Object Recognition Flashcards
THE 2 VISUAL SYSTEM PATHWAYS
- VENTRAL
- what aka. vision for perception
- occipital/temporal cortex - DORSAL
- where aka. vision for action
- parietal lobe
POHL (1973)
- what VS where in monkey visual cortex
A - specific pairs of objects predict food award
- lesions to inferotemporal/ventral cortex impair object recognition aka. WHAT
B - proximity of cylinder to foodwell predicts reward
- lesions to parietal/dorsal cortex impair spatial recognition aka. WHERE
WHAT VS WHERE: HUMAN NEUROPSYCHOLOGICAL EVIDENCE
OCCIPITOTEMPORAL/VENTRAL CORTEX LESIONS
- visual agnosia
- object recognition deficit
PARIETAL/DORSAL CORTEX LESIONS
- hemispatial neglect
- spatial awareness deficits
WHAT VS WHERE: KOHLER ET AL. (1995) PROCEDURE
TASK 1
- pps presented w/2 displays; judged if object locations were the same in both (aka. spatial locations)
TASK 2
- pps were again presented w/2 displays; had to judge if all pics = same objects in both (aka. object identities)
WHAT VS WHERE: KOHLER ET AL. (1995) RESULTS
- object > spatial = ventral activation (primarily fusiform gyrus)
- spatial > object = dorsal activation (inferior parietal cortex)
KARNATH ET AL. (2009): PROCEDURE
- effects of occipitotemporal/ventral visual cortex lesion on vision for action/perception
- tested a patient (JS) w/circumscribed lesion to ventral occipitotemporal cortex on 2 tasks:
1) required perceptual judgement
2) required motor action - performance compared against non-lesion controls
KARNATH ET AL. (2009): RESULTS
- perception task = impaired; BUT motor task = normal performance
- aka. ventral lesions impair vision for perception BUT not vision for action; suggests ventral/dorsal distinction may be more along such lines
- patient DF (dif study) = similar pattern w/lesser circumscribed lesion
VISUAL AGNOSIA
- impairment in visual perception BUT above lvl of basic sensory deficit (ie. visual field defect)
- patient cannot recognise/copy/match/discriminate simple visual stimuli OR recognise simple shapes (ie. triangles/circles)
- inability to group/integrate objects into whole
- shape processing deficit
ASSOCIATIVE AGNOSIA
- basic perception = fine BUT recognition cannot take place
- patient can make good object copies BUT cannot recognise them
- not due to language deficit (ie. anomia)
- anomia = describe dog pic; say it barks/is a pet
- associative agnosia = would not know if the dog is animate/inanimate object
- inability to associate items w/functions
TEUBER (1968) - “a normal perception stripped of its meaning”
ASSOCIATIVE AGNOSIA: MATCHING BY FUNCTIONS TASK
- patients asked to match 2 items most closely related by functions
- patients w/associative agnosia will choose 2 most visually similar items indicating that they’re unable to retrieve functions associated w/the objects
PROSOPAGNOSIA
- selective deficit in facial recognition
AGNOSIA & OBJECT RECOGNITION: MODULAR APPROACH
- object recognition = modular
- separate brain systems for dif processes
- dif agnosia types broadly highlight distinction between:
1) perceptual processing (shape analysis); impaired in visual form agnosia
2) semantic processing (associated knowledge activation (ie. function))
AGNOSIA & OBJECT RECOGNITON: CONSTRUCTIVE APPROACH
- object recognition = constructive process
- brain constructs representations of objects based on many dif sources of contextual info
- these representations (not simply retinal input) are what we’re consciously aware of
AGNOSIA & OBJECT RECOGNITON: SEMANTIC APPROACH
- object recognition = semantic process
- info about object meaning = automatically processed when we see it (ie. its function)
VISUAL PATHWAYS: RETINA -> PRIMARY VISUAL CORTEX (V1)
1) nasal retina
2) temporal retina
3) optic nerve
4) optic chiasma
5) lateral geniculate nucleus (LGN)
6) primary visual cortex (V1)
PRIMARY VISUAL CORTEX (V1)
- structured retinotopically
-dif visual field regions (IRL) = perceived in dif V1 regions - correspondence (mapping) between spatial structure of primary visual cortex/spatial structure of IRL
- cortical magnification: disproportionately large V1 area is dedicated to visual field centre (corresponding to eye’s fovea)
PROCESSING COMPLEXITY: V1 -> EXTRASTRIATE CORTEX
- processing complexity ^ V1 -> extrastriate cortex:
V1) neurons sensitive to simple visual features (ie. line orientation/spatial frequency/colour)
LO) begin to see sensitivity of more complex features (ie. geometric shape)
V5) plays key role in motion processing
KOURTZI & KANWISHER (2001): PROCEDURE
LATERAL OCCIPITAL COMPLEX (LOC)
- specialised brain region; integrates features into shapes according to fMRI
- subjects viewed 3 dif types of object (familiar/novel/non-objects)
- 2/3 required feature integration into shapes
- other objects = just disjointed features collection
KOURTZI & KANWISHER (2001): RESULTS
- lateral occipital complex activation (part of ventral processing stream) = ^ for familiar/novel objects > scrambled non-objects
- aka. lateral occipital complex plays role in integrating feature into whole shapes
LATERL OCCIPITAL COMPLEX (LOC): PROPERTIES
- largely non-retinotopic area activated by both contralateral/ipsilateral visual fields
- LOC = not simply sensitive to retinal input
- seems to encode higher-lvl shape representations even when not defined purely by retinal input
MENDOLA ET AL. (1999) - fMRI evidence that LOC responds to shapes defined by illusory contours
FMRI ADAPTATION
- used to map dif component processes of object recognition onto dif regions in ventral visual pathway
- can investigate sensitivity of neuron groups to dif object properties
- ie. a neuron sensitive to object identity “knows” pics of car from dif angles = all same object as it reduces its response w/repeated presentations
- reduction occurs even when repetition occurs from dif POVs aka. neuron = insensitive to POV
VUILLEUMIER ET AL. (2002): PROCEDURE
- fMRI adaptation; extent to which activation decreases w/repetitions
- when you present same object/word x2 -> activation tends to decrease as neurons “adapt” responses to object via varying dif stimuli properties you can access extent to which brain region processes the property
- eg. dif frequency tones; you may present via dif lengths/frequencies; neurons that process frequency will adapt response to tones of same frequency even if length = dif
VUILLEUMIER ET AL. (2002): RESULTS
- found reduced activation in left fusiform cortex to same object from dif POV relative to when dif objects were presented
- aka. region treats top 2 pics as the same even if presented from dif angles; indicates region may play role in object constancy (enabling us to recognise object under multiple dif contexts/POVs)
YEE ET AL. (2010): PROCEDURE
- presented pairs of words that had similar:
1) function (flashlight-lantern)
2) shape (marble-grape)
3) function & shape (pencil-pen)
4) unrelated (saucer-needle)
5) identical (drill-drill)
6) manipulation (defined as type of movement you make when manipulating it)
YEE ET AL. (2010): RESULTS
- several brain areas showed adaptation to function esp. in medial temporal lobe (ventral stream)
- suggests that these neurons actually represent object function aka. objects can be represented in these regions via their functions explaining how neuron loss can result in associative agnosia
VENTRAL VISUAL PATHWAY: SEPARATE MODULES
LATERAL OCCIPITAL CORTEX
- for integrating features into shapes
LEFT FUSIFORM CORTEX
- for viewpoint invariant representation of objects
LEFT MEDIAL TEMPORAL CORTEX
- for representation of objects’ functions
SUMMARY
- 2 visual systems (distinction supported by evidence):
1) ventral (what/perception)
2) dorsal (where/action) - agnosia:
1) visual form VS associative agnosia
2) implications for object recognition system - mapping function -> structure in ventral visual pathway; anatomically separate brain regions mediate dif object recognition aspects:
1) lateral occipital cortex (shape)
2) left fusiform cortex (object constancy (POV invariant representation))
3) left temporal cortex (function)
