On 2015 Nov 04, Lydia Maniatis commented:

Continuation of "area rule" story: Avoiding acknowledging the role of figure-ground in contrast effects also allowed advocates of "anchoring theory" to claim that "the debate seems to revolve around layer models and framework models" (Gilchrist, 2015), frameworks being adjacent "like countries on a map." Figure-ground structure means layers, and so would complicate this presentation.

On 2015 Sep 20, Lydia Maniatis commented:

In line with the earlier point about avoiding figure/ground issues, the "multi-sector" stimuli used here are of the sort used by Gestalt psychologists to demonstrate and study figure-ground effects (e.g. the role of sector color, size, orientation; bistability). But such effects and their potential lightness consequences are not acknowledged.

On 2015 Sep 17, Lydia Maniatis commented:

"Simple" vs "complex" stimulus is not a meaningful distinction, because there is no objective criterion for making it. Geometric simplicity may or may not produce complexity in the percept, of different kinds.

Which is more simple: a. the three pacmen of a Kanizsa triangle, that produce subjective contours/ contrast/amodal completion; b. the simultaneous contrast effect that produces contrast and amodal completion? c. the de Valois checkerboard that produces assimilation; d. a random array of variously shaped, overlapping white, grey, black shapes? e. the checkerboard used here, which produces transparency/inhomogenous illumination effects; f.a black shape that produces the impression of two overlapping black shapes; g. etc.

The authors offer their own “clear distinction” of simple vs complex. Not only is this proposed distinction unclear, it lacks theoretical content. The authors subsequently seem to ignore it, and use the terms in a loose and undefined way. The proposed distinction, attributed to “anchoring theory” is: “A simple image is one in which all the surfaces lie within a single illumination level (a single framework) whereas complex images contain multiple adjacent fields of illumination (multiple frameworks).” This distinction is theoretically hollow, because it references actual (actual level of illumination) and not perceived (perceived illumination) image features. By this definition, a photograph of sunlight and shadow should count as a single framework, provided it is being viewed under a homogeneous illumination. Even if we take the authors' distinction to refer to perceived illumination (which begs the question the “frameworks” argument is supposed to answer, i.e. how do we parse the scene into separate illumination “frameworks”), they don't stick to it. A little later, for example, we are told that the stimuli of Boyaci et al (2014) “are still fairly simple, [but] qualify [] as complex as they consist of more than one framework” (the distinction is invoked to explain discrepancies between those authors findings and the findings of R and G (2014). But Boyaci et al's stimuli are neither under inhomogeneous illumination nor are they designed to create such an impression. So the definition of “framework,” which wasn't explanatory to begin with, has shifted in the space of a few paragraphs, from “framework of illumination” to something else (what?). We are similarly told that “a dozen previous experiments...used images that qualify as complex images;” but many (perhaps all) of these studies did not involve inhomogeneous illumination, real or perceived.

On 2015 Sep 17, Lydia Maniatis commented:

A short, instructive history of the “area rule”

The “area rule” was born in an attempt to deny the role of figure-ground structure in lightness. This is its original sin, and has led to interesting distortions in theory and practice.

Classic disc-annulus experiments, which demonstrated the dependence of lightness on luminance ratios, also showed an asymmetrical influence of disc and annulus. The annulus would always look white, and push a lighter disc towards luminosity. In other words, raising the luminance of the annulus would lighten the disc, but not vice versa. The disc appeared to lie on top of an amodally-completed larger disc, so an easy provisional conclusion would be to assign different influences to figure and ground in mediating lightness perception of a surface (the same asymmetry applies to figure-ground contrast in general).

Gilchrist et al (1999) did not like this solution because it interfered with their preferred theoretical assumption that the highest luminance in a (vaguely-defined) “framework” would be white. Clearly, in the very simple, disc-annulus situation, the highest luminance was not necessarily white. Instead of acknowledging a role for figure-ground, Gilchrist et al (1999) created a new rule, stating that if the darker area was more than 50% larger than the smaller, then it would lighten, and progressively push the smaller, lighter area toward luminosity. They presented a speculative function, reprinted in Gilchrist and Radonjic (2009), that has never been corroborated, despite a number of attempts.

The Gilchrist group's own results constantly cried out for a figure-ground explanation. Tellingly, they were forced to modify their area claim to include “amodally-completed” area – thus in effect making the area rule indistinguishable from a figure-ground claim. Later, Economou et al (2007) acknowledged a similar, figure-ground-related asymmetry in the simultaneous contrast display. The team acknowledged the asymmetry but did not explore it further.

Preserving the highest-luminance-white rule was not the only or even most important incentive for rejecting a possible figure-ground role. Another fundamental claim of Gilchrist et al (1999) was that the classic simultaneous contrast demonstration is due to a process which, at a certain stage, treats each square and its interior as a separate “framework” and evaluates its contents based on the ratio principle and highest-luminance rule. The idea that the lightness of the targets is actually mediated by local luminance contrast between the apparent figure and its background was not compatible with this assumption. However, it is easy to show (Maniatis, 2015), by adding surfaces within each putative “framework” that border contrast between figure and ground, not the ratios with all surfaces contained in the background square, mediates this effect.

The commitment to avoiding acknowledging a role for figure-ground explains, I believe, the preference manifested by investigators with these theoretical commitments for stimuli which either did not produce figure-ground effects, or in which the contrast effects would average out. Specifically, they adopted the use of checkerboards or Mondrians, and random or semi-random selection of luminances. Such stimuli and choices muddy rather than clarify the role of structure in lightness. Thus, proponents of a “Gestalt” theory, were, paradoxically forced by their commitments to prefer stimuli in which image structuring could be ignored.

There was a second reason that this “Gestalt” theory needed to avoid confronting the role of structure in lightness, and this was that it did not/could not address the fact that we sometimes perceive surfaces as lying beneath transparent layers with their own lightness. As in the case of figure-ground/amodal completion effects, such layers arise when contours showing good continuation intersect, with the added proviso that the luminance structure is compatible with such a solution. Checkerboards, lacking such cues, avoid such effects.

Well, actually, they don't. They often produce multiple such effects, as well as luminosity. This latter result arose in Radonjic et al (2011). It was awkward and they tried to explain it away by selectively attributing inconvenient results to presentation on an “emissive” screen. Allred et al (2012) tentatively acknowledged, after much highly technical wrestling with very low-resolution data, the self-evident yet apparently unplanned-for fact that checkerboards do produce differential lightness impressions. So restricting the class of allowable stimuli (rationalized on the basis that they were “simple” and that results would carry over to more “complex” situations – a view codified in the oft-repeated “applicability assumption”) in order to avoid confronting figure-ground and transparency effects has nevertheless led researchers back to these same, unavoidable issues.

It is interesting that the checks on a checkerboard can coalesce into transparent overlays/underlying surfaces despite the absence of apparent overlap. It is surely not unrelated to the fact that checkerboards produce assimilation rather than contrast when we replace a black or white check with a grey one (the de Valois and de Valois checkerboard contrast demonstration). It would be worth analyzing.

On 2015 Aug 12, Lydia Maniatis commented:

This article is a bit misleading. The authors seem to be claiming to have corroborated the area rule, at least for the case where the darker area fills more than half the visual field.

As defined and illustrated by Gilchrist et al (1999) and Gilchrist and Radonjic (2009), the area rule predicts the onset of luminosity when the darker area exceeds half the display. But this prediction has never been corroborated, even for far, far larger darker area coverage, nor has the description of the rule changed. In addition, lightening of the darker are has never been shown to start until the latter covers far more than 50% of total area. So the claim that this study has corroborated the "area rule" needs to be qualified by the introduction of a new description of the area rule.

On 2015 Aug 12, Lydia Maniatis commented:

This article is a bit misleading. The authors seem to be claiming to have corroborated the area rule, at least for the case where the darker area fills more than half the visual field.

As defined and illustrated by Gilchrist et al (1999) and Gilchrist and Radonjic (2009), the area rule predicts the onset of luminosity when the darker area exceeds half the display. But this prediction has never been corroborated, even for far, far larger darker area coverage, nor has the description of the rule changed. In addition, lightening of the darker are has never been shown to start until the latter covers far more than 50% of total area. So the claim that this study has corroborated the "area rule" needs to be qualified by the introduction of a new description of the area rule.

On 2015 Sep 17, Lydia Maniatis commented:

A short, instructive history of the “area rule”

The “area rule” was born in an attempt to deny the role of figure-ground structure in lightness. This is its original sin, and has led to interesting distortions in theory and practice.

Classic disc-annulus experiments, which demonstrated the dependence of lightness on luminance ratios, also showed an asymmetrical influence of disc and annulus. The annulus would always look white, and push a lighter disc towards luminosity. In other words, raising the luminance of the annulus would lighten the disc, but not vice versa. The disc appeared to lie on top of an amodally-completed larger disc, so an easy provisional conclusion would be to assign different influences to figure and ground in mediating lightness perception of a surface (the same asymmetry applies to figure-ground contrast in general).

Gilchrist et al (1999) did not like this solution because it interfered with their preferred theoretical assumption that the highest luminance in a (vaguely-defined) “framework” would be white. Clearly, in the very simple, disc-annulus situation, the highest luminance was not necessarily white. Instead of acknowledging a role for figure-ground, Gilchrist et al (1999) created a new rule, stating that if the darker area was more than 50% larger than the smaller, then it would lighten, and progressively push the smaller, lighter area toward luminosity. They presented a speculative function, reprinted in Gilchrist and Radonjic (2009), that has never been corroborated, despite a number of attempts.

The Gilchrist group's own results constantly cried out for a figure-ground explanation. Tellingly, they were forced to modify their area claim to include “amodally-completed” area – thus in effect making the area rule indistinguishable from a figure-ground claim. Later, Economou et al (2007) acknowledged a similar, figure-ground-related asymmetry in the simultaneous contrast display. The team acknowledged the asymmetry but did not explore it further.

Preserving the highest-luminance-white rule was not the only or even most important incentive for rejecting a possible figure-ground role. Another fundamental claim of Gilchrist et al (1999) was that the classic simultaneous contrast demonstration is due to a process which, at a certain stage, treats each square and its interior as a separate “framework” and evaluates its contents based on the ratio principle and highest-luminance rule. The idea that the lightness of the targets is actually mediated by local luminance contrast between the apparent figure and its background was not compatible with this assumption. However, it is easy to show (Maniatis, 2015), by adding surfaces within each putative “framework” that border contrast between figure and ground, not the ratios with all surfaces contained in the background square, mediates this effect.

The commitment to avoiding acknowledging a role for figure-ground explains, I believe, the preference manifested by investigators with these theoretical commitments for stimuli which either did not produce figure-ground effects, or in which the contrast effects would average out. Specifically, they adopted the use of checkerboards or Mondrians, and random or semi-random selection of luminances. Such stimuli and choices muddy rather than clarify the role of structure in lightness. Thus, proponents of a “Gestalt” theory, were, paradoxically forced by their commitments to prefer stimuli in which image structuring could be ignored.

There was a second reason that this “Gestalt” theory needed to avoid confronting the role of structure in lightness, and this was that it did not/could not address the fact that we sometimes perceive surfaces as lying beneath transparent layers with their own lightness. As in the case of figure-ground/amodal completion effects, such layers arise when contours showing good continuation intersect, with the added proviso that the luminance structure is compatible with such a solution. Checkerboards, lacking such cues, avoid such effects.

Well, actually, they don't. They often produce multiple such effects, as well as luminosity. This latter result arose in Radonjic et al (2011). It was awkward and they tried to explain it away by selectively attributing inconvenient results to presentation on an “emissive” screen. Allred et al (2012) tentatively acknowledged, after much highly technical wrestling with very low-resolution data, the self-evident yet apparently unplanned-for fact that checkerboards do produce differential lightness impressions. So restricting the class of allowable stimuli (rationalized on the basis that they were “simple” and that results would carry over to more “complex” situations – a view codified in the oft-repeated “applicability assumption”) in order to avoid confronting figure-ground and transparency effects has nevertheless led researchers back to these same, unavoidable issues.

It is interesting that the checks on a checkerboard can coalesce into transparent overlays/underlying surfaces despite the absence of apparent overlap. It is surely not unrelated to the fact that checkerboards produce assimilation rather than contrast when we replace a black or white check with a grey one (the de Valois and de Valois checkerboard contrast demonstration). It would be worth analyzing.

On 2015 Sep 17, Lydia Maniatis commented:

"Simple" vs "complex" stimulus is not a meaningful distinction, because there is no objective criterion for making it. Geometric simplicity may or may not produce complexity in the percept, of different kinds.

Which is more simple: a. the three pacmen of a Kanizsa triangle, that produce subjective contours/ contrast/amodal completion; b. the simultaneous contrast effect that produces contrast and amodal completion? c. the de Valois checkerboard that produces assimilation; d. a random array of variously shaped, overlapping white, grey, black shapes? e. the checkerboard used here, which produces transparency/inhomogenous illumination effects; f.a black shape that produces the impression of two overlapping black shapes; g. etc.

The authors offer their own “clear distinction” of simple vs complex. Not only is this proposed distinction unclear, it lacks theoretical content. The authors subsequently seem to ignore it, and use the terms in a loose and undefined way. The proposed distinction, attributed to “anchoring theory” is: “A simple image is one in which all the surfaces lie within a single illumination level (a single framework) whereas complex images contain multiple adjacent fields of illumination (multiple frameworks).” This distinction is theoretically hollow, because it references actual (actual level of illumination) and not perceived (perceived illumination) image features. By this definition, a photograph of sunlight and shadow should count as a single framework, provided it is being viewed under a homogeneous illumination. Even if we take the authors' distinction to refer to perceived illumination (which begs the question the “frameworks” argument is supposed to answer, i.e. how do we parse the scene into separate illumination “frameworks”), they don't stick to it. A little later, for example, we are told that the stimuli of Boyaci et al (2014) “are still fairly simple, [but] qualify [] as complex as they consist of more than one framework” (the distinction is invoked to explain discrepancies between those authors findings and the findings of R and G (2014). But Boyaci et al's stimuli are neither under inhomogeneous illumination nor are they designed to create such an impression. So the definition of “framework,” which wasn't explanatory to begin with, has shifted in the space of a few paragraphs, from “framework of illumination” to something else (what?). We are similarly told that “a dozen previous experiments...used images that qualify as complex images;” but many (perhaps all) of these studies did not involve inhomogeneous illumination, real or perceived.

On 2015 Sep 20, Lydia Maniatis commented:

In line with the earlier point about avoiding figure/ground issues, the "multi-sector" stimuli used here are of the sort used by Gestalt psychologists to demonstrate and study figure-ground effects (e.g. the role of sector color, size, orientation; bistability). But such effects and their potential lightness consequences are not acknowledged.

On 2015 Nov 04, Lydia Maniatis commented:

Continuation of "area rule" story: Avoiding acknowledging the role of figure-ground in contrast effects also allowed advocates of "anchoring theory" to claim that "the debate seems to revolve around layer models and framework models" (Gilchrist, 2015), frameworks being adjacent "like countries on a map." Figure-ground structure means layers, and so would complicate this presentation.