You are on page 1of 1

Visual Backward Masking: Feed-forward or Recurrent?

Frouke Hermens (1), Gediminas Luksys (2), Wulfram Gerstner (2), Michael H Herzog (1)

1) Laboratory of Psychophysics, Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne, Switzerland
2) Laboratory of Computational Neuroscience, Brain Mind Institute, Ecole Polytechnique Fédérale de Lausanne, Switzerland

Introduction Simulation results: U-shaped and


Most models of visual backward masking either make monotonic masking functions
use of a feedforward or a recurrent structure.
Here, we show that a simple 2D Wilson-Cowan type

Target

Mask
network model, employing lateral connections only, can
account for many masking phenomena.

2D Wilson-Cowan model 50

45
High intensity mask
Low intensity mask

We
40

35

Covariation with template


Excitatory layer 30

25

20

15
Wi We
10

Inhibitory layer 5

0
0 10 20 30 40 50 60 70 80
V ISI (ms)

Wi

V Increasing the intensity of the mask changes the


masking function from U-shaped to mononotic.

Stimulus
(300x140
pixels) Simulation results: Common onset
masking and the effects of attention
Target

Mask

Results: Shine-through
20 ms
10
300 ms Early
Intermediate
Late

Percept 9.5

Psychophysical
Covariation with template

Threshold 9

T = 22 ms T = 30 ms T = 80 ms
8.5
25 grating

7.5
0 5 10 15 20 25 30 35 40
T = 22 ms T = 30 ms T = 80 ms Mask continuation after target offset (ms)
5 grating

Conclusions
+ A simple 2D Wilson-Cowan type model can explain
T = 22 ms T = 30 ms T = 80 ms
many basic temporal and spatial effects of masking
gaps in grating

+ No feedforward or recurrent structures are needed to


explain these effects

Contact: frouke.hermens@epfl.ch

You might also like