordinates. RF ‘moves’ relative to fixation, yet stays the same relative to screen (= head/ body)

Question 56

When measuring a range of these neurons in the VIP what was found?

Answer 56

VIP neurons show all the intermediate steps of transforming retinotopic (eye centered) RF’s into head centered RF’s.

Question 57

How has this (Transforming retinotopic RF’s into head centred RF’s) been shown outside of the brain?

Answer 57

When you train an AI network to transform retinal to head centered coordinates, you find ‘hidden units’ showing properties like gain modulation and intermediate forms (Zipser & Anderson)

Question 58

What does the transformation of retinal input to motor output require?

Answer 58

the transformation of eye-centered (retinotopic) visual information to motor vector outputs that are hand centered. This transformation requires the input about the positions of the eyes in the head, head relative to body, body relative to hand.

Question 59

Where are the cells which calculate this located

Answer 59

This is an example of how the coordinate frames of motor outputs are studied in area 5: the Parietal Reach Region (PRR).

Question 60

How are neurons encoded differently in the PRR

Answer 60

Some encode movement relative to the hand, others relative to the body, eye, hand & body or hand & eye

Look at and learn diagrams in docs

Question 61

To recap, name the brain areas involved with their functions in

1) Colour (2)
2) Motion (4)
3) Depth (2)
4) Shape (4)
5) Position (4)

Answer 61

Colour-
Wavelength (V1, V2)
Colour constancy (V4)

Motion-
Component (V1, V3, MT)
Pattern motion (MT)
optic flow (MST)
biological (STS)

Depth
Disparity (V1, V2, V3/V3A)
Perceived depth (IT)

Shape 
Orientation (V1, V2, V4)
Complex shapes (V4, LOC) 
Real world shapes (IT) (face, hands)
Invariant object coding (fusiform, frontal, inferior frontal and occipital gyri)

Position
Retinal (V1, V3, MT)
Head centred (VIP, LIP, 7a)
body centred (area 5, PRA)