On 2015 Dec 01, Lydia Maniatis commented:

Very brief summary: The central claim about the discovery of a perceptual difference in the effect on a target of a narrow gradient versus a wide gradient when both are invisible has not been corroborated because:

a. the wide gradient is visible;

b. the effect of the narrow gradient is the very same as that which would occur in the absence of this gradient. The attempt to spin it differently is based on past experiments in different conditions using narrow but visible, not invisible, surrounds.

On 2015 Nov 30, Lydia Maniatis commented:

his article claims to introduce a new illusion whose novelty and theoretical interest hinge on the presence of invisible luminance gradients - one shallow, one steep. The problem, as any reader can confirm for themselves, is that the shallow gradient is quite visible.

The discussion section contains an admission of the inaccuracy of the invisibility claim and an attempt to downplay the relevance of this failure to control for visibility: “A problem...is how to interpret the contrast effect obtained with the display pair C-D, in which the luminance gradients were barely visible...We ...conclude that the validity of a high-level interpretation of our results based on illumination cues does not hinge critically on whether the luminance gradients in our study were actually invisible or “almost” invisible.”

Contrast the above statements with others such as that “The experimental results clearly indicate that luminance gradients do not need to be visible in order to influence the lightness of embedded surfaces....the effect produced by pair C-D is not surprising. This effect is just an invisible gradient version of the effect demonstrated earlier by Agostini and Galmonte (1997, 2002);” “the luminance range of gradual transitions has been reduced to make them “invisible” and “unnoticeable;” “we manipulated the width of invisible luminance gradients.” The difference between invisible and barely visible is that the latter is visible, and the former invisible.

Given that the main empirical claim (from the summary) is that “the width of an invisible luminance gradient determines the lightness of a target that it surrounds...wide invisible gradients generate contrast effects, while narrow invisible gradients generate assimilation effects,” the visibility of the shallow gradient means that visibility/invisibility was, in fact, an uncontrolled confound in these experiments, and that the central claim has not been corroborated.

The meaningfulness of the “assimilation” description of the invisible narrow gradient is also called into question when the authors acknowledge that b/c the gradient was so narrow, “one might argue that the outer background could have had a larger effect on the disk appearance than the local surround.” They attempt to deflect this possibility by citing previous experiments in which a visible, narrow local surround produced assimilation. In this comparison between experiments, just as in the comparisons within this experiment, the visibility/invisibility factor constitutes a confound that undermines any conclusions as to a distinction between wide and narrow gradients.

But wasn't subjects' ability to perceive the shallow gradient an empirical question? The procedure used actually did not allow this key question to be settled, something which would not have been particularly difficult. Observers were simply asked to describe the displays. When they did not spontaneously report on the differences between the targets, they were directed by the experimenter to do so. Because they did not spontaneously report on the smoky or cloudy effect of the shallow gradients, the authors concluded (or at least state) that “none of the observers noticed the gradual luminance gradient.” But in this case, why do the authors themselves later describe the gradients as “barely visible”? They were obviously aware that a failure to spontaneously report on some aspect of the stimuli did not necessarily indicate a failure to perceive, and they took measures to fill in some relevant gaps. Since the shallow gradients are, in fact, visible, treating the failure to report as failure to perceive is transparently poor practice.

The discussion is conceptually confused, ad hoc and fully-hedged, but it does not seem worthwhile for the present purposes to analyze the various flawed attempts at interpretation of a confounded and biased dataset.

Finally, it should be noted that the authors refer to other “notable approaches” to lightness “including Gilchrist's Anchoring theory...critiqued by Rudd (2010, 2013, 2014)...” It would have been relevant here to mention that a paradigmatic claim of “anchoring theory” has been straightforwardly and definitively falsified (Maniatis, Lydia M. "A simple test of the “anchoring account” of simultaneous contrast." Journal of Vision 15.5 (2015): 13-13).

On 2015 Nov 30, Lydia Maniatis commented:

his article claims to introduce a new illusion whose novelty and theoretical interest hinge on the presence of invisible luminance gradients - one shallow, one steep. The problem, as any reader can confirm for themselves, is that the shallow gradient is quite visible.

The discussion section contains an admission of the inaccuracy of the invisibility claim and an attempt to downplay the relevance of this failure to control for visibility: “A problem...is how to interpret the contrast effect obtained with the display pair C-D, in which the luminance gradients were barely visible...We ...conclude that the validity of a high-level interpretation of our results based on illumination cues does not hinge critically on whether the luminance gradients in our study were actually invisible or “almost” invisible.”

Contrast the above statements with others such as that “The experimental results clearly indicate that luminance gradients do not need to be visible in order to influence the lightness of embedded surfaces....the effect produced by pair C-D is not surprising. This effect is just an invisible gradient version of the effect demonstrated earlier by Agostini and Galmonte (1997, 2002);” “the luminance range of gradual transitions has been reduced to make them “invisible” and “unnoticeable;” “we manipulated the width of invisible luminance gradients.” The difference between invisible and barely visible is that the latter is visible, and the former invisible.

Given that the main empirical claim (from the summary) is that “the width of an invisible luminance gradient determines the lightness of a target that it surrounds...wide invisible gradients generate contrast effects, while narrow invisible gradients generate assimilation effects,” the visibility of the shallow gradient means that visibility/invisibility was, in fact, an uncontrolled confound in these experiments, and that the central claim has not been corroborated.

The meaningfulness of the “assimilation” description of the invisible narrow gradient is also called into question when the authors acknowledge that b/c the gradient was so narrow, “one might argue that the outer background could have had a larger effect on the disk appearance than the local surround.” They attempt to deflect this possibility by citing previous experiments in which a visible, narrow local surround produced assimilation. In this comparison between experiments, just as in the comparisons within this experiment, the visibility/invisibility factor constitutes a confound that undermines any conclusions as to a distinction between wide and narrow gradients.

But wasn't subjects' ability to perceive the shallow gradient an empirical question? The procedure used actually did not allow this key question to be settled, something which would not have been particularly difficult. Observers were simply asked to describe the displays. When they did not spontaneously report on the differences between the targets, they were directed by the experimenter to do so. Because they did not spontaneously report on the smoky or cloudy effect of the shallow gradients, the authors concluded (or at least state) that “none of the observers noticed the gradual luminance gradient.” But in this case, why do the authors themselves later describe the gradients as “barely visible”? They were obviously aware that a failure to spontaneously report on some aspect of the stimuli did not necessarily indicate a failure to perceive, and they took measures to fill in some relevant gaps. Since the shallow gradients are, in fact, visible, treating the failure to report as failure to perceive is transparently poor practice.

The discussion is conceptually confused, ad hoc and fully-hedged, but it does not seem worthwhile for the present purposes to analyze the various flawed attempts at interpretation of a confounded and biased dataset.

Finally, it should be noted that the authors refer to other “notable approaches” to lightness “including Gilchrist's Anchoring theory...critiqued by Rudd (2010, 2013, 2014)...” It would have been relevant here to mention that a paradigmatic claim of “anchoring theory” has been straightforwardly and definitively falsified (Maniatis, Lydia M. "A simple test of the “anchoring account” of simultaneous contrast." Journal of Vision 15.5 (2015): 13-13).

On 2015 Dec 01, Lydia Maniatis commented:

Very brief summary: The central claim about the discovery of a perceptual difference in the effect on a target of a narrow gradient versus a wide gradient when both are invisible has not been corroborated because:

a. the wide gradient is visible;

b. the effect of the narrow gradient is the very same as that which would occur in the absence of this gradient. The attempt to spin it differently is based on past experiments in different conditions using narrow but visible, not invisible, surrounds.