Object Recognition

Psy391h - Classes 11-12

(This page last updated 27 Mar 2006.)
 

 What vs. Where Pathways

• What Pathway
  • Object recognition.
  • Receives information about color, orientation, form.
  • Damage can cause visual agnosia.
• Where Pathway
  • Representing locations of things.
  • Includes representation of speed and direction of motion.
  • Important for manipulating objects.
  • Damage can cause spatial neglect.

  Lesion studies with macaques

  • Temporal lesions impair performance in object recognition task.
    • Object recognition also depends on corpus callosum connections.
  • Parietal lesions impair performance in landmark task.
    • Landmark task depends less on corpus callosum connections.
  • double dissociation

  Similar dissociation between brain areas activated by What and Where tasks in human brain imaging studies.

What Pathway

• Hierarchical Coding Hypothesis
  • Each neuron at the bottom level represents simple feature such as an edge.
  • Neurons at next layer up represent combinations of features, such as corners.
  • Higher levels represent surfaces, and eventually objects.
  • Probably not that simple.
    • Could there be a separate neuron for every object you can recognize?
    • "grandmother" cell responds to any view of your grandmother.
• Ensemble Coding Hypothesis
  • Different neurons respond to different features, or properties.
  • The pattern of these neurons firing together represents Grandma.

Where Pathway

• Where Pathway function illustrated in Patient DF
  • Damage to What Pathway, but Where Pathway intact
  • Cannot copy pictures, but can draw objects from memory.
  • Cannot report slot orientation by matching it with card, but when "posting" card into slot, automatically adjusts orientation to match.
  • Cannot report width of block with fingers, but when picking it up, automatically adjusts width between fingers to match block.
  • Cannot draw object shape, but positions fingers on either side of center when picking it up.

Types of Visual Agnosia

• Apperceptive Agnosia
  • Associated more with right hemisphere damage
  • Difficulty recognizing objects or letters with degraded stimuli.
  • Difficulty recognizing objects at nonstandard orientations.
  • Inability to segment complex scenes into parts.
• Associative Agnosia
  • Able to reproduce stimuli by drawing.
  • Able to segregate complex stimuli into separate objects.
  • Nonetheless, unable to name objects.
  • Associated more with left hemisphere damage
• Integrative Agnosia
  • Can reproduce stimuli by drawing.
    • Slow process, with no apparent grouping into separate objects.
  • Unable to segregate overlapping objects.
  • May recognize parts without recognizing whole objects.
  • Unable to integrate parts into wholes.
• Category-Specific Agnosia
  • Some patients have more difficulty naming living things than nonliving objects.
    • A few patients have more difficulty naming nonliving objects.
  • One Explanation: Separate visual memory stores for living things and nonliving objects…
    • BUT visual discriminations between living things may be more difficult than between nonliving objects.
    • AND objects may be recognized using functional representations as well as visual representations.
  • Another Explanation
    • No separation between representations of living and nonliving things.
    • Separation between visual and functional info in semantic memory.
    • Model of visual object naming by Farah and McClelland
    • Connectionist architecture.
    • Verbal and visual representations only linked through semantic memory.
    • Semantic memory divided.
      • Visual memory
      • Functional memory
    • Damage to visual semantic memory produces bigger deficit for living things than for nonliving.
    • Opposite pattern with damage in functional semantic memory.

       

Prosopagnosia

• Inability to recognize faces.
• Other objects may still be recognized correctly.
• Is there a separate mechanism for recognizing faces?
• Fusiform Face Area
  • In the fusiform gyrus.
  • On right, very active in face recognition.
  • On left, active in both face and object recognition.
  • Some activity on right for other discriminations (birds, cars).
• Are Only Faces Affected in Prosopagnosia?
  • Evidence for shared mechanism across faces and other objects.
    • One farmer who became prosopagnosic was no longer able to recognize his cows.
    • One birdwatcher who became prosopagnosic was no longer able to distinguish between bird species.
  • Evidence for face-specific mechanism
    • Patient W.J.
      • Started sheep farming after becoming prosopagnosic.
      • Learned to recognize some sheep, while still unable to recognize humans.
      • Performs better at learning names for sheep faces than for human faces.
      • Seems to have lost ability for human face recognition without losing all ability to distinguish similar stimuli.
    • Patient L.H.
      • Can distinguish between different eyeglasses at a normal level, but is far below normal in face recognition.
      • Below normal with upright faces, but still better with upside down faces than with upright faces.

Faces Processed as a Complete Configuration

• Complete faces and houses studied
• Tested on recognition of individual parts
  • Some parts presented by themselves.
  • Others presesnted as part of a configuration.
• Face parts recognized better when part of configuration.
• House parts recognized equally well in isoloation.

One Possible Explanation: Separate Analytic and Holistic Recognition Systems

• Analytic
  • Components processed separately.
  • More left hemisphere.
  • Important for identifying words.
• Holistic
  • Entire configuration processed as a unit.
  • More right hemisphere.
  • Important for faces.

    •  

Imagery

We won't spend much time on imagery in class, but be sure to read the section on imagery in the textbook and bring any questions you have to class.

next: Attention

Psych 391h: Cognitive Neuroscience Kyle Cave Psychology Dept. U. Mass.