Response to Maniatis critique of anchoring theory

Perhaps her most important charge is that edge classification, a key part of my earlier intrinsic image theory (Gilchrist, 1979), was abandoned by anchoring theory but more recently, surreptitiously re-introduced. In fact, edge classification was never abandoned. It morphed into the idea of framework segregation. The source of the confusion lies in the following sentence: ‘‘...the belongingness construction allows us to bypass the problem of edge classification, even though factors like edge sharpness and coplanarity, formerly thought to underlie edge classification, now show up as grouping factors.’’ (Gilchrist et al., 1999, p. 805) Some background is needed here. Anchoring theory emerged as an attempt to account for the overall pattern of errors in surface lightness perception, on the belief that the pattern of errors is the signature of the visual software. The literature reveals many kinds of lightness errors. But they fall into two broad classes: illumination-dependent errors (historically called failures of constancy) and background-dependent errors (historically called illusions). Thus the main challenge for a theory of errors is to find a formula that can account for these two seemingly different classes. For anchoring theory, the link lies in the concept of frame of reference, which had earlier been applied only to lightness constancy (Koffka, 1935) and failures of lightness constancy (Kardos, 1934). However lightness illusions also typically seem to involve frames of reference. Simultaneous lightness contrast, the prototype, is composed of two obvious side-by-side frameworks. Frameworks are also prominent in the Benary effect (1924) and White’s illusion (1979), and various reverse contrast illusions, to name a few. As all these frameworks are supported by gestalt grouping principles, they can also be treated as perceptual groups, and the same can be said of illumination frameworks. Of course the frameworks in these lightness illusions are not bounded by illumination edges. But in the language of perceptual grouping, segregation and grouping are intimately related; one is roughly the flip-side of the other. Thus the frameworks, or perceptual groups, central to anchoring theory depend on both segregation between frameworks (including edge classification) and consolidation within frameworks. Viewing frameworks as perceptual groups supported by both segregation and grouping factors allows the Kardos idea of co-determination to be applied to both broad classes of lightness errors, failures of illumination-independent constancy and failures of background-independent constancy. Edge classification retains an important role, but is supplemented (bypassed if you will) by grouping processes. Many of the arguments made by Maniatis are predicated on her mistaken conclusion that edge classification had been rejected by anchoring theory.