This is an excellent review of the literature on Glass Patterns. The history section is especially interesting.
The article is both concise and comprehensive. However, I thought that there should be some mention of Wilson's 2-stage model. Otherwise, readers are left with the impression that nothing is known beyond the initial local stage of processing. In fact, Wilson's model provides a plausible mechanism for explaining the variation in detection threshold among linear, concentric, radial and spiral patterns.
Also, in discussing the role of higher cortical areas, I think it would be appropriate to reference Jack Gallant's work on responses to non-Cartesian gratings in V4. While his study did not actually use Glass patterns, the grating stimuli that were used had similar geometry and are clearly relevant.
In paragraph 6 of "Physiological basis...." I think it would be appropriate to also cite Wilson's 1997 paper as I believe this was the first to show the striking differences between concentric and translational patterns.
Concentric orientation summation in human form vision. Wilson HR, Wilkinson F, Asaad W. Vision Res. 1997 Sep;37(17):2325-30.
Author's reply: We appreciate these comments and agree with them, and have now included these citations and appropriate discussion in the article.
I think this is a generally well written article that provides an admirable description of Leon’s discovery of Glass patterns. The facts regarding Glass patterns are certainly all accurate and clearly presented. There are, however, a few points where I feel there should be changes or inclusions, and these are listed below.
1. Wilson, H. R., Wilkinson, F. & Asaad, W. (1997) Concentric orientation summation in human form vision. Vision Res., 37, 2325-2330 should certainly be cited, as it is the first quantitative model that accurately predicted Glass pattern thresholds for concentric patterns. In particular, the model used oriented V1 receptive fields for the first stage, and this was followed by full wave rectification (i.e. effectively creating both on-center and off-center cells), then by a global pooling stage that was sensitive to concentric curvature. We modified this to produce a model for global pooling in radial Glass patterns in our 1998 paper. Your article, as it now stands, seems to suggest that there is little neural modeling of the requisite global pooling in the literature, but our work shows that is not the case. Doubtless better models will emerge, but ours has been around for 14 years, and it accounts for the psychophysics.
2. When Leon and Switkes (1976) used opposite contrast dots (black & white, or different colors) to generate concentric Glass patterns, it is certainly true that the concentric patterns generated could not be perceived. However, a roughly orthogonal quasi-radial pattern is frequently perceived (I believe Leon still has one of these in blue and yellow on his living room wall). This can be explained at least qualitatively by V1 oriented receptive fields. Whereas two dots of the same polarity will most strongly stimulate a cell if both fall within the center zone of the appropriate cell (on- or off-center), if the polarities are opposite, the strongest stimulation will occur when one falls in the center and the other in the surround. This means that the cell must have a roughly orthogonal orientation. It might be worth mentioning this in the article.
3. The Dakin & Bex (2002) study is cited, but we have severely criticized it in print: Wilson, H. R. & Wilkinson, F. (2003) Further evidence for global orientation processing in circular Glass patterns. Vision Res. 43, 563-564. I’m sending Leon a pdf of this rejoinder. In fact, we showed that they provided further evidence for concentric summation in their own data. Finally, it has more recently been shown using VEPs that our rank ordering of sensitivity to Glass patterns (concentric most sensitive, radial next, and translational very poor) is correct, and these authors used square windows for all their patterns (Pei, Pettet, Vildaviski & Norcia, 2005). This again contradicts Dakin & Beck. In fact, Bex even pointed their poster out to me at a conference and indicated that he thought we had probably been correct. My suggestion is that you simply remover the one sentence and reference to Dakin & Bex. If you do prefer to include it, then I think it’s very important to include our rejoinder and mention of the VEP evidence. In fact, you might want to mention the VEP evidence in any case, as it further attests to the cortical significance of at least concentric and radial Glass pattern structures.
So that’s it. I have enjoyed reading and commenting on your article. I’ll look forward to your responses.
Pei, F., Pettet, M.W., Vildaviski, V.Y., & Norcia, A.M. (2005). Event related potentials show configural specificity of global form processing. Neuroreport, 16, 1427-1430.
Sincerely, Hugh R. Wilson
Author's reply: We're grateful for the suggestions on all three counts. We've now included discussion of the Wilson et al (1997) model, as well as citations to the Wilson & Wilkinson (2003) reply to this article and the Dakin & Bex (2003) rejoinder. In addition, we now discuss the case of opposite polarity Glass patterns, which we agree are interesting and illustrative in terms of mechanism.