typicality sharpens category representations in object...
TRANSCRIPT
Fixation
+ + +...
8 images
160 ms 590 ms 160 ms 160 ms12 s
+
Fixation FixationBoatBlock
12 s 6 s 12 s
Run1
Run2
Run3
Run 4
Run 5
Run 6
Run 7
Run8
2 sessions x 8 runs each (80 min total)
300 s
FixationPlaneBlock
6 s 12 s
FixationBirdBlock
6 s 12 s
FixationDogBlock
6 s 12 s
16 blocks
FixationCarBlock
6 s 12 s
Fixation
12 s
FixationDogBlock
6 s 12 s
...
MalamuteCanoe Biplane Hen Sedan Pug
12 s
+
Fixation
590 ms
Run9
Run10
Run11
Run 13
Run 14
Run 15
Run16
Run 12 p < 0.025 p < 0.050 p < 0.025p < 0.050p > 0.050
LH RH
More Typical > Less TypicalLess Typical > More Typical
Typicality Rankings for Real-World Categories
Typicality Sharpens Category Representations inObject-Selective Cortex
Marius Cătălin Iordan¹, Michelle R. Greene¹, Diane M. Beck², Li Fei-Fei¹1Department of Computer Science, Stanford University 2Beckman Institute and Department of Psychology, University of Illinois at Urbana-Champaign
[email protected] [email protected] [email protected] [email protected]
*
V1 V2 V3v hV4
*
LOC
0.00
0.03
0.06
0.09
-0.03
CATE
GO
RY B
OUN
DARY
EFF
ECT
(R)
0.00
0.03
0.06
0.09
-0.03
Most Typical Least Typical
**
n.s.n.s.n.s.n.s.
C G W
0.0
2.0
-2.0
V1 V2 V3v hV4
***
LOC
0.10
0.20
0.30
0.40
0.00
CORR
ELAT
ION
(R)
0.10
0.20
0.30
0.40
0.00
Most Typical Least Typical
**
**
****n.s.n.s.n.s.n.s.
C G W
0.0
1.0
-1.0
-0.10 -0.10
methods:16 images per exemplar (across 2 blocks)
passive viewing
analysis: use MVPA to characterize similarity and dissimilarity
of activity patterns within and between categories
Typical exemplars more similar to central category tendency in LOCTypical exemplars distinguish more strongly between categories in LOC
Suggests prototype representation for real-world objects
BirdsCatsDogsFishCarsBoatsPlanesTrains
more typical less typical
behavioral experimenttypicality rankings
for each 8 exemplars within their category
stimulus set8 categories
8 exemplars per category16 images per exemplar
ANIMALS(animate, natural)
VEHICLES(inanimate, man-made)
which is the best dog?faster recognition, faster categorization for typical items
Posner & Keele (1968), Rosch (1973), Rosch & Mervis (1975)
RT for recognition / categorization
Neural Typicality Hypothesis: Family Resemblance
relationship with central category tendencytypical items share more features in common with central tendency
differentiating between categoriessharper category boundaries between typical items compared to less typical items
typicality effect no typicality effect
LOC V1, V2,V3v, hV4
Whole-Brain Searchlight AnalysisfMRI Experiment Summary
Less typical exemplars more distinguishable in cIPLcIPL involved in category learning, memory of object context
Suggests contextual facilitation of categorization for atypical exemplars
Typicality may constitute a previously unexplored principle of organization for intra-category neural structure
New HypothesesIntermediate visual processing may prioritize the embedding of
behaviorally relevant dimensions of variance directly into neural representations
Selected References: Posner & Keele (1968) J. Exp. Psychol. ✦ Rosch (1973) Cogn. Dev. Acq. Lang. ✦ Rosch & Mervis (1975) Cogn. Psychol. ✦ Deng et al. (2011) CVPR ✦ Davis & Poldrack (in press) Cerebral Cortex ✦ Iordan, Greene, Beck, & Fei-Fei (in press) JOCN