On 2017 May 31, Lydia Maniatis commented:

p.s. The fact that references to decomposition into layers are references to perceptual facts also means that simply asking people to qualitatively describe their experience is a necessary and sufficient means for deciding between the authors' two proposed "models" of these experiences, as both refer to perceptual dimensions and only perceptual dimensions (regardless of the presumed physical causes). Thus: "The hybrid decomposition thus suggests that the human visual system may decompose compound BRDF's into two perceptual components..."

As this simple operation of asking people whether they perceive layers, etc, was not performed, the authors try to guess at whether their results are consistent with one or another description. "To this end, we have performed a number of linear regressions on our experimental data..." but of course "it should be noted that the evidence derived from the current experiment is mixed, so additional, more targeted [less confounded, better rationalized] experiments would be required..." But still, they go on to elaborate a post hoc description of results, for what its worth.

On 2017 May 31, Lydia Maniatis commented:

What do Vangorp et al (2017) have to say about the “perception of hazy gloss” (and why don’t they break the news up front)?

If the final statement of their discussion is any indication, it seems they ended up where they began,: “An exciting avenue of future work is thus to understand how the human visual system encodes hazy gloss and distinguishes it from glare in images.” (p. 14)

But this study also used images, and its stated goal was “to investigate the conditions in which haze gloss is encoded in the perceptual representation of glossy materials.” (p. 2)

As reported, the experiments performed were hopelessly confounded, their results uninterpretable, and their rationale conceptually confused. As is too often the case, all the authors can do under the circumstances is to describe their results in a purely self-referential way, offering vague and incoherent post hoc speculations as to the causes of their particular outcomes with respect to their particular conditions. Thus:

Experiment 1

Experiment 1 doesn’t actually evaluate perception of a quality that observers would necessarily describe (and which the authors vaguely describe) as “hazy gloss,” but, rather, the ability to match certain parameters of certain graphic images.

Results are highly variable and don’t possess a form the authors did or could have predicted or are able to interpret: “The space of average matches is severely compressed towards the middle of the valid space. The variance of these average matches is so large that representing it as ellipses would clutter the graph.” (p. 6) Vague speculations follow:

“One interpretation could be that a perceptual decomposition of sharp and wide specular components [being graphic images, the stimuli do not actually contain specular components, unless we are talking about reflections on the screen] is difficult…” (p. 6)

“However…these results do not prove that participants perceptually separated the material into two components.” This is partly due to confounds: “In each condition only a single parameter is kept correct, while the other three vary…This means that observers might attend to the narrow shading features to judge…Observers might also attend to other shading features…” (p. 6)

“Since there is no correct match [in the single component condition], success must be defined as the consistency between participants and consistency with plausible models…” (p. 6) There is, however, no discussion about the plausibility of models.

“The polar plots show skewed and even bimodal probability distributions…” (p. 6)

“The findings suggest that, at least for some parameter ranges, particularly in the upper edges of the stimulus space, participants are able to match the properties of two-component specular materials [assuming that the graphics program accurately mimics the optical effects of the referenced specular conditions].” (p. 6)

It should be clear that the muddled, ad hoc conclusions, to the extent that they are intelligible at all, refer narrowly to the particular experimental conditions and stimuli; the conceptual analysis is not such as to allow us to draw out implications outside of this self-referential space.

(And yet another confound is referred to in the description of Experiment 2: “The diffuse component in Experiment 1 could have obscured the intended differences in the specular [so-called] components…we performed the 4AFC task in a darkened room …in order to further optimize the performance of the observers.” (p. 9-10) (Further optimize....as in “even more perfect...”)

Experiment 2 This is a forced choice discrimination task. Forced choice means we don’t really care about what observers are actually perceiving, only that they give some manageable responses.

Again, there is much post hoc speculation (very briefly, to give the style of the account): “For all three classes of distractors there are ranges of corresponding two component materials that are easily distinguishable.” “The main differences between conditions may be explained by…This would explain the opposite slopes of the performance curves…However, Figure 10 suggests that a simple interpretation…is unlikely… Nevertheless, the discrimination experiment suggests…”

In any case: “However, this in itself does not tell us about the subjective interpretation of these differences. Does the [presumed, and variable] ability to distinguish between different BRDF’s reflect a distinct perceptual parameter, or are the image differences detectable but not interpretable?”

Answering this question is a job for Experiment 3, which, naturally, employs entirely different graphic material effects (silvery as opposed to plasticky...Why?) with different graphic lighting effects. (“This experiment used a different lighting environment representing another church interior, Galileo’s Tomb.” Well, as long as it's a church...We’re not told why, or how such a change might be expected to affect the phenomenon of interest.) The authors also give us various other details about the parameters employed, but what’s the point?

The results of Experiment 3 are as informative as those of the other two. The arguably tautological conclusion is drawn that “the subjective impression of hazy or layered materials is crucially associated with a “bloom” or “halo” around sharp reflections.” (p. 10). (The use of quotation marks indicates that a subjective impression is being explained in terms of a similar, differently-expressed subjective impression.)

The theoretical discussion generally is conceptually confused, essentially failing to make the crucial distinctions among the distal object, the proximal (retinal) stimulation and the perceptual experience. This is illustrated, for example, by the following phrase: “It is highly unlikely that the human visual system refers to only two specific angles [of reflections] to infer haze gloss…” The angles being referred to have no retinal correlate, so of course they could not be the cause of the percept. These angles of reflection may have a correlate in the percept, but this cannot be used as an explanatory principle for the formation of the percept.

Another funny thing the authors do (they’re not the only ones) is to express a perceptual fact as though it were a theoretical one, treating a perceptual fact as something yet to be ascertained experimentally. Thus:

“We therefore propose that the visual system may represent haze by decomposing the material response into two distinct components, or causal layers. One obvious choice for such a decomposition would be the physical components themselves (i.e. the broad and narrow specular terms). [Note that, as usual, the authors are inappropriately treating the physical features of the distal stimulus as explanatory variables]. This decomposition separates the composite reflection into two [and the authors can only be referring to perceived layers, as in their later reference to transparency] superimposed layers…much like the decomposition of image patches...in transparency perception. This would be broadly [a high standard!] consistent with our findings that observers base many of their judgments on the narrow component.”

We don’t have to guess as to whether the percept contains perceived layers; this is, by definition, a subjective fact and can be directly ascertained by personal experience and/or by asking other people what they see. If they answer in the positive, then any experiment that indicated otherwise would obviously be in error.

On 2017 May 31, Lydia Maniatis commented:

What do Vangorp et al (2017) have to say about the “perception of hazy gloss” (and why don’t they break the news up front)?

If the final statement of their discussion is any indication, it seems they ended up where they began,: “An exciting avenue of future work is thus to understand how the human visual system encodes hazy gloss and distinguishes it from glare in images.” (p. 14)

But this study also used images, and its stated goal was “to investigate the conditions in which haze gloss is encoded in the perceptual representation of glossy materials.” (p. 2)

As reported, the experiments performed were hopelessly confounded, their results uninterpretable, and their rationale conceptually confused. As is too often the case, all the authors can do under the circumstances is to describe their results in a purely self-referential way, offering vague and incoherent post hoc speculations as to the causes of their particular outcomes with respect to their particular conditions. Thus:

Experiment 1

Experiment 1 doesn’t actually evaluate perception of a quality that observers would necessarily describe (and which the authors vaguely describe) as “hazy gloss,” but, rather, the ability to match certain parameters of certain graphic images.

Results are highly variable and don’t possess a form the authors did or could have predicted or are able to interpret: “The space of average matches is severely compressed towards the middle of the valid space. The variance of these average matches is so large that representing it as ellipses would clutter the graph.” (p. 6) Vague speculations follow:

“One interpretation could be that a perceptual decomposition of sharp and wide specular components [being graphic images, the stimuli do not actually contain specular components, unless we are talking about reflections on the screen] is difficult…” (p. 6)

“However…these results do not prove that participants perceptually separated the material into two components.” This is partly due to confounds: “In each condition only a single parameter is kept correct, while the other three vary…This means that observers might attend to the narrow shading features to judge…Observers might also attend to other shading features…” (p. 6)

“Since there is no correct match [in the single component condition], success must be defined as the consistency between participants and consistency with plausible models…” (p. 6) There is, however, no discussion about the plausibility of models.

“The polar plots show skewed and even bimodal probability distributions…” (p. 6)

“The findings suggest that, at least for some parameter ranges, particularly in the upper edges of the stimulus space, participants are able to match the properties of two-component specular materials [assuming that the graphics program accurately mimics the optical effects of the referenced specular conditions].” (p. 6)

It should be clear that the muddled, ad hoc conclusions, to the extent that they are intelligible at all, refer narrowly to the particular experimental conditions and stimuli; the conceptual analysis is not such as to allow us to draw out implications outside of this self-referential space.

(And yet another confound is referred to in the description of Experiment 2: “The diffuse component in Experiment 1 could have obscured the intended differences in the specular [so-called] components…we performed the 4AFC task in a darkened room …in order to further optimize the performance of the observers.” (p. 9-10) (Further optimize....as in “even more perfect...”)

Experiment 2 This is a forced choice discrimination task. Forced choice means we don’t really care about what observers are actually perceiving, only that they give some manageable responses.

Again, there is much post hoc speculation (very briefly, to give the style of the account): “For all three classes of distractors there are ranges of corresponding two component materials that are easily distinguishable.” “The main differences between conditions may be explained by…This would explain the opposite slopes of the performance curves…However, Figure 10 suggests that a simple interpretation…is unlikely… Nevertheless, the discrimination experiment suggests…”

In any case: “However, this in itself does not tell us about the subjective interpretation of these differences. Does the [presumed, and variable] ability to distinguish between different BRDF’s reflect a distinct perceptual parameter, or are the image differences detectable but not interpretable?”

Answering this question is a job for Experiment 3, which, naturally, employs entirely different graphic material effects (silvery as opposed to plasticky...Why?) with different graphic lighting effects. (“This experiment used a different lighting environment representing another church interior, Galileo’s Tomb.” Well, as long as it's a church...We’re not told why, or how such a change might be expected to affect the phenomenon of interest.) The authors also give us various other details about the parameters employed, but what’s the point?

The results of Experiment 3 are as informative as those of the other two. The arguably tautological conclusion is drawn that “the subjective impression of hazy or layered materials is crucially associated with a “bloom” or “halo” around sharp reflections.” (p. 10). (The use of quotation marks indicates that a subjective impression is being explained in terms of a similar, differently-expressed subjective impression.)

The theoretical discussion generally is conceptually confused, essentially failing to make the crucial distinctions among the distal object, the proximal (retinal) stimulation and the perceptual experience. This is illustrated, for example, by the following phrase: “It is highly unlikely that the human visual system refers to only two specific angles [of reflections] to infer haze gloss…” The angles being referred to have no retinal correlate, so of course they could not be the cause of the percept. These angles of reflection may have a correlate in the percept, but this cannot be used as an explanatory principle for the formation of the percept.

Another funny thing the authors do (they’re not the only ones) is to express a perceptual fact as though it were a theoretical one, treating a perceptual fact as something yet to be ascertained experimentally. Thus:

“We therefore propose that the visual system may represent haze by decomposing the material response into two distinct components, or causal layers. One obvious choice for such a decomposition would be the physical components themselves (i.e. the broad and narrow specular terms). [Note that, as usual, the authors are inappropriately treating the physical features of the distal stimulus as explanatory variables]. This decomposition separates the composite reflection into two [and the authors can only be referring to perceived layers, as in their later reference to transparency] superimposed layers…much like the decomposition of image patches...in transparency perception. This would be broadly [a high standard!] consistent with our findings that observers base many of their judgments on the narrow component.”

We don’t have to guess as to whether the percept contains perceived layers; this is, by definition, a subjective fact and can be directly ascertained by personal experience and/or by asking other people what they see. If they answer in the positive, then any experiment that indicated otherwise would obviously be in error.

On 2017 May 31, Lydia Maniatis commented:

p.s. The fact that references to decomposition into layers are references to perceptual facts also means that simply asking people to qualitatively describe their experience is a necessary and sufficient means for deciding between the authors' two proposed "models" of these experiences, as both refer to perceptual dimensions and only perceptual dimensions (regardless of the presumed physical causes). Thus: "The hybrid decomposition thus suggests that the human visual system may decompose compound BRDF's into two perceptual components..."

As this simple operation of asking people whether they perceive layers, etc, was not performed, the authors try to guess at whether their results are consistent with one or another description. "To this end, we have performed a number of linear regressions on our experimental data..." but of course "it should be noted that the evidence derived from the current experiment is mixed, so additional, more targeted [less confounded, better rationalized] experiments would be required..." But still, they go on to elaborate a post hoc description of results, for what its worth.