Relative Efficiency for the Detection of Apparent Motion

1
Relative Efficiency for the Detection of Apparent
Motion

• Bryan L. Gross, David R. Pope, Theodore E. Cohn
• Vision research, volume 36, issue 15, Aug 1996
• Presented by Paul Smorenburg, in cooperation
  with Bert Kruyt and Thomas v. Wezel

2
To start with

• In this presentation we take a look at the
  system that detects (apparent) motion
• Introduction
• Experiment
• Discussion
• Conclusion

3
Introduction

• Do we use different channels for perception of
  motion and position?
• Study Motion Perception using experiences from
  contrast perception
• So using dots instead of gratings
• Thus allowing to make an relatively assumption
  free estimate of efficiency

4
-Introduction-Efficiency

• Informative way of looking at measured contrast
  treshold.
• Quantifies the amount of available information
  the observer has used.
• Use an Ideal observer to compare real observer
  performance to.
• Efficiency then is a function of stimulus
  detectability for the real observer and for the
  ideal observer. Ef(dR, dI)

5
-Introduction-Assumptions

• When measuring Efficiency, one has to make
  assumptions about the visual system and sources
  of information loss.
• Common Assumption
• Constant percentage of quanta is not used by the
  visual system.

6
-Introduction-This Experiment

• To minimize assumptions, design 2 tasks on wich
  the predicted performance of an ideal observer is
  the same.
• Thus, Efficiency is not quantified.
• Therefore no assumptions have to be made.
• Result Qualitative indication of relative
  efficiency of the observer.

7
Methods

• 2 tasks
• Motion discrimination (A)
• Position discrimination (B)
• Luminance is varied
• In each task Indicate which stimulus is presented

8
-Methods-Ideal observer performance

• For an ideal observer, the first flash in motion
  detection task does not provide any information.
• This becomes a position discr. task
• Operation performed by ideal observer is the same
  on both tasks.
• Expected performance is the same

9
-Methods-Stimuli and Procedure

• Spots produced by LEDs
• Six different ISIs between 1 and 150 msec
• 3 observers
• 20 threshold estimates for each task and each
  ISI value

10
-Methods-Predictions

• Ideal observer performance will be the same on
  both tasks.
• Human observer will use different systems for
  motion and position detection.
• Hence, performance may be different on the two
  tasks.

11
Results
12
Discussion

• Subjects perform better at discriminating motion
  from non-motion than discriminating one position
  from another.
• What explains this difference?

13
-Discussion-Spatial or Temporal uncertainty

• Could the first flash be a spatial cue reducing
  uncertainty in location?
• No, spots are visible when not lit.
• First flash cue to timing of second flash?
• No, when irrelevant flash occurs after
  information flash, No significant difference in
  threshold value.

14
-Discussion-Summation

• Could the stimulus with 2 flashes on the same
  spot look brighter due to summation of the
  flashes?
• Would make distinction easier.
• New experiment to test this idea

15
-Discussion-Summation

• Threshold for left-left stimulus IS lower.
• But this difference is not enough to account for
  the difference between motion perception and
  position perception
• Summation hypothesis can be rejected.

16
-Discussion-Motion Pathway

• Motion and position stimuli equally detectable
• Discriminable at treshold
• They must be detected by different systems
  Motion pathway

17
Relation with ISI

• As ISI increases treshold increases
• At longer ISI position pathway takes over
• At high ISI no sensation of motion
• 60-100 msec upper limit for short range motion
  system

18
Conclusion

• We are more sensitive to moving stimuli than
  equally intense non moving stimuli
• We attribute this advantage to use of position
  motion pathways.