Orientation-selective
adaptation to first- and
second-order patterns in
human visual cortex
Jonas Larsson, Michael S. Landy, David J. Heeger
Department of Psychology &
Center for Neural Science,
New York University, USA
jonas@cns.nyu.edu
What are first- and
second-order patterns?
•
•
First-order: vary in luminance; can be detected by linear filters
Second-order: do not vary in luminance; cannot be detected by linear filters
Filter-rectify-filter (FRF) model of
second-order vision
First stage: many
small-scale linear
filters
Rectify output of
first stage filters
Second stage: Sum
rectified output with
large-scale linear filter
Predicts separate mechanisms for first- and second-order vision
Separate mechanisms?
• Psychophysics: different mechanisms for first- and
second-order vision
• Electrophysiology: first-order neurons as early as V1;
second-order neurons in extrastriate visual cortex
of cats (area 18) and macaques (V2,V4,V5/MT)
• But many second-order neurons also selective for
first-order patterns (cue invariance) - not
predicted by FRF
• Neuroimaging: anatomical segregation of first- and
second-order vision in humans? conflicting results
Open questions
segregation? Are any human visual
• Anatomical
areas specialized for first- or second-order vision?
second-order mechanisms? Are
• Multiple
different types of second-order patterns (contrast,
orientation) processed by the same mechanism?
Are there neurons that respond
• Cue-invariance?
both to first- and second-order patterns?
Approach
• Adapt orientation-selective neurons
• Measure responses to adapted & orthogonal
stimulus orientation with fMRI
• Event-related design
• Use independently identified visual area ROIs
Stimulus conditions
Unimodal
Probe
Adapter
LM:LM
First-order
CM:CM
Cross-modal
OM:OM
Second-order
luminance
contrast
orientation
First-order
luminance
Second-order
contrast
orientation
LM:OM
First-order
luminance
Second-order
orientation
Trial types
Adapter orientation
Probe orientation
Parallel trials
Orthogonal trials
Blank trials
blank screen
Trial structure
Adapter
4s
Unattended
perifoveal stimulus
(1.5-5 deg)
Attended
foveal RSVP task
(<1 deg)
ISI
1s
Probe
1s
Response
1.2 s
how
many
X’s?
fMRI methods
•
•
•
3 subjects
•
Siemens Allegra 3T, quadrature
surface coil
•
BOLD EPI, 19 slices
perpendicular to calcarine
sulcus, TR=1.2s, TE=30ms,
FA=75
•
Bite bar & motion correction
with FSL
8 scanning sessions per subject
280 trials per stimulus
condition and trial type
Visual area ROIs
V7
IPS
dorsal
LO1/
V3B
V3A
V3
V2
medial
CS
lateral
ITS
CoS
V1
V2 V3
hV4
VO1
1 cm
left
V5/
MT
up
pLOC
right
down
Co
ventral
LO2
Results:
Unimodal adaptation first-order (LM:LM)
Orientation-selective
adaptation to first-order
patterns in V1
FMRI response
(% modulation)
0.8
Response
amplitude (A):
Parallel trials
Orthogonal trials
0.6
0.4
Adaptation
0.2
0
Adapter Probe
-0.2
0
5
10
Time (s)
15
Adaptation to first-order
patterns: luminance (LM:LM)
Parallel trials
1
0.8
LM:LM
Orthogonal trials
***
Adaptation
index:
*** ***
***
0.6
*** **
A
A
***
0.4
0.2
V5
V7
A
V3
2
LO
B
V3
hV
4
VO
1
pL
O
C
V3
V2
0
V1
FMRI response amplitude
(% modulation)
1.2
Adaptation indices: first-order
patterns (LM:LM)
• Adaptation indices
Adaptation index
0.8
LM:LM
• No significant
0.6
differences between V1
and extrastriate visual
areas
0.4
0.2
0
constant across visual
areas
V1
V2
V3
hV4 VO1 V3B V3A
Visual area
• Adaptation in V1 can
account for adaptation
in extrastriate visual
areas
Results:
Unimodal adaptation second-order
(CM:CM & OM:OM)
Adaptation to second-order
patterns: contrast (CM:CM)
CM:CM
*
1
Parallel trials
*
Orthogonal trials
***
0.8
***
0.6
***
***
**
0.4
0.2
V5
V7
A
V3
2
LO
B
V3
C
O
pL
1
VO
4
hV
V3
V2
0
V1
FMRI response amplitude
(% modulation)
1.2
Adaptation to second-order
patterns: orientation (OM:OM)
1
OM:OM
*
Parallel trials
***
Orthogonal trials
***
0.8
0.6
*** ***
0.4
*
**
0.2
*
V5
V7
A
V3
2
LO
B
V3
VO
1
pL
O
C
4
hV
V3
V2
0
V1
FMRI response amplitude
(% modulation)
1.2
Adaptation indices: second-order
patterns (CM:CM & OM:OM)
*
CM:CM
0.6
*
*
0.4
0.2
0
V1
V2
V3
hV4 VO1 V3B V3A
Visual area
0.8
Adaptation index
Adaptation index
0.8
OM:OM
0.6
*
0.4
0.2
0
V1
V2
V3
hV4 VO1 V3B V3A
Visual area
• Adaptation indices greater in extrastriate areas than in V1
• Suggests additional adaptation in extrastriate visual cortex
Cue invariance?
Cross-modal adaptation
(LM:OM)
LM:OM
1
Parallel
Orthogonal
0.8
0.6
0.4
0.2
Adaptation index
1.2
0.4
0.2
0
-0.2
V1
V2
V3
hV4 VO1 V3B V3A
Visual area
Visual area
V5
V7
A
V3
LO
2
VO
1
pL
O
C
V3
B
4
hV
V3
V2
0
V1
FMRI response amplitude
(% modulation)
No cross-adaptation between
first- and second-order
patterns (LM:OM)
Conclusions
• Results consistent with FRF model
• First-order neurons in V1
• FRF second-stage neurons both in V1 and
extrastriate areas (feedback to V1?)
• Different second-order patterns processed
in same visual areas
• No evidence of cue-invariant neurons
• Selective attention is not required for firstor second-order pattern adaptation
Acknowledgements
Center for Neural Science/
Department of Psychology:
Denis Schluppeck
Barbara Knappmeyer
Justin Gardner
NYU Center for Brain
Imaging:
Keith Sanzenbach
Souheil Inati
jonas@cns.nyu.edu