|
Orientation variance| old_uid | 6215 |
|---|
| title | Orientation variance |
|---|
| start_date | 2009/02/09 |
|---|
| schedule | 11h-12h30 |
|---|
| online | no |
|---|
| summary | When I was a graduate student, I had a long argument with Denis Pelli about using letters to study vision. I said that gratings were better. Contrary evidence is the citation count for our paper about letter identification. Oh well. When I got interested in visual crowding, which was traditionally examined with letters, I decided it was time to vindicate gratings. Like letters, the orientation of a little patch of grating becomes hard to identify when you donπt look straight at it and other gratings are nearby. What you tend to see is a semi-homogenous texture, in which all the little gratings are more-or-less parallel. This phenomenon seemed to suggest that we are particularly insensitive to orientation variance. I got so wound up trying to document this insensitivity that I have all but forgotten about crowding. I will briefly describe three ongoing projects related to my new obsession. Well, the first is pretty much finished. I had observers discriminate between various amounts of orientation variance in textures composed of little gratings. Turns out weπre not so insensitive after all. Results of another experiment, using the visual search paradigm, suggest that weπre actually hard-wired to detect orientation variance: plaids "pop out" from gratings, but not vice versa. The third project has by far the most surprising results. When observers are asked to ignore orientation variance and just report the average orientation in a texture made of gratings, their estimates become more accurate when the variance increases. Specifically, they lose their bias for obliquity. My presentation will end when I explain why this final result is so hard to reconcile with simple Bayesian models of orientation bias. |
|---|
| responsibles | Waszak |
|---|
| |
|