On 2016 Dec 29, Lydia Maniatis commented:

The title of this article indicates that the authors may have something to say about “lightness perception for surfaces moving through different illumination levels”, but leaves us in the dark what that might be.

The abstract isn’t much more illuminating. The somewhat vague message seems to be that the perceived lightness of a patch in the visual field depends on the structure of the “light field,” the “choice of fixation positions,” and whether the scene is viewed freely or not, and that “eye movements in [dynamic scenes and nonuniform light fields] are chosen to improve lightness constancy.”

Unfortunately and fatally absent from the terms of the discussion is any reference to shape. Yet, shape (i.e. the organization (segregation/unification, 3D interpretation), via the visual process, of the points in the retinal projection into perceived forms) is the only available means to the goal of creating percepts of lightness as well as relative illumination of surfaces. This is obvious with respect to the authors' sitmuli, which are images on a computer screen. The luminance structure of the light emitting points on that screen is the only information the visual system has to work with, and unless those points are grouped and boundaries and depth relations inferred there is no basis for designating continuous surfaces, their lightness, their relative illumination. Whether areas of the visual are interpreted as changing in reflectance or illumination is contingent on which parts of the field are eligible to be grouped into perceived physical units, with a homogeneous surface.

In other words: When the luminance of a surface in a part of the visual field changes, (e.g. from lighter to darker), the change may be interpreted as being due to a change in illumination of a surface in that location, a change in the color of the surface at that location, the presence of a fog overlying the surface at that location, etc., or a combination of these possibilities. How is the solution (the percept) arrived at? For example, at the lower left side of Toscani et al’s (2016) Figure 1, an edge between a dark area (the “wall)” and a lighter area (the “side of a cube”) to its right is perceived as a lightening in terms of both perceived illumination and perceived reflectance) while a change from same lighter area to a darker area to its right is seen as a change in illumination only. The reason is structural, based on the very principles of organization not mentioned by the authors.

The consequence of the failure to consider principles of organization in any study of lightness perception is that ANY resulting claims can be immediately falsified. It is impossible to predict how a surface will look when placed in any given location in the visual field by referring only to the distribution of incident illumination, since this information doesn’t in the least allow us to predict luminance structure. And a description of luminance structure doesn’t help us if we don’t consider visual principles of organization. The former fact should be particularly obvious to people using uniformly illuminated pictorial stimuli, whether on a page or on a screen, which produce impressions of non-uniform illumination. Like reflectance, the perception of illumination is constructed, it isn’t an independent variable for vision; so it makes no sense, in the context of perception experiments, to refer to it as though it is – as the authors do in the phrase “moving through different illumination levels” - especially if we aren’t even talking about actual illumination levels, but only visually-constructed ones! The perception of changing illumination levels is the flip side to the perception of unchanging surfaces, and vice versa. Like lightness, perceived illumination is dependent on principles of organization, starting with figure/ground segregation.

So, for example, when the authors say that the brightest parts of a (perceived) surface’s luminance distribution is “an efficient…heuristic for the visual system to achieve accurate…judgments of lightness,” we can counter (falsify) with the glare illusion (http://www.opticalillusion.net/optical-illusions/grey-glow-illusion-the-glare-effect/) in which the brightest area is not perceived as the plain view color of the surface, which appears black and obscured by a glare or bright fog.

With respect to eye movements and fixation: It seems to be the case that fixations are the product, not the cause, of perceptual solutions. For example, it has been shown that while viewing the Muller-Lyer illusion, eye movements trace a longer path when we’re looking at the apparently longer figure and vice versa. Another problem with the claim that eye movements have a causal role by sampling “more relevant” parts of the field is that all parts of the field are taken into account in the generation of a percept, e.g. in order for the visual system to conclude that a particular patch is the lightest part of a homogeneously-colored but differently-illuminated physical unit, rather than a differently colored patch on a different unit. Since the perceived relative lightness/illumination of that particular patch is related to the perceived lightness/illumination of the whole visual field, isolating that patch by fixation can’t be uniquely informative. As we know, reduction conditions can transform the perception of surfaces.

I would note that the emphasis on “lightness constancy” rather than “principles of lightness perception” is common but ill-conceived. With respect to understanding perception, understanding lightness constancy is no more informative than understanding lightness inconstancy. (For a great example, complete with movement, of lightness INconstancy, see https://www.youtube.com/watch?v=z9Sen1HTu5o). In either case, what is constant are the underlying perceptual principles; to understand one effect is to understand the other. This is another reason the claim that eye movements are chosen “to improve lightness constancy” is ill-conceived. Only an all-knowing homunculus can know, a priori, which areas of the visual field represent stimulation from physical surfaces with constant reflectance x, which represent physical surfaces obstructed by fog or in shadow, which areas represent physical surfaces that are actually changing in their light reflecting properties (a squid, for example - do we want to improve his or her constancy?), etc. The visual system has to go where the evidence goes, as interpreted via the evolved process. This process achieves veridicality – e.g. seeing surface properties as unchanging when they’re unchanging, and as changing when they’re changing - in typical conditions.

Ironically, observers in Toscani et al’s (2016) experiments are not perceiving surfaces veridically, since, for example, parts of the screen surface that are actually varying in their color are perceived as unchanging in that respect, while, correspondingly, they are seen (incorrectly) as experiencing changing illumination. So we’re actually talking about the converse of lightness constancy; the authors are equating a physical surface that is unchanging in its light-reflecting properties, but experiencing changing illumination, with a surface (a section of the screen) that is actually changing in its light reflecting/emitting properties independently of incident illumination, which is constant. In the former case, seeing the surface as unchanging parallels the physical situation, while in the latter case it is opposite to the physical situation. Calling both situations examples of “lightness constancy” only confuses the issue, which is: “Why does a retinal projection with a particular luminance structure result in patch x looking the way it does.” The question, again, cannot be answered reliably without invoking principles of organization, i.e. the consequences of that luminance structure for perceived shape.

Short version: Shape.

On 2016 Dec 29, Lydia Maniatis commented:

The title of this article indicates that the authors may have something to say about “lightness perception for surfaces moving through different illumination levels”, but leaves us in the dark what that might be.

The abstract isn’t much more illuminating. The somewhat vague message seems to be that the perceived lightness of a patch in the visual field depends on the structure of the “light field,” the “choice of fixation positions,” and whether the scene is viewed freely or not, and that “eye movements in [dynamic scenes and nonuniform light fields] are chosen to improve lightness constancy.”

Unfortunately and fatally absent from the terms of the discussion is any reference to shape. Yet, shape (i.e. the organization (segregation/unification, 3D interpretation), via the visual process, of the points in the retinal projection into perceived forms) is the only available means to the goal of creating percepts of lightness as well as relative illumination of surfaces. This is obvious with respect to the authors' sitmuli, which are images on a computer screen. The luminance structure of the light emitting points on that screen is the only information the visual system has to work with, and unless those points are grouped and boundaries and depth relations inferred there is no basis for designating continuous surfaces, their lightness, their relative illumination. Whether areas of the visual are interpreted as changing in reflectance or illumination is contingent on which parts of the field are eligible to be grouped into perceived physical units, with a homogeneous surface.

In other words: When the luminance of a surface in a part of the visual field changes, (e.g. from lighter to darker), the change may be interpreted as being due to a change in illumination of a surface in that location, a change in the color of the surface at that location, the presence of a fog overlying the surface at that location, etc., or a combination of these possibilities. How is the solution (the percept) arrived at? For example, at the lower left side of Toscani et al’s (2016) Figure 1, an edge between a dark area (the “wall)” and a lighter area (the “side of a cube”) to its right is perceived as a lightening in terms of both perceived illumination and perceived reflectance) while a change from same lighter area to a darker area to its right is seen as a change in illumination only. The reason is structural, based on the very principles of organization not mentioned by the authors.

The consequence of the failure to consider principles of organization in any study of lightness perception is that ANY resulting claims can be immediately falsified. It is impossible to predict how a surface will look when placed in any given location in the visual field by referring only to the distribution of incident illumination, since this information doesn’t in the least allow us to predict luminance structure. And a description of luminance structure doesn’t help us if we don’t consider visual principles of organization. The former fact should be particularly obvious to people using uniformly illuminated pictorial stimuli, whether on a page or on a screen, which produce impressions of non-uniform illumination. Like reflectance, the perception of illumination is constructed, it isn’t an independent variable for vision; so it makes no sense, in the context of perception experiments, to refer to it as though it is – as the authors do in the phrase “moving through different illumination levels” - especially if we aren’t even talking about actual illumination levels, but only visually-constructed ones! The perception of changing illumination levels is the flip side to the perception of unchanging surfaces, and vice versa. Like lightness, perceived illumination is dependent on principles of organization, starting with figure/ground segregation.

So, for example, when the authors say that the brightest parts of a (perceived) surface’s luminance distribution is “an efficient…heuristic for the visual system to achieve accurate…judgments of lightness,” we can counter (falsify) with the glare illusion (http://www.opticalillusion.net/optical-illusions/grey-glow-illusion-the-glare-effect/) in which the brightest area is not perceived as the plain view color of the surface, which appears black and obscured by a glare or bright fog.

With respect to eye movements and fixation: It seems to be the case that fixations are the product, not the cause, of perceptual solutions. For example, it has been shown that while viewing the Muller-Lyer illusion, eye movements trace a longer path when we’re looking at the apparently longer figure and vice versa. Another problem with the claim that eye movements have a causal role by sampling “more relevant” parts of the field is that all parts of the field are taken into account in the generation of a percept, e.g. in order for the visual system to conclude that a particular patch is the lightest part of a homogeneously-colored but differently-illuminated physical unit, rather than a differently colored patch on a different unit. Since the perceived relative lightness/illumination of that particular patch is related to the perceived lightness/illumination of the whole visual field, isolating that patch by fixation can’t be uniquely informative. As we know, reduction conditions can transform the perception of surfaces.

I would note that the emphasis on “lightness constancy” rather than “principles of lightness perception” is common but ill-conceived. With respect to understanding perception, understanding lightness constancy is no more informative than understanding lightness inconstancy. (For a great example, complete with movement, of lightness INconstancy, see https://www.youtube.com/watch?v=z9Sen1HTu5o). In either case, what is constant are the underlying perceptual principles; to understand one effect is to understand the other. This is another reason the claim that eye movements are chosen “to improve lightness constancy” is ill-conceived. Only an all-knowing homunculus can know, a priori, which areas of the visual field represent stimulation from physical surfaces with constant reflectance x, which represent physical surfaces obstructed by fog or in shadow, which areas represent physical surfaces that are actually changing in their light reflecting properties (a squid, for example - do we want to improve his or her constancy?), etc. The visual system has to go where the evidence goes, as interpreted via the evolved process. This process achieves veridicality – e.g. seeing surface properties as unchanging when they’re unchanging, and as changing when they’re changing - in typical conditions.

Ironically, observers in Toscani et al’s (2016) experiments are not perceiving surfaces veridically, since, for example, parts of the screen surface that are actually varying in their color are perceived as unchanging in that respect, while, correspondingly, they are seen (incorrectly) as experiencing changing illumination. So we’re actually talking about the converse of lightness constancy; the authors are equating a physical surface that is unchanging in its light-reflecting properties, but experiencing changing illumination, with a surface (a section of the screen) that is actually changing in its light reflecting/emitting properties independently of incident illumination, which is constant. In the former case, seeing the surface as unchanging parallels the physical situation, while in the latter case it is opposite to the physical situation. Calling both situations examples of “lightness constancy” only confuses the issue, which is: “Why does a retinal projection with a particular luminance structure result in patch x looking the way it does.” The question, again, cannot be answered reliably without invoking principles of organization, i.e. the consequences of that luminance structure for perceived shape.

Short version: Shape.