L7&8 Perception in the Visual System

Question 1

Why is colour important ?

Answer 1

Colour vision aids discrimination and detection
Important for key tasks e.g. picking what to eat, visual memory, camouflage, mating rituals etc.

Question 2

What are the important terms to describe colour ?

Answer 2

Hue (H) - the quality that distinguishes red from blue
Brightness (V) - the perceived intensity of light
Saturation (S) - characterises a colour as pale or vibrant

Question 3

What is colour ?

Answer 3

A purely psychological phenomenon (entirely subjective)
Objects only appear coloured because they reflect different wavelengths of light
A property of our neural apparatus
We need to have the correct photoreceptors and neurons

Question 4

What is a metamer ?

Answer 4

A sensory stimulus that is perceptually identical to another stimulus but physically different
e.g. Newton - can’t discriminate between an orange light and a red and yellow light mixed together

Question 5

What is colour coding in the retina ?

Answer 5

Physically different wavelengths can result in identical colour experience
Photoreceptors are the first stage in colour coding

Question 6

What are the cones do we have ?

Answer 6

S cones - short, blue (420nm)
M cones - medium, green (530nm)
L cones - long, red (565nm)

Question 7

What is the principle of univariance ?

Answer 7

An infinite set of different wavelength intensity combinations can elicit the same response as a single photoreceptor

Question 8

How do we see more colours ?

Answer 8

We need a comparison of signals from two or more cone classes, each with a unique spectral sensitivity
Wavelength discrimination improves with the number of cone classes
Humans are trichromatic

Question 9

What is retinal topography ?

Answer 9

Cone mosaic
Layout of cone cells on the retina, colour coded for pigment
There are fewer S cones
There are no S cones in the fovea
Randomly distributed, clumping is common
No two humans have the same cone layout

Question 10

What is opponent coding theory ?

Answer 10

Colours are grouped into opposing pairs (blue/yellow, red/green)
This produces afterimages - adapting to one colour produces its opposite in the after image
Purely a result of our physiology

Question 11

What is the physiology of opponency ?

Answer 11

Parvocellular RGC have chromatically opponent RFs
The centre may be excited by red light while the inhibitory surround is excited by green light

Question 12

What is colour turning in the LGN ?

Answer 12

Layers 1 and 2 get their input rom M RGCs - input for achromatic luminance channel
Layers 3-6 get theirs from P RGCs - input for the two chromatic channels called cardinals

Question 13

What is colour turning in the visual cortex ?

Answer 13

Cortical cells show a preference for a wide range of hues, not just cardinals
Tuning width remains fairly consistent across cortical areas
Some cortical cells have double opponent RFs, the centre is excited by red and inhibited by green, while the surround is the opposite

Question 14

What is colour constancy ?

Answer 14

The ability to assign a fixed colour to an object even though the actual spectral information entering the eye changes due to illumination
The change in perceived colour is called chromatic induction

Question 15

What are colour vision disorders ?

Answer 15

Deficiency can be congenital or acquired
Acquired cerebral achromatopsia is typically due to damage to V4
Congenital is an X-linked recessive gene - usually affects M or L cones (green and red gets confused), if S is missing then blue is hard to distinguish
XY - 8% chance of colour blindness
XX - 0.8% chance of colour blindness
People affected have normal cone numbers but fewer photopigments available

Question 16

What is a dichromat ?

Answer 16

Lacking a pigment
Trichromat has all three pigments just one is defective

Question 17

How do we detect colour vision disorders ?

Answer 17

Ishihara colour plates
The images differ in hue
The combination of colours that are visible will indicate the type of deficiency

Question 18

What is many-to-one-mapping ?

Answer 18

There are many different objects that occupy the same cognitive category e.g. chairs
This also applies to poses, distance, lighting, position, viewpoint
This is needed to cope with degraded images

Question 19

Where does object recognition take place ?

Answer 19

The two stream model
Ventral stream - the what pathway (from V1, V2, V4 and IT cortex) associated with object recognition and memory
Dorsal stream - the where pathway (from V1 to V2 then V6 and V5)
associated with motion, location and saccadic control
Complex networks link to the frontal lobe

Question 20

What is object agnosia ?

Answer 20

Damage to the ventral stream can create a deficiency in object recognition
We can use other senses

Question 21

What are the three levels of analysis ?

Answer 21

Computational - what machinery is required
Algorithmic - what processes and sequences are there
Implementational - how does this produce a recognised object

Question 22

What is the template matching model ?

Answer 22

The simplest model for object recognition
When an object appears in the RF of this detector that matches the template it signals
For this to work, we would need a detector for every single possible orientation
The computer vision equivalent works well but they must be in the exact location

Question 23

What is the feature detection model ?

Answer 23

Selfridges Pandemonium model
1. The feature demons look at the image and simply write down how many examples of their feature they see e.g. horizontal lines
2. The cognitive demons shout if they think that combination of features applies to their letter , the more confident, the louder
3. The decision demons listen to the cognitive demons and decide who is shouting the loudest, providing that as the perceived letter

Question 24

What is the structural description model ?

Answer 24

Marr and Nishihara
The goal of the model is to describe the object unambiguity
The system must be invariant to transformations in viewpoint, illumination etc.
The system knows which properties are invariant under transformation and how other properties might vary

Question 25

What are the 3 important factors of the structural description models ?

Answer 25

Should the coordinate system be viewer-centred or object-centred
What are its primitives
How is that information organised into an object description ?

Question 26

What does Biedermann (1987) say about the structural description model ?

Answer 26

Primitive volumes to which objects are decomposed (not just cylinders)
Geons 1,296 pairs of geons

Question 27

What are the pros and cons of the structural description model ?

Answer 27

Pros - invariance is well explained, recognition relies on description, graded representations cope with discrimination and generalisation
Cons - extracting model parameters can be hard, structural description is difficult for some objects, driven by theoretical desirability rather than behavioural or physiological evidence

Question 28

What is the view-dependant model ?

Answer 28

Viewer based coordination system
There are abstract features that might consist of lines, curves, texture etc.
Feature sensitive units combine into each other in a weighted way
Size and position invariant
These feed into view tuned object recognition cells
Recognition by matching input to the closest stored view
Human recognition is not perfectly viewpoint invariant

Question 29

What are the pros and cons of the view dependant model ?

Answer 29

Pros - straightforward, minimises transformations that must be performed, newer models are based directly on what we know, good behavioural, physiological and simulation based evidence
Cons - humans often show quite good generalisation across viewpoints even for novel objects, still more memory intensive than e.g. geon model

Question 30

Why is face perception interesting ?

Answer 30

Faces are uniquely rich in information
There are practical applications

Question 31

What sort of object is a face ?

Answer 31

One with extremely similar distractions - it is within class, not between class recognition
Changeable - head movements, viewpoint, shape, aging etc

Question 32

What is pareidolia ?

Answer 32

When we see faces that aren’t even there, we are very good at facial recognition

Question 33

Where are faces processed ?

Answer 33

The fusiform face area (FFA)
Neural signalling in monkey was the highest for faces

Question 34

What is the domain specificity hypothesis for facial recognition ?

Answer 34

Facial recognition is innate
Faces are special
Some people are bad at recognising faces
Inversion makes it harder
Holistic processing

Question 35

What are the 4 pieces of evidence for the domain specificity hypothesis ?

Answer 35

Neonatal face discrimination
Prosopagnosia - some people cant recognise faces when damage to the occipital-temporal regions
The inversion effect - it is hard to see faces upside down
Sensitivity to facial configuration - change spacing between facial features (holistic processing

Question 36

What is the expertise hypothesis ?

Answer 36

Faces are not special
Familiarity is important
Parts of an object are also recognised better in their original environment
FFA supports general expertise

Question 37

What is the evidence for the expertise hypothesis ?

Answer 37

The effect of unfamiliarity - so much better at identifying people we’ve seen
The other race effect - sensitivity to differences between faces seems to require specific experience
Object inversion experts - worse of recognising faces upside down, but dog experts were worse at recognising dogs inverted
The part whole effect - parts are often recognised better in their original context, not just faces