Experience Dependent Object Perception

1
Experience Dependent Object Perception

• Richard Zemel
• Computer Science Department
• University of Toronto

2
Two Sets of Experiments
1. To what degree is object perception
invariant? Familiarity Naming Reference-Frames
.2. What is the role of experience in
completion? Novel occluded shapes Occlusion
without occluders
3
Collaborators

• Mary Peterson
• Larry James
• Dave Towers
• Marlene Behrmann
• Mike Mozer
• Daphne Bevalier

4
Sample Stimuli
5
Familiarity Methods

• Training Phase 16 stimuli, fixed locations
• passive viewing (6 blocks)
• active old/new discrimination (2 blocks)

• repeat

• Testing Phase (2 blocks)
• new trials -- 16 distractors
• old trials -- half of learned objects
• stay in learned location
• shift to diagonally-opposite location

6
Familiarity Accuracy
7
Familiarity RT
8
Naming Methods

• Training Phase 8 rounds 8 named objects
• passive 2 blocks (160 ms) GIX
  (500 ms)
• active 1 block (160 ms) keyboard
  (feedback)

• Testing Phase 6 rounds half shift sides, half
  stay
• passive 2 blocks
• active 1 block (no feedback)

9
Naming Accuracy
10
Reference-Frames Variations
11
Reference-Frames Methods

• Training Phase 2 rounds 16 objects
• passive 6 blocks
• active 2 blocks
  old/new (f-back)

• Testing Phase same/different retinal screen
  locs
• active 2 blocks
  old-new

12
Reference-Frames Conditions
13
Reference-Frames Results
14
Experience Dependence When Objects Irrelevant?

• Many properties of objects not invariant, but
  rather depend on experience
• Evidence from experiments in which object memory
  directly relevant to task
• Is experience important when the object is
  unnecessary to accomplish task?

15
Object attention occlusion
16
Object attention sensitive to layout
17
Object attention experience
18
Experience affects object attention
19
Completion without occluder?
Subjects complete fragments given experience with
potential linking shape Is evidence of
occlusion required?
20
Completion w/o occluder Methods

• Phase 1 Ends displays

• Phase 2 V displays
• Phase 3 Ends and Vs

21
Missing Occluder Results
22
Conclusions

• Memories of novel objects are specific to learned
  locations
• Both retinal screen coordinates involved, so
  not just episodic memory

• Object attention applies to recently viewed novel
  shapes
• Experience-dependent object effects can apply to
  fragments without occluder

23
Current Directions

• Duration of effects
• Generalization
• Amount of experience required