global shape processing involves feature-selective and feature-agnostic coding mechanisms
TRANSCRIPT
Global shape processing involves feature selective and feature agnostic coding
mechanisms
Jason BellMimosa ForsythDavid R. BadcockFred A.A. Kingdom
An object’s shape is a strong cue for its recognition.
Experimentally: Attneave, 1954
…and even earlier…
Specialised shape detectors in area V4:macaque
Gallant et al. (1993, 1996): global form detectors in macaque V4
Neuron 1: concentric & spiral Neuron 2: Radial
Shape coding modelsPasupathy & Connor, 2002; Poirier & Wilson, 20062 critical attributes of a shape1) Curvature of features2) Angular position of curved
featuresStep 1: sample curvatureStep 2: code curvature x angleResulting shape code is unique
Are global shape detectors selective for shape orientation?
To test this we use:a) Radial Frequency (RF)
patterns: widely used to study global shape processing
b) Adaptation and after-effects
Measuring shape orientation selectivity of the RFAAERF3 rotated through 60 degrees:
RF4 rotated through 45 degrees:
4 observers & 6 runs of each condition
Feature invariant processingThe results of Experiment 1 suggest two distinct mechanisms1) Shape orientation selective
mechanism2) Shape orientation invariant
mechanism
Is this invariant mechanism insensitive to other shape attributes?
For example, luminance polarity
Luminance polarity selectivity?Yes and no
All Observers Polarity
Adaptor Rotation Difference
RFAA
E
-0.05
0.00
0.05
0.10
Adaptor
Test
(a) (b)Interaction Between Lum inance Polarity
and Shape Orientation
Adaptor Rotation DifferencePo
larity
Transfer
0.0
0.5
1.0
1.5
black/whitewhite/white
Summing up: two stage global shape coding
Global shape mechanism A: orientation selective and luminance-polarity selective: as the models predict.
Global shape mechanism B:insensitive to orientation and luminance polarity
A A A
A
B
A A
Feature invariant processingCue convergence in object sensitive LOC. 1) Luminance, motion and texture-defined form: Grill-
Spector et al., 1998; 2) Contrast invariant activation to attended object:
Murray & He, 2006
fMRI studies show RF patterns are processed in V4 and the LOCBetts et al. 2008; Rainville et al., 2005; Wilkinson et al., 2000
Consistent with two stage processing of RF shapes