On 2015 Nov 26, Lydia Maniatis commented:

We can get a sense of the intellectual poverty of this study simply by reading the "Conclusions" section:

"Our experiments show that luminance edges play a central role in White's illusion. The illusion seems to be predominantly caused by the luminance edge between the test patch and its background bar, while the edge contrast to neighboring bars is largely ignored. The effect of contour adaptation on White's illusion could not be replicated by spatial filtering models, which adds further evidence against the adequacy of such models as a mechanistic explanation of White's illusion in particular, and lightness perception in general. Our results highlight the importance of further investigating the question of how surface lightness is computed from edge contrast. "

There is no content here because:

1) There is no alternative to the idea that "luminance edges play a central role in White's illusion." Both sides of the display are the same, other than the (perceived) location of the targets as lying on white or on black. Note that there is NO REFERENCE to the authors' orientation claims.

2) The inadequacy of a priori inadequate models that weren't even properly tested (assuming they could be) is not an argument for anything.

3). The statement that "Our results highlight the importance of further investigating the question of how surface lightness is computed from edge contrast" is meaningless. If there are questions to be answered, this study did not address them nor make them seem any more interesting.

The ad hotness of the authors' orientation claims is not only refuted by other displays, such as the Benary cross, but can also be refuted by versions of White's illusion in which the edges of the targets are curved so as to collectively be consistent with, e.g. an interrupted circle. Try it. Such a display, for me, makes more evident the fundamentally bistable character of the lighter-looking group of targets. They can either appear to be part of an opaque surface that lies behind the white stripes, on an amodally-completed black background, or they can appear to form part of a transparency that passes over the black and white stripes. The transparency aspect of White's illusion, which has been noted before, and is noticed by naive observers, is, of course, never touched on in this paper.

On 2015 Nov 08, Lydia Maniatis commented:

This article should probably not be read from the beginning, but from the end – from the last section of the discussion, titled “Difficulties with matching tasks.” The major theoretical problems discussed earlier seem moot once you appreciate that problems with method render the data virtually worthless..

The stimulus effects were not robust but, rather, highly ambiguous and liable to produce theoretically relevant effects that the authors did not bother to consider. This led to highly variable responses that they could not interpret. The authors try and put the blame for their data problems on the lightness matching technique itself, but the fault lies in them: If a task (or a stimulus) is not fit for purpose, it is the fault of those who chose it when it fails to deliver.

The problem is, in fact, the stimuli, which, again, were highly unstable and often did not produce the percepts that the authors predicted and which needed to arise reliably in order to allow them to validate their predictions. Four of ten observers, we are told, did not even perceive White's illusion! These observers were “following some strategy, but [our data] does not allow us to understand what exactly they are doing.”

In general, there was “large variability across and within observers....With simple stimuli of the kind used here, the perceived lightness of different image parts can change over time....In that case, the mean across trials may not be a good indicator of subjective experience [i.e. perception].” So the authors do not really know what their subjects are perceiving.

It seems, further, that the “simple stimuli” produced perceptual phenomena that the authors were or should have been aware are possible – either through the literature or simply on the basis of looking at their stimuli (the least a perception scientist should be doing when planning an experiment is to notice obvious effects of their stimuli) - but did not take into consideration: “Ekroll and Faul (2013) observed an additional transparent layer for similar stimuli...” This might “explain the inconsistent results...” So the “simple stimuli” could produce complex percepts - precepts that, for these authors simply amounts to noise.

In addition: “...observers sometimes selected match lightnesses that were outside the luminance range spanned by the grating. This is surprising....” They really have no idea what their data mean.

The authors' stunning conclusion: “This is just one further example that lightness matching is not such an easy and straightforward task as it might appear on the surface.” They don't seem to realize that no method is secure if your stimuli are ambiguous and unstable and you have not inspected them and factored the potential effects into your theory and method.

Unfortunately, lesson seems not to have been learned, as revealed by the final discussion section of Betz, Shapley, Wichmann and Maertens, 2015, which similarly describes devastating methodological problems as an afterthought. Why are reviewers setting such an obviously low bar for publication?

On 2015 Aug 30, Lydia Maniatis commented:

The authors of this study construct an untenable, ad hoc account of a myopic observation.

The observation is that the contrast effect in White's illusion “is largely determined by edge contrast across the edge orthogonal to the grating, whereas the parallel edge has little or no influence.”

This is a correct literal description of White's illusion. If we had no knowledge of any other contrast illusions, we might overgeneralize from this and conclude that, in general, orthogonal edges produce contrast effects while parallel edges do not. But we do know more. We know that contrast effects are not tied to edge orientation in this way. Consequently we would not reasonably attempt to construct a neural model of contrast based on a principle of “Orthogonal edges produce contrast, parallel edges don't,” since such an ad hoc model, shrink-wrapped to fit White's illusion, would instantly be challenged and falsified by any number of other contrast effects. However, this is precisely what Betz et al propose. (Not to mention that, being “low-level,” the authors' proposed mechanism would also have to be compatible with every other “low-level” as well as every “high-level” effect, e.g. transparency effects, since information from the early parts of the visual system is the basis for the high-level effects.)

In addition, the authors tailor their account to a very weak version of White's illusion (as the authors show, its effect is comparable to the classic slc demo), composed of single target bars rather than columns of aligned bars. The latter produce a much larger lightness differences as well as transparency effects (in the case of the lighter-seeming bars). Does the proposed model work for this classic version as well? Does it work for the classic simultaneous contrast illusion? Does it work for round targets on square backgrounds? Does it explain the Benary cross? If not, then how does such an account contribute to theoretical or practical understanding of the phenomenon (lightness contrast) under consideration? What are the chances that the visual system possesses a mechanism especially for producing a version of White's illusion?

As far as their experimental observations, the authors take a little too much credit for themselves when they say that their experiments have shown that “not all luminance borders that enclose a surface are treated equally in White's illusion.” This is an unnecessarily narrow framing of a many-decades-known, fundamental fact of lightness perception.

If there is one safe statement we can make about simultaneous contrast, it is that it is closely correlated with the appearance of a figure-ground relationship. Specifically, the surface that appears as figure lightens against a darker (apparent) background, and darkens against a lighter one. Mere (apparent) adjacency does not produce contrast. So when the authors claim to have shown that “the luminance step between the test patch and the grating bar on which it is [apparently!] placed is the critical condition for perceiving White's illusion” they a. Have not shown anything new and b. seem naive to the theoretical implications and generality of their own casual description (“on which it is placed”).

The implications are that, given the robustness of the figure-ground/simultaneous contrast link, the challenge of predicting simultaneous contrast reduces to the challenge of predicting figure-ground relations. The latter, in contrast, is not reducible to structure-blind theories involving inhibition/filtering based on luminance ratios. And it is certainly not reducible to an orthogonal/parallel edge dichotomy.

On 2015 Aug 30, Lydia Maniatis commented:

The authors of this study construct an untenable, ad hoc account of a myopic observation.

The observation is that the contrast effect in White's illusion “is largely determined by edge contrast across the edge orthogonal to the grating, whereas the parallel edge has little or no influence.”

This is a correct literal description of White's illusion. If we had no knowledge of any other contrast illusions, we might overgeneralize from this and conclude that, in general, orthogonal edges produce contrast effects while parallel edges do not. But we do know more. We know that contrast effects are not tied to edge orientation in this way. Consequently we would not reasonably attempt to construct a neural model of contrast based on a principle of “Orthogonal edges produce contrast, parallel edges don't,” since such an ad hoc model, shrink-wrapped to fit White's illusion, would instantly be challenged and falsified by any number of other contrast effects. However, this is precisely what Betz et al propose. (Not to mention that, being “low-level,” the authors' proposed mechanism would also have to be compatible with every other “low-level” as well as every “high-level” effect, e.g. transparency effects, since information from the early parts of the visual system is the basis for the high-level effects.)

In addition, the authors tailor their account to a very weak version of White's illusion (as the authors show, its effect is comparable to the classic slc demo), composed of single target bars rather than columns of aligned bars. The latter produce a much larger lightness differences as well as transparency effects (in the case of the lighter-seeming bars). Does the proposed model work for this classic version as well? Does it work for the classic simultaneous contrast illusion? Does it work for round targets on square backgrounds? Does it explain the Benary cross? If not, then how does such an account contribute to theoretical or practical understanding of the phenomenon (lightness contrast) under consideration? What are the chances that the visual system possesses a mechanism especially for producing a version of White's illusion?

As far as their experimental observations, the authors take a little too much credit for themselves when they say that their experiments have shown that “not all luminance borders that enclose a surface are treated equally in White's illusion.” This is an unnecessarily narrow framing of a many-decades-known, fundamental fact of lightness perception.

If there is one safe statement we can make about simultaneous contrast, it is that it is closely correlated with the appearance of a figure-ground relationship. Specifically, the surface that appears as figure lightens against a darker (apparent) background, and darkens against a lighter one. Mere (apparent) adjacency does not produce contrast. So when the authors claim to have shown that “the luminance step between the test patch and the grating bar on which it is [apparently!] placed is the critical condition for perceiving White's illusion” they a. Have not shown anything new and b. seem naive to the theoretical implications and generality of their own casual description (“on which it is placed”).

The implications are that, given the robustness of the figure-ground/simultaneous contrast link, the challenge of predicting simultaneous contrast reduces to the challenge of predicting figure-ground relations. The latter, in contrast, is not reducible to structure-blind theories involving inhibition/filtering based on luminance ratios. And it is certainly not reducible to an orthogonal/parallel edge dichotomy.

On 2015 Nov 08, Lydia Maniatis commented:

This article should probably not be read from the beginning, but from the end – from the last section of the discussion, titled “Difficulties with matching tasks.” The major theoretical problems discussed earlier seem moot once you appreciate that problems with method render the data virtually worthless..

The stimulus effects were not robust but, rather, highly ambiguous and liable to produce theoretically relevant effects that the authors did not bother to consider. This led to highly variable responses that they could not interpret. The authors try and put the blame for their data problems on the lightness matching technique itself, but the fault lies in them: If a task (or a stimulus) is not fit for purpose, it is the fault of those who chose it when it fails to deliver.

The problem is, in fact, the stimuli, which, again, were highly unstable and often did not produce the percepts that the authors predicted and which needed to arise reliably in order to allow them to validate their predictions. Four of ten observers, we are told, did not even perceive White's illusion! These observers were “following some strategy, but [our data] does not allow us to understand what exactly they are doing.”

In general, there was “large variability across and within observers....With simple stimuli of the kind used here, the perceived lightness of different image parts can change over time....In that case, the mean across trials may not be a good indicator of subjective experience [i.e. perception].” So the authors do not really know what their subjects are perceiving.

It seems, further, that the “simple stimuli” produced perceptual phenomena that the authors were or should have been aware are possible – either through the literature or simply on the basis of looking at their stimuli (the least a perception scientist should be doing when planning an experiment is to notice obvious effects of their stimuli) - but did not take into consideration: “Ekroll and Faul (2013) observed an additional transparent layer for similar stimuli...” This might “explain the inconsistent results...” So the “simple stimuli” could produce complex percepts - precepts that, for these authors simply amounts to noise.

In addition: “...observers sometimes selected match lightnesses that were outside the luminance range spanned by the grating. This is surprising....” They really have no idea what their data mean.

The authors' stunning conclusion: “This is just one further example that lightness matching is not such an easy and straightforward task as it might appear on the surface.” They don't seem to realize that no method is secure if your stimuli are ambiguous and unstable and you have not inspected them and factored the potential effects into your theory and method.

Unfortunately, lesson seems not to have been learned, as revealed by the final discussion section of Betz, Shapley, Wichmann and Maertens, 2015, which similarly describes devastating methodological problems as an afterthought. Why are reviewers setting such an obviously low bar for publication?

On 2015 Nov 26, Lydia Maniatis commented:

We can get a sense of the intellectual poverty of this study simply by reading the "Conclusions" section:

"Our experiments show that luminance edges play a central role in White's illusion. The illusion seems to be predominantly caused by the luminance edge between the test patch and its background bar, while the edge contrast to neighboring bars is largely ignored. The effect of contour adaptation on White's illusion could not be replicated by spatial filtering models, which adds further evidence against the adequacy of such models as a mechanistic explanation of White's illusion in particular, and lightness perception in general. Our results highlight the importance of further investigating the question of how surface lightness is computed from edge contrast. "

There is no content here because:

1) There is no alternative to the idea that "luminance edges play a central role in White's illusion." Both sides of the display are the same, other than the (perceived) location of the targets as lying on white or on black. Note that there is NO REFERENCE to the authors' orientation claims.

2) The inadequacy of a priori inadequate models that weren't even properly tested (assuming they could be) is not an argument for anything.

3). The statement that "Our results highlight the importance of further investigating the question of how surface lightness is computed from edge contrast" is meaningless. If there are questions to be answered, this study did not address them nor make them seem any more interesting.

The ad hotness of the authors' orientation claims is not only refuted by other displays, such as the Benary cross, but can also be refuted by versions of White's illusion in which the edges of the targets are curved so as to collectively be consistent with, e.g. an interrupted circle. Try it. Such a display, for me, makes more evident the fundamentally bistable character of the lighter-looking group of targets. They can either appear to be part of an opaque surface that lies behind the white stripes, on an amodally-completed black background, or they can appear to form part of a transparency that passes over the black and white stripes. The transparency aspect of White's illusion, which has been noted before, and is noticed by naive observers, is, of course, never touched on in this paper.