} PSY2010
Cognition and Neuroscience
Cognition and Neuroscience
} Constructing our Visual World
} VISUAL PERCEPTION IS NOT A
PASSIVE PROCESS …
PASSIVE PROCESS …
} i.e. we do not simply see what is there …
} Our brains CONSTRUCT part of what we see,
based on what we expect to see
based on what we expect to see
} Perception based on external visual stimulià BOTTOM UP PROCESSING
} Construction based on knowledge/expectationsà TOP DOWN PROCESSING
} Visual perception incorporates BOTH
} 2 components of perception …
} Another EG of construction …The Blind Spot
}
Important structures in the eye …
Important structures in the eye …
} The primary visual pathway …
} Each eye contains a left and a right visual field
} Information from both fields travels to the optic chiasm
} After this crossover, information from the right visual field travels to LEFT visual cortex, while info from left visual field travels to RIGHT visual cortex
} The primary visual pathway …
} Primary visual cortex is in the occipital lobe
} The dominant pathway
(in humans) for visual information passes
through the LATERAL GENICULATE NUCLEUS of the thalamus, and then to
primary visual cortex
(in humans) for visual information passes
through the LATERAL GENICULATE NUCLEUS of the thalamus, and then to
primary visual cortex
} Known as GENICULOSTRIATE pathway
} The primary visual pathway …
} Geniculostriate pathway contains 90% of the axons in the optic nerve
} Other 10% of fibres ennervate other pathways – these also important
} EG Pathway via superior colliculus very important role in visual attention
} The lateral geniculate nucleus …
} LGN
} Visual field information now lateralized
◦ R LGN has left visual field info only
◦ L LGN has right visual field info only
} Each LGN has 6 layers
◦ 4 upper layers = P layers
– Parvocellular (small cell bodies)
– Most responsive to color; smaller detail
◦ 2 lower layers = M layers
– Magnocellular (large cell bodies)
– Most responsive to movement; larger areas of visual field
} Lateral Geniculate Nucleus cont …
} Most neurons in LGN have a center-surround receptive field
} Receptive field = region of space that elicits response from a particular neuron
} Don’t respond to light per se
} They respond to differences in light across their receptive field (e.g. presence of light in centre, absence in surround)
} Lateral Geniculate Nucleus cont …
} Light falling across the entire receptive field would à NO RESPONSE, as the center and surround fields
inhibit each other
inhibit each other
} Parallel processing in Visual System
} Visual perception understood as analytic
} Visual information distributed across
distinct subsystems
distinct subsystems
} Early processes analyze attributes
} Eg shape, color, movement each
processed separately
processed separately
} Primary visual cortex – V1
} What do visual neurons need to do?
} Represent how light/dark something is
} Represent color of an object
} Detect edges (abrupt changes in color/brightness) as these help separate out visual objects
} Detect movement
} Represent depth
} CELLS IN V1 TRANSFORM INFO FROM LGN à
BASIC CODE THAT ENABLES ALL THIS INFO TO BE EXTRACTED AT LATER STAGES OF PROCESSING
BASIC CODE THAT ENABLES ALL THIS INFO TO BE EXTRACTED AT LATER STAGES OF PROCESSING
} Primary visual cortex – V1cont …
} Hubel & Wiesel got Nobel Prize for research on responses of single cells in visual cortex
} Center-surround receptive fields of cells in LGN ideally suited to signalling CHANGES in illumination – like those that arise from stimulus edges, but these cells responsive to points of light
} Similarly, cortical visual neurons respond to EDGES
} FUNDAMENTAL PRINCIPLE OF PERCEPTION:
} CNS most interested in change
} Recognize elephant not by homogenous grey surface, but by contrast between grey edge of its shape against
a background
a background
} Primary visual cortex – V1cont …
STRUCTURE OF THE SYSTEM
} Several LGN cells will synapse with single
cortical neuron
cortical neuron
} The neuron called SIMPLE CELL
} This neuron also has center-surround receptive field, but optimal stimulus = EDGE
} Combines responses from several center-surround receptive fields in LGN
} Some also respond to wavelength (thus involved in color perception)
} Primary visual cortex – V1cont …
} SIMPLE CELLS
Their simple organization extracts a fundamental feature for shape perception – the border of an object
Their simple organization extracts a fundamental feature for shape perception – the border of an object
} COMPLEX CELLS
Again using the principle of combination, these cells integrate information from several simple cells to detect more complex features, eg corners or edge terminations
Again using the principle of combination, these cells integrate information from several simple cells to detect more complex features, eg corners or edge terminations
} Primary visual cortex – V1cont …
} COMPLEX CELLS
} Combination of simple cells
} Also respond to particular orientations
} BUT have larger receptive fields
AND must be stimulated across their entire length
AND must be stimulated across their entire length
} Can detect corners; edge terminations
} NB Terminology
} Striate cortex refers to V1, or primary visual cortex
} = 1st processing region in cortex
} Extrastriate cortex refers to visual association areas,
outside of V1
outside of V1
} Here numbers should not be interpreted as indicating sequential processing
} Primary Visual Cortex (V1) – Summary
} Extracts basic information from the visual scene (e.g. edges, orientations, wavelength of light)
} This information is used by later stages of processing to extract information about shape, colour, movement, etc.
} Single-cell recordings by Hubel and Wiesel lead to a hierarchical view of vision in which simple visual features (e.g. points of light) are combined into more complex ones (e.g. adjacent points of light may combine into a line)
} PSY2010
Cognition and Neuroscience
Cognition and Neuroscience
} Constructing our Visual World
} Primary Visual Cortex (V1) – Summary
} Extracts basic information from the visual scene (e.g. edges, orientations, wavelength of light)
} This information is used by later stages of processing to extract information about shape, colour, movement, etc.
} Single-cell recordings by Hubel and Wiesel lead to a hierarchical view of vision in which simple visual features (e.g. points of light) are combined into more complex ones (e.g. adjacent points of light may combine into a line)
} V1 & beyond …Cortical visual areas
} Spatial Arrangement of Primary Visual Cortex (V1)
} Retinotopic organization – the spatial arrangement of light on the retina is retained in the response properties of V1 neurons (except inverted)
} Damage to parts of area V1 results in blindness for the corresponding region of space
(e.g. hemianopia)
(e.g. hemianopia)
} Spatial Arrangement of other visual areas
} Each visual area has topographic representation of external space
} As one area projects to another,
retinotopic arrangement is maintained
retinotopic arrangement is maintained
} These multiple retinotopic maps
retain precise spatial information
retain precise spatial information
} Why has the primate brain evolved so many visual areas?
} EARLY IDEA
} HIERARCHICAL PROCESSING
– Each area elaborates on processing done in
previous area
previous area
– Simple cells calculate edges
– Complex cells use these to calculate corners &
edge terminations
edge terminations
– These used by higher neurons to represent shapes
– BUT if you look back at the diagram, see there is no simple, sequential, hierarchical pathway in later/
higher visual areas
higher visual areas
} Why has the primate brain evolved so many visual areas?
} NEWER IDEA
} ANALYTIC PROCESSING – DIVIDE & CONQUER
– Each area provides a map of external space
– BUT maps differ ito the information they represent
– EG neurons in V4 highly sensitive to color
– Neurons in V5 highly sensitive to movement
– Each area thus provides its own limited analysis of particular information
– PROCESSING IS DISTRIBUTED & SPECIALIZED
} Recognizable percepts achieved at
higher areas
higher areas
} Area V4 and Area V5/MT
Note: V5 often called MT (for middle temporal) because that’s where its found in primates; location is not same in humans
} Colour Perception and Area V4
} Why does the brain need a specialized colour centre given that the retina is sensitive to different wavelengths of light?
– The problem is that wavelength depends on the composition of the light source (e.g. daylight, electric light) as well as the colour of an object
– Area V4 tries to compute the colour of the object taking into account variations in lighting conditions
– This is called color constancy
} Color constancy
} Color Perception & Area V4
} Cells in V4 continue to respond to the same surface colour if the light source is changed, whereas cells in V1 do not
} Patients with damage to area V4 see the world in black and white – they are called achromatopsic (not to be confused with colour blindness due to cone deficiency)
} Although achromatopsic patients fail to see colour, their retina and their V1 cells still respond to different wavelengths of light
} This is another example of how visual perception is constructed, rather than the mere detection of physical properties in the environment
} Achromatopsia
} How do people with this disorder see the world?
} Seems like in greyscale
} Some cases report hating perceptions
– Tomato juice seemed black, the dog was grey, human skin “an abhorrent grey” – patient couldn’t look in
mirror or at other people without revulsion
mirror or at other people without revulsion
– “..patients with achromatopsia see with their cones and wavelength-sensitive cells in V1, they have access to color and wavelength; however …their visual perception of color and wavelength is that of V1, the output of which never normally appears in our awareness; and the sensation is prechromatic, neither colored, nor colorless” (Devinsky & D’Esposito p.138)
} Imaging V4 & color perception
} Zeki (1993) PET study
} Subtractive logic: activation in control task factored out from activation in experimental task
– Control task: view collage of achromatic rectangles
– Experimental task: each grey rectangle replaced
by color exactly matched in luminance
by color exactly matched in luminance
– Neurons sensitive to luminance should thus be =lly active in both conditions
– Neurons sensitive to chromatic information should be more active in the experimental condition
} à activation in V4
} Movement Perception & Area V5/MT
} Cells in V5/MT do not respond to color
but 90% of them respond to particular directions of movement
but 90% of them respond to particular directions of movement
} Patients with bilateral damage to this region see the world in a series of still frames
} They are said to be akinetopsic
} The patients can detect movement in other senses (e.g. hearing, touch)
} Color and form perception intact
} Akinetopsia
} How do patients with this disorder see the world?
} Seems its like seeing a series of still frames; or seeing things through a strobe light
} See series of static images
} Patient describes overfilling cup; being afraid to cross the road as a moving car seems far, then right on top of her
} VERY difficult to direct your own movements if you cannot perceive movement of other components of your environment
} Imaging V5 & perception of movement
} Zeki (1993) PET study
} Control stimulus: complex collage of black and white squares
} Experimental stimulus: same collage, BUT squares set in motion
} à Activation in human MT/V5 –
bilaterally near the junction of temporal,
parietal and occipital cortices
bilaterally near the junction of temporal,
parietal and occipital cortices
} More Imaging V5 & perception of movement
} The Enigma pattern
} Illusion of motion
} PET and fMRI studies both show viewing visual displays like this one à marked activation in V5
} V1 is silent during illusory movement
} Color & Movement Perception
} Clinical literature – (ie lesion studies) indicate double dissociation
– Color perception can be impaired while perception of movement is intact
– Perception of movement can be impaired while color perception is intact
} This supported by functional imaging studies – activity in regions implicated by lesion studies is associated with particular types
of perception indicated in clinical population
of perception indicated in clinical population
} CONVERGENT EVIDENCE RE FUNCTIONAL
SPECIALIZATION
SPECIALIZATION
} Subcortical visual pathways
Pathway involving SUPERIOR COLLICULUS
} Midbrain structure
} Critical role in producing eye movements
(ie orienting to a stimulus)
(ie orienting to a stimulus)
} NB role in spatial orientation
} Schneider (1969) showed that cortically blind hamsters were able to locate/orient to
sunflower seeds
sunflower seeds
} Subcortical visual pathways
} Weiskrantz (1986) reported on patient DB – wanted to see if similar ability present
– Stimuli presented in blind hemifield
– Asked to move his eyes à stimulus (ie orient)
– Buzzer indicated should move eyes
– Control trials – no stimulus in blind field
– Experimental trials – stimulus in blind field
} Performance random on control trials
} BUT able to orient on experimental trials
} Blindsight
} Damage to V1 leads to a clinical diagnosis of blindness (the patient cannot consciously report objects presented in this region of space)
} However, the patient is still able to make some visual discriminations in the "blind" area (e.g. orientation, movement direction) – called blindsight
} This is because there are other routes from the eye to the brain
} The geniculostriate route may be specialized for conscious vision but other routes act unconsciously
} Subcortical pathways
} Patients with blindsight can localize stimuli they are not aware of
} Ability to orient fits known function of retino-collicular pathway
} Seems possible that subcortical pathway collaborates with conscious, cortical
geniculostriate pathway
geniculostriate pathway
} Detects object on periphery à generates eye movement that brings object to center of view so that cortical system can analyze and identify
consciously
consciously
¨ Cognition & neuroscience
Higher visual perceptual functions
¨ Recognizing objects
¨ Product of perception STRONGLY
interwoven with memory
interwoven with memory
¨ Consider the computational problems inherent in
a system that processes sensory information
AND then links this information to memory
a system that processes sensory information
AND then links this information to memory
¨ Perception and recognition are NOT a single, unitary process
¨ Are there separate memory stores for each sensory modality, or is there a modality-independent knowledge base?
¨ Are different kinds of objects processed differently?
¨ A Model of Object Recognition
¨ Four broad stages…
(1)Early visual processing (colour, motion, edges, etc.)
(2)Grouping of visual elements (Gestalt principles, figure–ground segmentation)
(3)Matching grouped visual description onto a representation of the object stored in the brain
(called structural descriptions)
(called structural descriptions)
(4)Attaching meaning to the object (retrieved from semantic memory)
The first stage was considered in previous lectures…
¨ Stage 2: Combining Parts into Wholes - Gestalt Grouping
(a) Law of proximity, (b) law of similarity,
(c) law of good continuation, (d) law of closure
¨ A Model
of Object Recognition
of Object Recognition
¨ Adapted from
¨ Humphreys and
¨ Riddoch, 2001
¨ Agnosia – failure to recognize
¨ Agnosia means ‘no knowledge’
¨ It means the patient fails to recognize objects
¨ We will be talking about VISUAL agnosias
¨ CASE
¨ GS – stroke
¨ Visual acuity intact – could see shape, form, could draw objects
¨ HOWEVER could not recognize what they were from purely visual information
¨ Agnosias cont …
¨ COULD name objects if they were described
in words
in words
¨ COULD recognize objects through other modalities (ie if allowed to touch or hear)
¨ HAD A MODALITY SPECIFIC DEFICIT
IN RECOGNITION
IN RECOGNITION
¨ Seems unable to link conceptual knowledge with visual percept, and use this knowledge to ID object
¨ Various different kinds of visual agnosia exist
¨ By studying these, parse out stages/processes involved in object recognition
¨ ‘What’ & ‘Where’ pathways
¨ So what happens to visual information outside of the visual areas we talked about before?
¨ As mentioned, complex convergence and divergence of pathways
¨ Makes researching this very difficult
¨ Know that information from occipital cortex moves to higher heteromodal association areas
via 2 bundles of fibers
via 2 bundles of fibers
¨ 1 takes a ventral route à temporal lobe (‘what’)
¨ Other takes dorsal path à parietal lobe (‘where’)
¨ ‘What’ & ‘Where’ pathways
¨ ‘What’ & ‘Where’ pathways
¨ WHAT and WHERE are the 2 basic questions the visual system has to answer
¨ ‘What’ pathway
ú Ventral
ú Occipito-temporal pathway
ú Specialized for object perception and recognition –
tells us what we are looking at
tells us what we are looking at
¨ ‘Where’ pathway
ú Dorsal
ú Occipito-parietal
ú Specialized for spatial perception, determining where an object is, and the spatial relationships between objects in a scene
¨ ‘What’ & ‘Where’ pathways
¨ Differences in receptive fields/neurons evident
¨ Both have large receptive fields
¨ Parietal neurons respond in non-selective way:
ú May activate in response to small or
large stimuli
large stimuli
ú Activate in response to stimuli that occur
in periphery of visual field
in periphery of visual field
ú Thus (like superior colliculus) ideal for detecting presence of stimulus, esp one
that has just entered visual field
that has just entered visual field
¨ ‘What’ & ‘Where’ pathways
¨ Temporal neurons different
ú Receptive fields always encompass fovea
(ie center of visual attention)
(ie center of visual attention)
ú Thus ideal for object recognition
ú Tend to look directly at things we want
to identify
to identify
¨ Imaging ‘what’ and ‘where’
¨ Kohler et al (1995): PET
¨ Presented identical sets of stimuli; with identical changes in stimuli
¨ Instructions differed
¨ In the ‘where’ condition, participants had to decide if an object was in the same place as on the previous stimulus presentation
¨ In the ‘what’ condition, subjects had to decide if the object was the same as on the previous stimulus presentation
¨ Imaging ‘what’ and ‘where’
¨ Activation on the two conditions compared
¨ Similar activations across tasks (eg decision making; similar perceptual processing) would thus cancel each other out
¨ Only activation unique to each condition would stand out
¨ ‘Where’ condition à activation in R parietal
¨ ‘What’ condition à bilateral activation at junction of occipital and temporal lobes
¨ Perception for identification
vs perception for action
vs perception for action
¨ We see functional dissociations in performance of patients with agnosia
¨ DF – carbon monoxide poisoning à
bilateral occipital lesions
bilateral occipital lesions
¨ Basic visual perception intact
¨ However could not recognize objects
¨ EG Would call a cup an ashtray
¨ Could not identify objects from pictures
¨ Gave only crude descriptions of real objects
¨ Perception for identification vs perception for action
¨ Could ID the objects easily if allowed to hold them – ie recognition from tactile info intact
¨ Dissociation between ability to recognize/identify object; and act using object demonstrated
¨ Wanted to test ability to perceive orientation of
a slot
a slot
¨ Would be given a card, and asked to hold the card in the same orientation as the slot
¨ Failed
¨ However, when asked to insert card à slot, reached forward and correctly inserted
¨ DF and controls’ performance
¨ Perception for identification vs perception for action
¨ Indicates that information for action systems can be dissociated from information accessible to knowledge systems and consciousness
¨ Patients with optic ataxia have exactly the opposite problem to DF
¨ Can recognize objects, but cannot use visual information to guide actions
¨ EG misreach; grope about; eye movements are not directed appropriately
¨ Computational problems in object recognition
¨ How do we process the shape of an object when our position in relation to the object constantly varies?
¨ We are able to recognize shapes from infinite array of positions/orientations
¨ Also, recognize object regardless of scale differences due to differences in our distance from the object
¨ OBJECT CONSTANCY
¨ Object Constancy
¨ We need to be able to recognize an object regardless of our perspective; and regardless
of distance
of distance
¨ Other factors we have to accommodate for:
ú Differences in luminance
ú Occlusion – other objects in environment occlude (hide/mask) parts of an object
¨ Object Constancy
¨ Object recognition must be general enough to allow us to recognize the same object under all these various conditions
¨ It must also be specific enough to be able to pick up on small differences in form so that we can discriminate between objects that are very similar, but are not in fact the same
¨ View-dependent or view-invariant recognition?
¨ View-dependent theories of object recognition
¨ Have multiple stored representations of objects
¨ EG view of bicycle from side; another from
top; etc
top; etc
¨ Recognition simply requires matching stimulus to stored representation
¨ PROBLEM:
Heavy burden on perceptual memory
Heavy burden on perceptual memory
¨ View-dependent or view-invariant recognition?
¨ View-invariant theories
¨ EG Marr’s theory
¨ Recognition does not happen simply by analyzing stimulus properties
¨ Sensory input defines basic properties – primal sketch
¨ 2 ½ D sketch – viewer centered sketch
¨ 3D sketch – here object constancy achieved; viewer independent representation
¨ Critical is identifying the component axes of the shape – this enables moving from 2 ½ D à 3D rep
¨ How do we get from parts to wholes?
¨ We have some understanding of how the very basic features of visual stimuli are processed – orientation, edges, movement, color
¨ How do we process the principle axis of
an object?
an object?
¨ How and where does grouping of features happen?
¨ Is there a hierarchy of processing, as suggested by Marr?
¨ Grandmother Cells and ensemble coding
¨ Certain neurons in the temporal lobe have been found to be responsive to particular objects
¨ EG some respond to hands; others to faces
¨ These cells called GNOSTIC UNITS – they signal the presence of a known stimulus
¨ Grandmother cell = neuron that fires when you see your grandmother
(like the Jennifer Anniston cell)
(like the Jennifer Anniston cell)
¨ Grandmother Cells and ensemble coding
¨ Surely this system – where single neurons represent units of specific information – wouldn’t work very well
¨ Very susceptible to error
¨ If cell dies à sudden loss of that information
¨ Cannot account for our ability to perceive novel objects
¨ Grandmother Cells and ensemble coding
¨ Alternative idea – ENSEMBLE CODING
¨ Recognition results from activation across
multiple complex feature detectors
multiple complex feature detectors
¨ To continue to grandmother example:
¨ Some neurons respond to face shape; others to hair color; others to particular markings on face; ETC
¨ This account explains why we easily confuse similar objects (both activate many of the
same neurons)
same neurons)
¨ Also explains how we recognize novel objects – they share features with familiar objects
¨ If a few cells die, the remaining neurons sufficient to carry out processing successfully
No comments:
Post a Comment