1. New Data:
Irrelevant Objects & Orientations
Disrupt the
Intermediate-level Layout Representation
Functional Representations of Scene Layout Are Disrupted by Irrelevant Objects Thomas Sanocki & Noah Sulman, Univ. of South Florida
Vision
Cognition
DESIGN
USF
# 295
Can we influence (disrupt) the layout representation?
>> I.e., reduce the scene prime benefits
We added irrelevant objects to the scene primes,
which disrupt the purity of the layout information.
Supportive Data: Prime Time Functions
Overview
Brief exposure to a scene prime activates an intermediate-level representation, which provides a head start on subsequent spatial processing. Recent results separate this early priming effect from later processes. New results implicate orientation information in the intermediate representation, showing that the presence of irrelevant objects and orientations cancels out the benefit of the early information.
Reading these RT Data:
3 Types of Scene Prime (no control prime)
1. Scene prime only,
and scene prime with irrelevant objects,*
either:
2. Same orientation as prime and target (X)
3. Inverted relative to prime and target (IXI)
In figures, prime time is maximum at the left and decreases to 0 ms at right.
The main benefit of the scene prime is its
overall RT relative to background prime.
The benefit was maximal with only 200 ms of prime time, suggesting a fast early process:
X Scene
* one of 4 pairs of
irrelevant objects
appeared randomly
Orientations
slow
Reaction
Time
fast
Background / control
Prime Examples:
Relative
benefit
of scene
prime
“IXI /upright” - Irr Objs are opposite in orientation to prime/target scene (which were both upright)
“X /inverted” - Irr Objs are same orientation
as prime/target scene (inverted)
Method and Theory
We show the method for each trial and discuss the assumed processes.
Scene prime
800 ms ms
Prime Time
Events on Each Trial
850
900
950
1000
1050
1100
1150
800
600
400
200
Prime Time
Reaction Time
Background (control)
Scene prime (identical)
INVERTED
UPRIGHT
54 ms
58 ms
64 ms
46 ms
Sanocki & Sulman (2007, submitted)
1. Fixation
2A. Scene prime
3. Target, until “right/left oval closer” reaction time response
(Irrelevant objects did not appear in target) RT
850
900
950
1000
1050
1100
800 & 600
400 & 200
Prime Times
Reaction Time X
Scene prime only
IXI
INVERTED
UPRIGHT
2B. Background / control prime
<< KEY (examples above)
X = Irrelevant objects in prime
IXI = Irrelevant objects misoriented
(opposite orientation to
prime and target scene) RT
800 ms
Prime Duration
Fixation + prime = 1200 ms total time
Priming as Early Process
Prime activates intermediate-level representation,
>> provides “head-start” on target processing,
resulting in faster RT responses.
Early processes are (see data in middle column):
Full data (noisier):
means for each prime time
Early priming process was:
850
900
950
1000
1050
1100
800
600
400
200
Prime Time
Reaction Time X
Scene prime only
IXI
Fast (required only 200 ms prime time to be maxed)
Broad in scope scene primes benefited 16 spatial relations
Created efficiently / in parallel prime time functions similar for upright and inverted scenes, as shown above
Created efficiently / in parallel prime time functions similar for simple and complex scenes, as shown below
Priming independent of target difficulty additivity of prime x target (data not shown but available)
Fast (maxed with only 200 ms prime time)
Broad in scope > entire scene had 16 spatial relations
Created efficiently / in parallel > no effect of scene complexity
Created efficiently / in parallel > no effect of scene inversion
Priming independent of target difficulty
Priming not sensory (Sanocki, 2003)
Later Processes
Refine intermediate-level representation, eventually resolving the spatial relation between the probes. These processes can be separated from the early processes:
1. Slower
2. Narrow scope (other data)
3. Limited by complexity
4. Limited by inversion
5. Longer with difficult targets
750
800
850
500
300
150 50
Lead Time RT
Complex scenes
Simpler scenes
Conclusions
Scene prime functions
in an experiment with
even shorter durations.
(Effects were quicker but
smaller in magnitude,
due to reduced prime
validity.)
<<
Scene priming is fast (strong at 200 ms) and similar with upright and inverted stimuli.
Irrelevant objects disrupt scene priming and implicate orientation in the intermediate representation that causes priming:
When same orientation as scene, moderate disruption;
When opposite in orientation, all of scene prime benefit was eliminated (resulting in
same level of benefit as with background prime in previous experiments).