Reason for seeing grid-like patterns when strobing a solid black-white image?

Question

I was trying to do some stimulation for SSVEPs by strobing a solid black square and a solid white square. Weird thing was, I ended up seeing weird grid like patterns that looked like this, always of the inverted color.

  _        _
 / \      / \
 | |      | |
 \ / _--_ \ / _--_
  O |    | O |    |
 / \ ---- / \ ----
 | |      | |
 \_/ _--_ \_/ _--_
  O |    | O |    |
     ----     ----

And that sort of pattern repeats itself around the strobing area (where the convex areas are always of opposite color than the current color on the screen).

I'm inclined to think about light and dark adaptation, but it still seems weird that it'd be shaped like that in particular. So... What's going on?

EDIT: Additional info that I should've realized was necessary >_<:

• Strobe area: ~14x10.5 cm

• Strobe frequency: 2 frames/half-cycle @ Laptop Vsync (~60 Hz) ~ 15 Hz

• Viewing distance: ~55-60 cm and shorter.

Screen is backlit. Not sure about brightness metrics.

The pattern isn't offset to one corner like I've drawn, it's more evenly spaced. Horizontal bars are more accentuated than vertical ones (and the dark-on-light phenomenon is more visible for me than the light-on-dark). Moving in closer increases the visibility of the effect (I guess that was to be expected), and changes the number of times the pattern repeats across the screen (but it's hard to count them). The larger the area in my visual field the more times the pattern repeats. I'd say that at about 60 cm it's just one box, but with my nose a handful of cm away from the screen (it's hard to measure by myself >_<), there are maybe three in the horizontal direction... Or at least tens of them (increasingly large as it gets toward the center in both dimensions), depending on how focused in on the screen I am.

It's really really hard to count them.

I haven't had the chance to check with other people yet, but I'll try some time (this is just something I saw while working on a personal project).

Answer 1

Seeing in animals is at some level using Gabor wavelets (Marčelja 1980 and later authors). Your images reminds of a combination of two low frequency Gabor basis vectors. Apparently the nervous system does not have time to model the incoming image into Gabor components of other frequencies.

Here is a picture of two low frequency Gabor basis vectors added. (In this image rotated 45 degrees compared to your image.) http://medicalxpress.com/news/2013-01-feature-combination-perceptual-identification.html

This is a signal processing or a linear algebra question as much as a biology question.

One could argue that even lower frequency Gabor vectors would model the original images even better.