Transcript Slide 1

Masking occurs when
the perception of a brief stimulus (the target)
is impaired by another stimulus (the mask)
presented in close temporal and spatial contiguity
Why study masking?
Because it opens a window on the mind/brain
Masking occurs in two main flavours:
a) by integration (camouflage)
b) by interruption:
i) metacontrast masking
ii) pattern masking
iii) object-substitution masking
Let’s start with integration…
Masking by temporal integration
CW:
Masking by temporal integration is mediated
by visible persistence
(iconic memory)
Visible Persistence: an image that remains visible for a brief
period after the display has been turned off.
(NOT a retinal afterimage)
Storage model of visible persistence
(hydrodynamic metaphor)
Sensory store
Storage model of visible persistence
(the metaphor can also be electrodynamic)
sensory store
visible persistence
time
(10 ms)
How can we measure the duration of visible persistence?
Task: name the matrix location of missing dot
Present the matrix in two sequential frames, separated by an ISI
ISI
F1
F2
Short ISI
F1
ISI
F2
time
F1
F2
time
Long ISI
F1
ISI
F2
time
No overlap
No integration
F1
F2
time
F1
ISI
F2
time
F1
ISI
F1
F2
F2
ISI
Percent correct
100
VDL
JH
BJ
DR
80
60
40
20
0
40
80
120 160
ISI (ms)
200
mat1 noarg space
Charging the sensory store (increasing the duration of F1)
F1
F1
F1
ISI
F2
ISI
F2
ISI
F2
time
Percent correct
100
VDL
RGB
SKP
80
60
40
20
0
10 40
80
120
160 200
Duration of F1 (ms)
mat1 noarg space
F1 duration = 16 ms
F1 F2
F1 duration = 300 ms
F1
F2
(ISI = 0)
mat1 noarg space
When F1 is long, integration breaks down.
F1 and F2 become segregated
In fact, when F1 is long, F1 and F2 become segregated
even when displayed concurrently
F2
F1
Demo: mat2 20-1-0
Visible persistence is time-locked
to the onset of the stimulus, not to its offset
This means that temporal integration follows
an SOA law, not an ISI law
This also means
that visible persistence
is not based on a decaying sensory store.
Visible persistence is linked to a burst of neural activity
time-locked to stimulus onset
(Duysens, Orban, & Maes, 1985)
.
In brief,
The duration of visible persistence varies inversely
with stimulus duration
(inverse-duration effect)
F1
F1
F1
F2
F2
F2
time
All these effects
(integration at short SOAs and segregation at long SOAs)
can be explained
based on ideas of von Holst (1954) and MacKay (1957)
The onset of a stimulus creates an image of itself
(opens an object file)
that accepts new data for a very brief period
(the critical period of integration)
After that, the image acts as a filter that passes
only new stimuli
Short SOA:
F1
F2
The critical period is up:
the object is complete.
No more parts can be
added.
Long SOA with ISI:
ISI
F1
F2
The critical period is up:
the object is complete.
No more parts can be
added.
Long SOA, no ISI:
F2
F1
Applying the principles of temporal integration
to forward masking by pattern
mask
target
Masking by temporal integration
mask
target
msk
tgt
time
Demo: int1-4
In the matrix task a long F1 produced segregation.
A long leading mask produces unmasking.
tgt
mask
time
mask
mask + target
Demo
tgt
mask
time
seg1-4
msk
tgt
DURATION OF LEADING MASK (ms)
tgt
msk
DURATION OF LEADING MASK (ms)
An account based on ideas of von Holst (1954) and MacKay (1957)
(Long SOA)
mask
mask msk+tgt
tgt
On the basis of this evidence, one might conclude that forward masking never
occurs when the leading mask is displayed for longer than about 100 ms
But note the effect of task switching (e.g. conventional display sequence in masked priming)
5
420
C
440
420
Reaction Time (ms)
7
3
2
9
5
7
C
5
440
Reaction Time (ms)
Stream
No
stream
400
380
360
340
320
300
Grid Noise Digit Script
Leading Distractors
400
380
360
340
320
300
No Stream
Stream
No forward masking in either case when no backward mask is presented
RT differences mediate more backward masking in the Stream condition
Masking occurs in two main flavours:
a) integration (camouflage)
b) interruption:
i) metacontrast masking
ii) pattern masking
iii) object-substitution masking
Metacontrast masking
No spatial overlap between target and mask
When target-mask SOA is short, the target’s visibility is reduced
(but not when SOA is equal to zero)
Conventional demonstration of metacontrast masking
(visibility ratings)
TARGET
Conventional demonstration of metacontrast masking
(visibility ratings)
MASK
SOA = 0
No masking
(10 ms)
(10 ms)
Stimulus sequence
in metacontrast masking
SOA
Visibility ratings are subjective
Target
Mask
Target + Mask
SOA = 0
No masking
(10 ms)
Target
(10 ms)
Mask
sim
sim
sim
sim
Stimulus sequence
in metacontrast masking
ISI
Target
SOA
Mask
seq
seq
seq
seq
U-shaped curve
SOA
tgt
ed
ed
0
varied
10
100
10
100
100
100
% correct responses
ondition
tch. Gap
msk
80
80
60
60
40
40
Average
Br.ISIMtch.
No
With
GapISI
20
20
0
40
80
SOA (ms)
120
160
Effect of varying the SOA
SOA = 0
SOA = 100
SOA = 200
SOA = 300
SOA = 500
Effect of
target-mask separation
(SOA = 100 ms)
Sep = 1
Sep = 2
Sep = 3
Sep = 4
Sep = 5
Reducing the mask’s contours
(SOA = 100 ms)
4 dot 1
4 dot 2
4 dot 3
4 dot 4
Function of SOA, not ISI (SOA law)
With ISI
No ISI
ed
ed
0
varied
10
100
10
100
tgt
msk
msk
100
100
% correct responses
ondition
tch. Gap
ISI
tgt
80
80
60
60
40
40
Average
Br.ISIMtch.
No
With
GapISI
20
20
0
40
80
SOA (ms)
120
160
NOTE: no masking when SOA = 0
A theory of metacontrast masking:
Cross-Channel Inhibition
(Breitmeyer & Ganz, 1976, 2005)
1.
A brief display generates activity in two visual channels:
a) Transient channels (short latency; short duration; low SF; mediate perception of stimulus onset)
b) Sustained channels (long latency; long duration; high SF; mediate perception of stimulus identity)
2.
Activity in the transient channels suppresses activity in the sustained channels
no masking when SOA = 0
time
target
mask
+
+
1.
A brief display generates activity in two visual channels:
a) Transient channels (short latency; short duration; mediate perception of stimulus onset)
b) Sustained channels (long latency; long duration; mediate perception of stimulus identity)
2.
Activity in the transient channels suppresses activity in the sustained channels
time
target
mask
Support for the cross-channel inhibition hypothesis
inhibition vanishes in scotopic viewing (von Békésy)
and so does metacontrast masking
Light-adapted
Dark-adapted
Masking of a pattern by parts of itself reveals a new form of masking:
Object substitution
F1
(30 ms)
F2
F1
F2
(300 ms)
F2
Demo:
mat2 0-2-0
mat2 0-2-20
(vary overlap)
This is unquestionably a form of masking but:
The mask consists of a portion of the target
The target-”mask” SOA is equal to zero
Remember: no metacontrast masking when SOA = 0
So, let’s turn this into a classical metacontrast paradigm,
and see if masking occurs when SOA = 0
Metacontrast paradigm
target
mask
time
SOA = 0
demo
0
1x
2x
3x
4x
The finding that metacontrast masking occurs when T-M SOA = 0
is inconsistent with accounts based on
inhibitory contour interactions
+
+
In addition,
common-onset masking
occurs without substantial masking contours
(four-dot masking)
1dotx
2dotx
3dotx
4dotx
Common-onset masking occurs with overlearned geometrical shapes
The target can be seen easily when the
entire display is presented briefly
(10 ms)
os16-0a
os16-0b
os16-0c
os16-0d
Display sequence for
object-substitution masking
(10 ms)
(320 ms)
os16-600a
os16-600b
os16-600c
os16-600d
How does object substitution
come about?
?
Object-substitution masking does not occur when set size = 1
(10 ms)
(320 ms)
os1-600a
os1-600b
os1-600c
os1-600d
Object substitution masking does not occur
when attention is directed to the target location
X
os16-600a
X
os16-600b
Expectation based on learned regularities
(occlusion of far objects by near objects)
influences OSM
Kahan & Lichtman (P&P, in press)
30 ms
600 ms
Expectation based on learned regularities
(occlusion of far objects by near objects)
influences OSM
Kahan & Lichtman (P&P, in press)
Masking Effect
% C (common - delayed offset)
far
near
10
8
6
4
2
0
-2
n-n f-f f-n n-f
Target-mask depth
Common-onset masking has two components
Inhibitory contour interactions
early
photopic
independent of attention
requires substantial masking contours
does not occur at SOA = 0
Object-substitution
relatively late
scotopic
critically dependent on attention
requires minimal masking contours
occurs at SOA = 0
?
Separating the two components of common-onset masking:
(inhibitory contour interactions and object substitution)
Photopic
% correct responses
100
Scotopic
100
80
80
60
60
Set Size
(No. elements)
40
1
2
4
8
16
20
0
80 160 240 320
0
80 160 240 320
Duration of mask (ms)
Separating the two components of common-onset masking:
(inhibitory contour interactions and object substitution)
photopic
scotopic
Masking occurs in two main flavours:
a) integration (camouflage)
b) interruption:
i) metacontrast masking
ii) pattern masking
iii) object-substitution masking
Pattern masking is probably the most commonly used form of masking
but, in my opinion, it is the least interesting
Depending on viewing conditions, it includes aspects of both
integration and object substitution
Here is one way in which the processes of integration and interruption
(object substitution) may combine to yield masking
(Michaels & Turvey, 1979)
Ciao!