Cortical Pathways for Visual Perception

1
Cortical Pathways for Visual Perception

• Output from occipital lobe follows two major
  fiber tracts
• Superior longitudinal fasciculus to parietal lobe
• Inferior longitudinal fasciculus to temporal lobe
• Ungerleider Mishkin propose that these pathways
  form functionally distinct processing systems
• Dorsal (occipito-parietal) is the where system
  specialized for spatial analysis
• Ventral (occipito-temporal) is the what system
  specialized for object perception and recognition

2
Dorsal and Ventral Functional Pathways

• Pohl experiments reveal double dissociation
• Landmark task monkeys with bilateral parietal
  lesion have deficit, but monkeys with bilateral
  temporal lesion can learn task
• Object discrimination task monkeys with
  bilateral temporal lesion have deficit learning
  task, but monkeys with bilateral parietal lesion
  do not

3
Dorsal and Ventral Functional Pathways

• Pohl experiments reveal double dissociation
• Other bilateral lesion experiments show
  dissociation within temporal lobe
• anterior temporal lesions disrupt visual memory
• posterior temporal lesions disrupt visual
  discrimination

4
Dorsal and Ventral Functional Pathways

• Pohl experiments reveal double dissociation
• Other bilateral lesion experiments show
  dissociation within temporal lobe
• Must lesions be bilateral to cause deficits?
• Are dorsal and ventral pathways bilateral?

5
Dorsal and Ventral Functional Pathways

• Ungerleiter and Mishkin combination lesion
  experiments
• Ventral pathway is bilateral
• Right striate left inferior temporal lesions
• No deficit in object discrimination
• Severing corpus callosum creates deficit
• Dorsal pathway is primarily unilateral
• Right striate left parietal lesions
• Deficit in landmark task
• Severing corpus callosum increases deficit

6
Dorsal and Ventral Functional Pathways

• Neuron receptive field differences
• Parietal lobe neurons
• Large receptive fields
• Specific to hemifield
• More neurons have receptive fields outside the
  fovea than inside the fovea

7
Dorsal and Ventral Functional Pathways

• Neuron receptive field differences
• Parietal lobe neurons
• Temporal lobe neurons
• Large receptive fields
• Not specific to hemifield
• More neurons have receptive fields inside the
  fovea than outside the fovea
• Majority of neurons respond selectively to
  complex stimuli

8
Dorsal and Ventral Functional Pathways

• Kohler et al. PET study in humans supports
  "what/where" distinction
• Position task greater rCBF in right parietal
  lobe
• Object task greater rCBF biaterally at
  occipito-temporal areas

9
Dorsal and Ventral Functional Pathways

• Kanwisher et al. PET study in humans fails to
  isolate cognitive functions during passive
  viewing of shapes

10
Computational Problems in Object Recognition

• How to account for shape-based encoding?
• Objects are more than the sum of their parts
• Properties of object constancy
• Viewing position
• Illumination
• Occlusion

11
Computational Problems in Object Recognition

• Theoretical issues in object recognition
• Frame of reference object constancy across
  orientation
• View-dependent
• Separate representation of an object for each
  viewpoint
• View-invariant
• Critical properties used for object recognition
• Major/minor axes, etc.

12
Computational Problems in Object Recognition

• Theoretical issues in object recognition
• Frame of reference
• Shape encoding hierarchical representations of
  increasing complexity
• Salient features
• Invariants parallelism, symmetry, T-junctions,
  occlusion, etc

13
Computational Problems in Object Recognition

• Theoretical issues in object recognition
• Frame of reference
• Shape encoding hierarchical representations of
  increasing complexity
• Salient features
• Recognition by parts
• Beiderman's geons 24 fundamental 3-D shapes

14
Computational Problems in Object Recognition

• Theoretical issues in object recognition
• Frame of reference
• Shape encoding
• Neurophysiological representations
• Hierarchical coding hypothesis
• Object defined by Gnostic (or grandmother) cell
  single neuron that represents"granny" activated
  by outputs from increasingly more complex
  detectors

15
Computational Problems in Object Recognition

• Theoretical issues in object recognition
• Frame of reference
• Shape encoding
• Neurophysiological representations
• Hierarchical coding hypothesis
• Ensemble coding hypothesis
• Object defined by co-occurrence of complex
  feature detectors

16
Failures of Object Recognition

• Visual Agnosia Failure to recognize visual
  objects
• Limited to visual modality
• Not a sensory deficit
• Not a memory deficit (anomia)
• But deficits can co-occur
• What can disorders of perception tell us about
  perceptual function?

17
Failures of Object Recognition

• Patient GS
• Normal visual acuity
• Normal verbal memory

18
Failures of Object Recognition

• Patient studies of visual perception deficits
  that support "what/where" distinction
• Patient DF bilateral occipital lobe lesions
• Differentiates "perception for identification"
  from "perception for action"

19
Failures of Object Recognition

• Subtypes of agnosia
• Apperceptive agnosia deficit in perceptual
  processing
• Associative agnosia "normal" perceptual
  processing, but deficit in linking percept to
  name

20
Failures of Object Recognition

• Apperceptive agnosia
• Deficits in perceptual processing may be subtle
  and may not be apparent in standard clinical
  tests

21
Failures of Object Recognition

• Apperceptive agnosia
• Warrington
• RH patients had greater deficits in the Gollin
  Picture Test and Incomplete Letters Test than LH
  patients
• RH patients with posterior damage have difficulty
  with object constancy
• Unusual Views Test can't recognize objects from
  unusual view
• Shadows Test can't recognize objects when
  illumination changes

22
Failures of Object Recognition

• Associative agnosia
• Warrington Deficits are not at a perceptual
  level but at a semantic level
• Patient FRA
• LH occipito-temporal lesion
• Could normally parse complex drawing into
  constituent objects, but could not name objects

23
Failures of Object Recognition

• LH and RH patients both fail Matching-by-Function
  Test but for different reasons

24
Failures of Object Recognition

• LH and RH patients both fail Matching-by-Function
  Test but for different reasons
• Warrington's two-stage model of object
  recognition
• Perceptual categorization occurs in the RH
• Semantic categorization occurs in the LH
• Perceptual categorization precedes semantic
  categorization
• But, unilateral LH patients don't always have
  associative agnosia, usually requires bilateral
  lesions

25
Failures of Object Recognition

• Integrative agnosia some elements of
  apperceptive and associative agnosias
• Patient HJA
• Bilateral occipito-temporal lesion
• Could recognize individual objects, but not line
  drawing of objects
• Could perform Unusual Views Test
• Could not recognize overlapping objects
• Deficit in integrating and grouping features

26
Failures of Object Recognition

• Integrative agnosia
• May common problem in most agnosics
• Other Integration failures
• Patient CK
• Closed head injury
• Patient JR

27
Failures of Object Recognition

• Category-specific agnosia associative agnosia
  where deficit is specific to a semantic category
• Patient JBR
• Herpes simplex produced severe associative
  agnosia that was worse for living things than for
  inanimate objects

28
Failures of Object Recognition

• Category-specific agnosia associative agnosia
  where deficit is specific to a semantic category
• Semantic knowledge is structured
• Object categories
• Tools, vehicles, etc.
• Living vs nonliving

29
Failures of Object Recognition

• Category-specific agnosia
• Damasio suggests that nonliving activate
  kinesthetic associations that living things do
  not

30
Failures of Object Recognition

• Category-specific agnosia
• Gaffan and Heywood showed that normals made more
  errors to short latency exposures of living
  things than of nonliving things
• Living things are more similar and share more
  features than nonliving things and are inherently
  more difficult to discriminate

31
Failures of Object Recognition

• Farah and McClellands computational explanation
  for category-specific agnosia
• Property-based organization of semantic system
• Two main layers semantic and input
• Semantic has 31 ratio of visual and functional
  properties
• Input is verbal and visual
• Objects have visual and functional property codes
• 7.71 for living things
• 1.41 for nonliving things
• Train model to discrimination 20 living and
  nonliving things
• "Lesion" model and evaluate recognition accuracy

32
Prosopagnosia

• Inability to recognize faces
• Is there a separate face recognition system?
• Evolutionary support
• Phylogeny of face recognition
• Primates show similar activation to faces
• Development of face recognition
• Neonates track faces longer than other stimuli
• Cross-cultural interpretation of facial
  expression
• Universal recognition of facial expressions

33
Prosopagnosia

• Is there a separate face recognition system?
• Evolutionary support
• Neurophysiological support in primates
• Baylis et al study on neuronal specificity to
  faces in the macaque
• Farah's taxonomy of lesion foci

34
Prosopagnosia
Majority of lesions in occipito-temporal area
35
Prosopagnosia

• Is there a separate face recognition system?
• Evolutionary support
• Neurophysiological support in primates
• Farah's taxonomy of lesion foci
• Function neuroimaging in humans
• Kanwisher et al. fusiform face area

36
Prosopagnosia

• Can face and object perception be dissociated?
• Dissociations between object and face perception
• Right side up faces vs upside down faces
• Patient CK
• Could not identify objects in Arcimbaldo painting
  when right side up, but could identify face when
  painting was upside down
• Patient WJ
• Sheep farmer who had prosopagnosia for human
  faces but could still recognize sheep

37
Prosopagnosia

• Can face and object perception be dissociated?
• Dissociations between object and face perception
• Tanaka and Farah face recognition is more than
  the sum of parts

38
Prosopagnosia

• Two systems for object recognition
• Farah study of patterns of co-occurrence of
  prosopagnosia, visual agnosia, and alexia
  (acquired dyslexia)
• Analytic processing
• Holistic processing

39
Co-occurence of Prosopagnosia, Visual Agnosia,
and Alexia
Number of patients possibly
40
Visual Imagery and Perception

• Does visual imagery use the same neural
  mechanisms as visual perception?

41
Synesthesia

• Involuntary experience of cross-modal association
• 0.05 occurrence in population
• More common in artists, novelists, poets, etc.
• May be related to variation of fusiform area of
  temporal lobe