This week I’m attending the annual meeting of the Vision Sciences Society in Naples Florida. Every day or so, I’ll post about a subset of the cool, interesting, funny, or quirky (I won’t say which) talks/posters I happened to catch. You can read the first installments here.
This morning had a great new color illusion/demo presented in a talk by Jordan Suchow and George Alvarez entitled “Silent updating: Cross-dimensional change suppression.” The effect is pretty similar to one presented by Jun Saiki and Alex Holcombe a couple years ago at Vision Sciences (“Surface-based, unpaired feature representations mediate detection of change to feature pairings”), but this one was a particularly dramatic demonstration.
If you show people a dot that continuously cycles through all the colors in the spectrum, you can see it changing systematically. If you arrange 300 such dots in a thick ring around the center of the display (imagine Saturn and one of its rings) and you maintain your gaze at the center of the display (e.g., on Saturn), you can see all of the colors flickering and changing. But, if you start the entire ring rotating around where you’re looking, all the color-changing appears to halt. The effect is dramatic, and Suchow and Alvarez generate several possible explanations for it. They argue that people are silently updating the colors of the objects in the ring so that if the display halts and suddenly reverts to the original colors, people notice. I didn’t find that explanation particularly compelling — it’s unlikely people can update all of the individual colors simultaneously, and reverting to the original colors will produce a big luminance signal. My bet is that they store little if anything about the colors, but they are really good at seeing motion. The explanation I prefer is one suggested by a questioner in the talk: The rotation of the entire ring produces a big motion signal for the visual system, and that signal masks the smaller changes of the individual dots. That principle–big motions hide smaller ones–is used in magic as well. A big arm movement can hide a subtle change to something held in your hand.
Thanks for commenting. I really loved the demo even though I’d beg to differ with the interpretation. I am familiar with both of the alternatives you raise, of course. However, I don’t think either really applies here. Both assume that the dot colors are actually being represented for more than an instant. The alternative I favor is that subjects need not have much of a representation of the dots at all. If they’re just living in the moment, they would notice large instantaneous changes but not small ones, and they wouldn’t notice progressive changes that take a while to occur. The change from the current state back to the original state is a large, instantaneous change that produces big luminance changes in that instant. The change from the current state to a nearby current state produces less of a signal.
Note that this sort of change detection doesn’t require a detailed comparison to a long-gone representation. It just requires perception of a luminance signal at that moment (one instant to the next).
Note that this same critique applies to the earlier studies of progressive orientation and luminance changes by Andrew Hollingworth. If people are just comparing the current state to the immediately preceding one, then a change back to the original scene is a huge change. They’re not really updating. Their just representing in the moment.
Glad you enjoyed the demo.
We aimed to distinguish two accounts of our observers’ failure to notice the rapid color changes: freezing and silent updating. (Here, freezing means having an outdated representation, and silent updating means having a current one. For an example of color-freezing in vision, see [1].) You correctly note that there is a big change in luminance at the moment the display reverts to the old colors. This isn’t a defect — it’s critical to the design. A change in luminance implies a difference in luminance before and after the change. However, from the vantage point of an observer who holds an outdated representation, the reverted display is identical to his outdated representation. With no differences, there is no change to detect. Since our observers do detect a change, it confirms that their representation is not outdated, ruling out the freezing account.
Like you, we are also intrigued by interactions between the mechanisms that detect motion and flicker. We didn’t have time to talk about it, but we’ve done experiments to address this directly — stay tuned.
-Jordan Suchow & George Alvarez
[1] Motoyoshi, I. (2007). Temporal freezing of visual features. Current Biology, 7, 404-406.