Visual perception Flashcards
How much of the cortex is concerned with vision?
25%
How much of the cortex is concerned with vision?
25%
Visible spectrum for humans
380-760nm
Retina
Photosensitive - converts light to electrical signals
Pupil
2-8mm changes 16x. Can detect anything from a few photons to bright sunlight
Lens
Accommodation, myopia and hypermetropia
Rods
120 million, contain rhodopsin on outer segments which isomerises and breaks down into retinene and opsin. Scotopic vision.
Cones
6-7 million, packed in fovea, size = 2 wavelengths of red light. Concerned with colour vision. 3 types: S, M, L (blue, green, red). Photopic vision.
Dark and light adaptation
Reflects photoreceptors - division of labour demonstrated by special sensitivity, purkinje shift (less sensitive to long wavelengths (red) when dark adapted)
What do photoreceptors synapse with?
Ganglion cell -> bipolar cell. Lateral connections via horizontal and amacrine cells.
Ganglion cells
Project via optic chasm to lateral geniculate nucleus and then to area 17 of occipital lobe
Experiment of hecht, pirenne and schlaer 1942: state of the subject
Dark adapted - 50% chance of rhodopsin molecule regenerating in 5 mins. Half life of recover is 5 mins. Virtually complete after 40 mins.
Experiment of hecht, pirenne and schlaer 1942: location of test flash
20 degrees to left of fixation point i.e. 20 degrees to right of fovea
Experiment of hecht, pirenne and schlaer 1942: size of test flash
10’ spot diameter. 300 rods connected to a single nerve fibre
Experiment of hecht, pirenne and schlaer 1942: colour of test flash
Wavelength of 510n (peak spectral sensitivity of rods)
Experiment of hecht, pirenne and schlaer 1942
Results 90 quanta is lowest light threshold, 3% reflected away at cornea, 50% absorbed by ocular media. only 20% are collected by rods. Hence 9/10 quanta sufficient. A typical torch radiates 2 x 10^15 per ms. Therefore, if 10 rods are activates, a single rod can be activated by a singe quantum
Experiment of hecht, pirenne and scholar 1942: subject variability
1200 million pigment molecules - only 10 have to change state - spontaneous isomerisation.
Techniques of studying infant vision
Forced-choice preferential looking - dis: no preference but still discriminate, adv - indicates what is important to infant
Infant controlled habituation/novelty
Physiological measures e.g. heart rate an Visual Evoked Potentials (VEPs)
Basic abilities of infant vision
Visual acuity, Contrast sensitivity
What limits poor vision in infants?
Accommodation is 1/3 adult level at 2 weeks old (Bookman, 1980). Eye is 1/2 size but good optics. Must be result of poor acuity - imposed by immaryurity of photoreceptors - 20fold increase in foveal cone density between infancy and adulthood
Colour vision in babies
Spectral sensitivity of babies similar to adults (Dobson, 1976)
Perception of pattern and form in babies
Preferences found at birth e.g. moving, high contrast, curved contours (Frantz and Miranda 1975) - prefer complexity and symmetry
Respond to external surround of a pattern (milewiski 1976)
Form discrimination present at birth - Slater, Morison and Roase 1983
Face perception in babies
3D, mobile, high contrast and regulate behaviour continent on baby’s activities.
Depth perception in babies - Visual Cliff of Gibson and Walk 1960
Infants 6-14 mo. Concluded depth vision is innate.
Depth perception in babies - Held and Hein 1976
Used newborn cats to show importance of experience of touch for perceptual development
Face perception in babies - (Goren, Sarty and Wu 1975)
Infants 9mins old tracked face-like stimuli further than 2 others or unnatural arrangements of the same features
Face perception in babies (Field, Greenberg, Cohen 1982)
Neonates (3 days) habituated to happy, sad and surprised expressions made by a real model - looked more at mouth for happy/sad and mouth and eyes for surprised. May have been performed on local cues BUT observers can guess which expression babies had seen - produced matching expressions
Face perception in babies (Meltzoff and Moore 1983)
3 day old infants, independent observer determined which expression presented on non-emotional expressions - evidence of early facial expression matching and discrimination
Illusions explanation
The percept is different from the retinal image - inadequacy of a sensation based on visual perception e.g. ambiguous figures: Frazer’s spiral, Parthenon, vase/face figure, duck/hawk figure and necker cube
Have to interpret retinal image
Principles containing how an image is organised - Gestalt psychologists and law of pragnanz
(Koffka, Kohler, Wertheimer) - grouping principles operated - on the basis of these principles, proposed law of pragnanz - the geometrical organisation that will occur posses the best, simplest and most stable same. - existence of field forces
Why does the visual system have to interpret images?
Ambiguous figures could be said to be in the real world. One source of ambiguity is because the world is 3D. Retinal images are 2D - if depth cues reduced, percept can be ambiguous.
Size constancy
Size of an object = visual angle that it subtends at the eye.
Retinal image = smaller for distant objects, but they still appear to be the appropriate size - we use cues to depth to automatically correct the retinal image size for distance
Emmert’s law
If 2 images of identical retinal size are perceived to be at different distances, then the more distant will appear larger
Size visual illusions
Explanation of misapplied constancy - Gregory 1973, moon illusion. The horizon moon looks larger because the moon is perceived to be closer. Occluding surface = closer
Mechanistic visual illusions
Angle expansion
How do we classify colours?
Change brightness, saturation and hue (150) - 7 million colours
Colour wheel
Colours arranged according to perceptual similarity - organised in terms of hue = mixture of wavelengths e.g. purple is not a spectral colour - mix of blue and red
Not adequate for all colours - better described as a colour solid - colour of an object is determined by the amount he object absorbs or reflects different wavelengths. For liquids = degree to which they transit different wavelengths
Additive colour mixture
Lights of different wavelets overlap - additive = all wavelengths are present -> red and blue = magenta and red and green = yellow, green and blue = cyan. All 3 = white
Subtractive colour mixture
When pigments are mixed. Points opposite each other on colour circle are complementary colours as adding them = white
Non-Spectral purples
Points between blue and red - cant be produced by passing light through a prism. Hue can be matched by the addition of 3 others
Thomas Young 1802 and Helmholtz 1852 colour matching and trichromatic theory
WE can match any colour with the addition, in appropriate proportions of 3 coloured lights, as long as none could be matched by the addition of the other 2. Colour vision depends on 3 receptors with 3 different spectral sensitivities
Hering 1878 - phenomenology and Opponent theory
Pairing of blue/yellow and red/green = 3 mechanisms - R/G. B/Y B/W
Werner and Wooton 1979
Describe all colours by a comb of red green yellow blue, black and white. WE cant visualise reddish-greens or bluish-yellows
Physiology of of colour vision
DeValois and Devalois 1975
Trichromatic and opponent colour theory both proved. The retina contains 3 cone types with different spectral sensitivity. However, connections to cells in LGN are such that these cells show colour opponent processing.
Devalois recorded cells in the Lateral geniculate nucleus and showed opponent colour processing.
Colour deficiency
When a subject is able to match any wavelength in the spectrum with fewer than three lights
Monochromat
Needs one light - 10 people/million
Dichromat
Requires 2 lights
Protanopia
Missing red - 1/100 males and 2/10000 females
Deuteranopia
Missing green 1/100 males and 1/10000 females
Tritanopia
Missing blue 1/50000 makes and 1/100000 females
The thought experiment
Put observer in room, luminous circle, one eye, head still. Size and distance uncertain, which is expected. Observers locate object at 6-8ft
Retinal disparity
When the eye views several objects then different objects fall on non-corresponding parts of the two retina - the degree of non-correspondnace is called retinal disparity and increases with increasing relative distance between 2 objects.
Stereopsis
The mechanism enabling depth to be inferred from disparity
Stereoscopic depth constancy
Disparity between 2 objects decreases with square of distance. Yet relative depth is constant - stereoscopic depth constancy
Wheatstone 1838 - mirror stereoscope
and Brewster - refracting stereoscope
Used geometrical patterns to show that retinal disparity is a cue to dept by showing disparity alone can register depth - not pectoral info.
Refracting relies on using lenses that refract, or bend the light to make it appear that they emanate from a single source.
Jules 1971
Object recognition achieved by both eyes and then fusion takes place. Thus stereopsis is not dependent on object recognition and is a powerful cue to depth.
Oculomotor cue: accommodation
Cue for distance perception. Several objects cannot all be focussed - degree of blurring should give info about depth. Accommodation only effective for a few feet, then all objects seen equidistant. Weak cue to depth.
Oculomotor cue: convergence
The fact that the eyes converge more for nearer objects than far in order to fixate objects on the fovea. The vergence angle could be a cue to distance. Weak cue to depth.
Motion parallax - rock 1984
Displayed several discs at different distances, viewed monocularly. Subjects perceived at same distance, but moving at different speeds when they moved they heads. Both eyes = different depth planes
Linear perspective e.g. necker cube
Lines parallel in a scene converge
Aerial perpsective
More distance objects more blurred and appear to be tinged with blue because of atmosphere disturbance
Shadow
Shading that results from depth = cue to depth i.e. attached shadow. Telling the difference between depression and elevation relies on knowing light source direction. Assume light comes from above
Occlusion: interposition
A powerful cue. Not dependent on familiar shape, can override stereopsis
Familiar size: epstein 1965
Used a dime, quarter and half dollar. Viewed monocularly and illuminated in darkened room. Subjects responses about distance were as expected
Height in the field of view
Objects high in field of view = more distant, opposite true for horizon. Only true for ground objects. For it to be a cue, we must know the objects on the ground - must have perceived the ground plane, which recedes into distance
Ecological (Gibson J.J 1950) vs Constructivist approach
Constructivists sau since retinal image is ambiguous, then size cant be established without knowing dance.
Gibson says laboratory experiments are impoverished and argues that:
perception is best understood under the moving observer
perception best studies in optic array not static image
enough info in optic array to make calls necessary
i.e. perception is direct
Newcombe, Ratcliff and Damasio 1987
VRD right parietal, poor onMaze learning, good on mooney faces
JS right temporal good at maze learning, poor on mooney faces
Bay 1953 refuted by Etllinger 1956
Appears to be a separate stage in object recognition as patient cant see during visual agnosia
Underleider and mishkin 1982
Visual discrimination vs the landmark test in monkeys with temporal and parietal lesions
Visual agnosia
A disorder of recognition confined to the visual realm, in which an alert individual fails to arrive at the meaning of previously known nonverbal visual stimuli
Apperceptive agnosia: impairment in visual perception
Objects not perceived normally. can describe common objects, but cant recognise objects, numbers, face or geometrical figures. Cant discriminate shape, copy or match
Apperceptive agnosia: perceptual cateorgisation deficit
Cant match 3D objects across a change in perceptive
Associative agnosia
A patient stripped of meaning - can copy and match, but frequently make visual eros. Main difference is that they use shape info
Stimulus equivalence
Must be a store of objects or representations that aids recognition: template matching, feature descriptions (stores as a list of features) or structural descriptions
Marr 1982 structural descriptions
Proposed visual system proceeded through 4 levels of representation:
Raw primal sketch - intensity changes across retina
Full primal sketch - contours defined (apperceptive agnosia)
21/2D sketch - visible surfaces from observers PoV
3D model representation - independent of observers position
The perceptual categorisation deficit
Reflect a difficulty of assigning an object centred frame of reference
Developmental face perception studies: Goren, Sarty and Wu 1975 and Bushnell 1989
Babies less than 10 min old track faces further than jumbled faces.
Newborns discriminate their mother from a stranger
Normal observer facial recognition
Recognition of familiar face in less than 0l5s - over 90% correct in electing 50 faces from 50 distractors when each seen for 5s. - right hemisphere = face processing
Barrack et al 1975
90% correct in ID of school peers independent of class size (90-900) and number of elapsed years (3mo - 35yr)
Yin 1969 inversion and expertise
Diamond and Careye 1986
Recognition memory for faces better than line drawings of houses and stick figures disrupted by inversion.
Effect of experts in that dog breeders more adversely affected by inversion.
Faces undergo non-rigid transformations - mobile. Important for identity, age, gender judgements, mood etc.
Representations for recognition
Task of the visual systems is to detect invariance’s to determine ID regardless of pose and facial expression and detect variant info - e.g. mood regardless of ID
Endow 1982 - different head types
Dolichocephalic long/narrow - leptoprosopic vs Brachycephalic wide/short- euryproscopic
Determination of age: Shaw and Pittenger 1977
Examined non-rigid transformation of head during ageing. Showed a single transformation mapped one age to another. Showed subjects a series of profiles. 91% ages judgements based on degrees of cardiodal strain.
Determination of facial expression: beckman and fiesen 1971
Basil 1978
facial expression = universal categories but these are posed. Expressions are dynamic
By filming faces in the dark, subjects could ID the facial expression without structural info about features. -> facial action coding system. Each action unit = movement of particular far emucles.
Configurations to ID faces: Sergent 1984
Same/different photofit faces varying chin/eyes and internal space. Processed as individual features = fastest different judgment must be determined by the most salient. Faster still if something else changed. Inverted faces did not show this effect - so processed feature by feature
Configurations to ID faces: Young and Hay 1986
Measured accuracy and latencies for top/bottom when combined. Launches much slower but improved when faces misaligned
Caricatures
Exaggerate differences from ‘average’ face
Brennan 1985
186 points then multiply departure of each from norm. High number = greater caricature - better recognised
Valentine and Bruce 1986
Faces rated as typical or distinctive - recognition faster for distinctive but slower when mixed with jumbled faces and question is ‘is it a face?’
Shepherd 1981
Ability to discriminate amongst faces of a different race is poor. Try to use same dimensions of encoding which are inappropriate.
Relation between object and facial agnosia
Prosopagnosic difficult with other categories (not face)
Differential difficult if damage slight then faces affected first. But there is double association
Exemplars within a category? Maybe bad at any category with may egg
Special stimulus specific recognition system used for things difficult to discriminate. Not good explanation as things they cant do can be relegated to this system.
