From Receptor Signal to Perception Flashcards
1
Q
INTRO
A
- linking retinal filtering of edges to perception
- visual processing beyond retina in primate/human brain
- human/animal colour vision
2
Q
WHY DOES RETINA FILTER EDGES SO EARLY IN VISUAL PATHWAY?
A
RAW IMAGE PROJECTION
- receptor signals are distributed spatially in continuous/noisy patterns; threshold required to detect contrasts
PROCESSED IMAGE
- contrast edges (areas of change) contain info
3
Q
VISUAL PROCESSING: EARLY STAGES
A
- involve edge filtering & enhancement
- lateral inhibition (found in centre-surroind (CS) receptive fields of many retina ganglion cells) explain hard-wired phenomena in perception ie. Herrman grid
- we see grey dots that aren’t there; during eye movements, same part of image is viewed by foveal receptors & ganglion cells/peripheral ones
- CS receptive fields = smallest in fovea (highest spatial resolution) & larger the further in periphery of retina they are
4
Q
HUMAN SPATIAL CONTRAST SENSITIVITY (CSF) FUNCTION
A
- ganglion cells w/CS receptive fields:
1. enhance edges
2. compresses info (only respond when in receptive field)
3. filters info according to spatial frequencies (dif sized receptive fields/varying sensitivity across retina)
5
Q
OWSLEY (2016); GHIM & HODOS (2006)
A
- vision & aging
- inter/intraspecific variations in acuity/contrast sensitivity
- both humans (Owsley) & birds (Ghim & Hodos) have similar negative correlation between contrast sensitivity & spatial frequency
6
Q
CROWDED VISUAL SCENES
A
- seeing/recognising objects/mates/predators/prey = important for many tasks
- BUT visual scenes = oft crowded (typically not monochrome); contrast enhancement of edges = important for many visual tasks
- major tasks of visual system = segregating objects/backgrounds automatically/quickly
- other tasks require further computations to extract info
7
Q
EGELHAAF ET AL. (2012); KANG ET AL (2012)
A
- spatial vision in insects = fascilitated by shaping dynamics of visual input through behavioural action
- camoflauge through active choice of resting spot/body orientation in moths
- insects land preferably at edge of objects
- moths actively choose spots; vary orientation to align w/lines in background for better camoflauge against avian predators
8
Q
HATEREN ET AL. (1990)
A
- insects discriminate/generalise stripe patterns & recognise illusory contours
- bees were trained w/variations of stripe pattern; each tested group learned to discriminate particular orientation of these patterns; experiment trained bees to Kanisza rectangles
9
Q
LOOKING: PHOTORECEPTOR SIGNALS
A
- change very quickly; are noisy
- movements (eyes/head/body) change/stabilise gaze for very short periods of time; fast main function filtering decomposing images into elementary features in peripheral visual system layers
- early segregation fed into parallel visual streams for unconscious/conscious visual perception
10
Q
ACTIVELY LOOKING & SEEING: UNCONSCIOUS PERCEPTION
A
- can be fast/slow/filtered depending on task/pathway can be invariant/selective/less noisy
- guides fast/slow actions; analyses scenes/objects; changes are perceived via task outcomes (ie. consequence of beh response/change in internal state/top-down control (ie. gaze/conscious decision-making in humans))
11
Q
SEEING: CONSCIOUS VISUAL PERCEPTION
A
- in humans = stable/slow/invariant/affected by filtering BUT less selective than unconscious perception/low noise
- only pronounced eye/head/body movements result in perceived change of viewed scene/object; conscious vision seems relevant for some specific tasks; can exert some top-down control but mostly results from processes at lvl 1/2; some = hard-wired
- humans = difficult to generate evidence clearly separating domains 2/3
- animals = lvl 3 studied w/brain imaging in monkeys/modelling
12
Q
METHODOLOGICAL APPROACHES TO STUDYING SENSING/PERCEPTION
A
PSYCHOPHYSICS
NEUROANATOMY
FUNCTIONAL NEUROPHYSIOLOGY/NEUROGENETICS
THEORETICAL NEUROSCIENCE/COMPUTATIONAL MODELLING
13
Q
METH APPROACHES: PSYCHOPHYSICS
A
- links variation in stimulus w/changes in beh (ie. eye/body responses/verbal responses/task acquisition/execution)
14
Q
METH APPROACHES: NEUROANATOMY
A
- provides info about connectivity in sensory organs/brain/motor systems
15
Q
METH APPROACHS: FUNCTIONAL NEUROPHYSIOLOGY/NEUROGENETICS
A
- links neural response patterns to connectivity/behaviour
- tests concepts/algorithms
16
Q
METH APPROACHES: THEORETICAL NEUROSCIENCE/COMPUTATIONAL MODELLING
A
- proposes concepts
- tests mechanistic/functional hypotheses; formulates methematical algorithms to show how nerons do/may interact within/between brain areas
17
Q
THE CORTICAL PATHWAY
A
- cortical pathway = processes image info
- lateral geniculate nucleus (LGN) aka. thalamus
- subcortical pathways = pretectum (mediates pupillary reflex) & superior colliculus (controls saccadic eye movements)
18
Q
THALAMUS
A
- principal synaptic relay before sensory info reaches verebral cortex
- high input = 10,000 ganglion axons
- thalamus gates/modulates info flow to cortex
- LGN = parvocellular (small cell) laminae/magnocellular (large cell) laminae
19
Q
STRUCTURE/RETINOTOPIC ORGANISATION OF PRIMARY VISUAL CORTEX
A
- retinotopic = relative position of object in visual field is preserved in retina; also preserved in LGN & primary visual cortex
- BUT… mape = upside down (due to focusing on lens/cornea); aka. disproportionate; more cells = devoted to processing info from fovea aka. cortical magnification
- V1 contains large diversity of cells incl. simple/complex cells
20
Q
DETECTING OBJECTS THAT LACK CONTINUOUS EDGES
A
- early 20th century; School of Gestalt psychology proposed shape/object perception = underpinned by processes in mind that are characterised by Gestalt laws ie. proximity/similarity (ie. in colour/size)
- local cues/features = integrated across long distances in image into global features
- aka. is slow! requires lot more computations/eye scannings of whole scene to make sense of world & resolve ambiguities
21
Q
RENOIR (1885)
A
- coding of contours/boundaries analysis of objects
- shape/positional invariance
- contour enhancement required for object identification
- aka. V1 = fundamentally important for conscious vision/perception
22
Q
MISHKIN, UNGERLEIDER & MACKO (1983)
A
- what/where streams in cortical processing of visual info in primates/humans
A. MAKING SENSE OF WORLD - aka. object discrimination
- bilateral remova of area TE in inferior temporal cortex produces severe impairment on object discrimination aka. 1-trial object recognition task based on principle of non-matching to sample where monkeys familiarised first w/1 object of pair in central location then rewarded in choice test for selecting unfamiliar objects
B. INTERACTING W/WORLD - aka. landmark discrimination
- bilateral removal of posterior parietal cortex produces severe impairment on landmark discrimination
- monkeys rewarded for choosing covered foodwell positioned randomly trial to trial
23
Q
NICHOLLS ET AL. (2011)
A
- V1 = higher visual areas generate conscious percepts
- parallel processing = ganglion cells receive inpit from same photoreceptors
- serial processing = signals transmitted from photoreceptors to ganglion cells to LGN -> V1
- parietal lobe = dorsal (where) stream; association cortex -> retinal ganglion cells
- temporal lobe = ventral (what) stream
24
Q
DI CARLO ET AL. (2012)
A
- macaque monkey brain object recognition/categorisation
- discrimination (<200ms)
- recognition of objects (also when changes in object position/size/viewpoint/visual context)
- ventral cortical stream = critical for object recognition
- V1-V4 = occipital lobes
- IT = inferior temporal cortex (temporal lobes)
25
Q
HOW CAN FILTERED/DECOMPOSED ELEMENTS BE COMBINED IN BRAIN?
A
- receptive fields in interneurons of brain can be varied in many ways to combine info provided within/between visual pathways
- cells in serially connected layers directly/simply cause elements of perception
26
Q
CONCEPTS OF PERCEPTUAL INFO CODING
A
- redundancy reduction/sparse coding/selective attention keep energetic cost of processing low
1. GRANDMOTHER/GNOSTIC CELLS - ie. face-selective cells in interotemporal cortex of monkeys
- sparsely coded representation of object
2. DISTRIBUTED NETWORKS - represent objects by binding features coded in receptive fields across neural brain centres; form distributed representation of object
27
Q
HUMAN COLOUR VISION
A
- originates in P pathway
- S/M/L cones contribute to colour vision
- fovea only contains M/L cones
- rods mediate night vision aka. achromatic
- V1 = primary visual cortex
- M pathway = colour/fast/low acuity/high contrast sensitivity
- P pathway = colour sensitive/slow/high acuity/low contrast sensitivity
28
Q
“WHERE” SYSTEM AKA. MT (V5)
A
- motion perception
- depth perception
- spatial organisation
- figure/ground segregation
29
Q
“WHAT” SYSTEM AKA. V4
A
- object recognition
- face recognition
- colour perception
30
Q
COLOUR = USEFUL INFO
A
- colour vision = ability to discriminate light stimuli by spectral content (colour cue) NOT intensity (brightness cue)
- widespread in animals; number of photoreceptors/spectral sensitivity differs
- ie. bees see rockrose flowers as pink/red rather than gold aka. via primary colours
31
Q
COLOUR-BLINDNESS
A
- human colour vision polymporphism
- most mammals have also 2 spectral types of photoreceptors ie. colour-blind humans
- marine mammals = truly colour-blind; have only 1 spectral type of photoreceptors
32
Q
COLOUR CONSTANCY
A
- ability to recognise colours under dif illuminations
- sunlight = slightly colours during dusk/dawn
- visual system compensates for such slow changes by global adaptation
- during visual search in natural scene; visual system can also adapt quickly BUT within spatially restricted area of visual search
33
Q
LEHRER ET AL. (1988); KINOSHITA & ARIKAWA (2000)
A
- insects have perceptial constancy/code depth from 2D retinal image
- size constancy = honeybees can discriminate between objects based on size; motion cues provide bee’s visual world w/third dimension
- honeybees/butterfly have colour constancy; recognise rewarded colour in Mondrian stimulus under dif coloured illuminations
34
Q
SUMMARY
A
- sensory systems extract info from sensory cues to generate unconscious/conscious perception depending on task type/priorities for decision making/task execution in dif areas/systems of brain
- visual/perceptual info (motion/colour hue/saturation/brightness/edges/shape/spatial & temporal patterns) = computed from photoreceptor signals; processed in various parallel working pathways hierarchically organised (aka. serial connectivity)
- evidence generayed in research not always links perception to tasks thus studying isolated components of visual system/brain; incorporating comparative approached/ecological/evolutionary considerations (ie. by considering properties of natural visual scenes (beh & ecological needs of humans/animals) can help better understand functions/limitations of brain mechanisms
