On 2015 Aug 28, Lydia Maniatis commented:

There seems to be a chicken and egg problem going on here as well. Here: https://www.youtube.com/watch?v=QORWM3Pl760, one of the authors seems to be defining "stimulus" as something that is constructed by the visual system, as opposed to a light -reflecting object in the world, and at the same time he is saying that the nature of this stimulus is determined in some way by its frequency of occurrence. But the latter has no existence prior to the former. There has to be some reference at some point to an objective situation, a stimulus in that sense, for the conversation to work.

Do the authors consider there to be a link between the topography of the light energy on the retina and the topography of the light energy in the world one moment before it strikes the retina?

On 2015 Aug 28, Lydia Maniatis commented:

The claims in this article (and related past publications) are not credible, and barely defended.

First, there is a little bit of sleight-of-hand in the authors' definition of “wholly empirical,” with the result that it includes both wholly acceptable and wholly unacceptable assertions. After describing their view that the visual system assigns perceptual values “without ever recovering or statistically estimating the properties of objects and conditions in the visual environment,” they label this view “wholly empirical” on the basis that it “depends entirely on feedback from trial and error experience.” But unlike the former claim, which is highly disputable, the latter seems simply to be a description of evolution by natural selection, a trial and error process in which adaptive features are preserved and less adaptive ones rejected. I cannot imagine that any scientist today would propose a mechanism for a biological function that they do not believe could have arisen via natural selection. Thus, it does not seem fair on the part of the authors to monopolize the concept for a specific set of proposals.

They reinforce this apparent monopoly by implying that a lack of correlation between the percept and the physical world is a necessary consequence of the link between perceptual processes and behavior: “...the biological feedback loop...would progressively order the basic visual qualities we perceive (apparent length, lightness, etc) according to their impact on biological (reproductive) success, rather than with the generative properties in the world (actual physical lengths, the reflective and illumination values underlying luminance, etc) [On what basis do the authors refer to reflectance and illumination as objective properties of the world (see below)?]” The “rather than” term is a verbal trick, implying that a perceptual representation that is correlated with (selected) physical properties would be less adaptive than one that is not. But such an assumption is not warranted.

The “evidence” for the position is also contingent on looseness in the supporting arguments. We are told, for example, that “since luminance measures the number of photons falling on the retina, common sense suggests that measurements of light intensity and perceived lightness should be proportional. Thus, if two surfaces return the same amount of light to the eye, they “should” be perceived as equally light.” This is a straw man. In the most extreme cases of same-luminance surfaces producing different lightness, the different surfaces, in addition to producing lightness percepts, also produce illumination percepts. Higher perceived lightness correlates with lower perceived illumination. The visual system is attempting to provide relative estimates for reflectance and illumination. This could not be achieved by directly representing luminance, whatever “common sense” might say. It is not correct to say that the visual system never correctly (or to a good approximations) represents relative reflectance/illumination values across surfaces. When it does, it is not by accident, it is what the system is designed (so to speak) to achieve. The idea that all images that elicit reflectance/illumination percepts do so on the basis of the frequency of evolutionary experience - behavioral responses to each particular stimulus, and their consequences for the reproductive success of the individual - is not credible. Even if we take apparent illumination out of the picture, the problem is just as big. Are we supposed to explain e.g. every Kanisza figure, on the basis of how frequently it occurred?

And, for that matter, the notion that “sensory stimuli [are] associated with biologically useful responses” does not justify the claim that stimulus frequency determines perceptual values, since a stimulus pattern may be very frequent but have little biological consequence, or rare but have a larger biological impact. So, in another sleight-of-hand, the authors have slipped an unwarranted frequency-biologically-relevant link into their argument.

In addition to the lightness example, the other piece of “evidence” offered is equally problematic on a number of counts. First, we are told that psychophysical experiments indicate that lines oriented at about thirty degrees from the vertical appear longer than lines at any other orientation. But the perceived length of a line is entirely contingent on the figure in which it is incorporated. So even if the authors claim that, over evolutionary time, line segments with the thirty degree orientation occurred (and were reacted to in an evolutionary dispositive way), more often than any other orientation, this would not explain why an oblique with a greater deviation from the vertical will reliably yield a longer percept than a vertical when incorporated in, e.g. a Shepard box. Even if we take the authors' natural scene statistics at face value, they are not relevant, since edges always occur in objects, and the shapes of these objects mediate the perception of length of an edge. Assumptions regarding the shapes of objects even determine whether or not a physically present or absent edge will or will not be seen.

The natural scene statistics offered are also not credible. Human ancestors varied differed greatly in size form the present, individual humans differ in height based on age, the eyes and body are constantly in motion, the distance to the object being viewed is constantly changing, and all of these things affect the orientation of the projection of a physical edge. Most of our time is arguably spent looking at close range, and at people. A statistical distribution developed on the basis of measurements from a fixed height at a relatively great distance of groomed gardens is arguably not a good approximation of human experience. Finally, are the authors really claiming that an isolated line segment of thirty degrees looks longer than others because it (supposedly) projected more frequently on the retinas of humans and their ancestors? If all percepts are drawn from frequency distributions, then how do we account for their qualitative differences? Are colors also labelled on the basis of the frequency that each collection of wavelengths occurs?

There is, finally, a fundamental contradiction in the argument that perceptual values don't ever recover or estimate properties of objects and conditions in the environment (I discuss this contradiction in Maniatis (in press). The problem is that if we choose to adopt this view, then we are not entitled to refer to the physical world as though we did, in fact, have knowledge about it. Thus, statements such as “consider the objective length of a line, e.g. the edge of a ruler...” are paradoxical because they imply that the authors have access to the very objective facts that they claim are inaccessible. It does no good to argue that we detect objective facts using objects called instruments, since the properties of these objects, too, are only accessible through perception. Given their position, the authors are not entitled to make any reference to the objective world and its properties, since such references directly contradict it.

On 2015 Aug 26, Lydia Maniatis commented:

In a search of the book "Perceiving Geometry," the term 'shape' comes up three times, twice in reference to the shape of distributions and once in a plain assertion that some neurons respond selectively to "higher-order stimulus characteristics...such as shape...". The assertion is not accompanied by arguments. The book has many references, are there any in particular that address the issue of shape?

The term or concept of 'figure-ground' also does not arise in the book. Do you consider the problems of perceptual organisation to have been solved on the basis of frequency distributions?

On 2015 Aug 24, Dale Purves commented:

Shape is a higher order construct of "size, distance, orientation" and geometrical factors such as the length of intervals (lines) and angles. See Howe and Purves "Perceiving Geometry" (Springer, 2005) for a full account and references to original papers.

On 2015 Aug 24, Lydia Maniatis commented:

In their summary, the authors state that:

"Visual perception is characterized by the basic qualities of lightness, brightness, color, size, distance, orientation, speed and direction of motion ordered over some range."

They have left out the most basic quality of all, and the one that largely mediates all the others: Shape. Was this an oversight, or a purposeful or principled omission?

They also say that: "These perceived qualities and their order within these ranges, however, do not align with reality."

They should clarify whether they mean for this second statement to apply also to shape.

On 2015 Aug 24, Lydia Maniatis commented:

In their summary, the authors state that:

"Visual perception is characterized by the basic qualities of lightness, brightness, color, size, distance, orientation, speed and direction of motion ordered over some range."

They have left out the most basic quality of all, and the one that largely mediates all the others: Shape. Was this an oversight, or a purposeful or principled omission?

They also say that: "These perceived qualities and their order within these ranges, however, do not align with reality."

They should clarify whether they mean for this second statement to apply also to shape.

On 2015 Aug 24, Dale Purves commented:

Shape is a higher order construct of "size, distance, orientation" and geometrical factors such as the length of intervals (lines) and angles. See Howe and Purves "Perceiving Geometry" (Springer, 2005) for a full account and references to original papers.

On 2015 Aug 26, Lydia Maniatis commented:

In a search of the book "Perceiving Geometry," the term 'shape' comes up three times, twice in reference to the shape of distributions and once in a plain assertion that some neurons respond selectively to "higher-order stimulus characteristics...such as shape...". The assertion is not accompanied by arguments. The book has many references, are there any in particular that address the issue of shape?

The term or concept of 'figure-ground' also does not arise in the book. Do you consider the problems of perceptual organisation to have been solved on the basis of frequency distributions?

On 2015 Aug 28, Lydia Maniatis commented:

The claims in this article (and related past publications) are not credible, and barely defended.

First, there is a little bit of sleight-of-hand in the authors' definition of “wholly empirical,” with the result that it includes both wholly acceptable and wholly unacceptable assertions. After describing their view that the visual system assigns perceptual values “without ever recovering or statistically estimating the properties of objects and conditions in the visual environment,” they label this view “wholly empirical” on the basis that it “depends entirely on feedback from trial and error experience.” But unlike the former claim, which is highly disputable, the latter seems simply to be a description of evolution by natural selection, a trial and error process in which adaptive features are preserved and less adaptive ones rejected. I cannot imagine that any scientist today would propose a mechanism for a biological function that they do not believe could have arisen via natural selection. Thus, it does not seem fair on the part of the authors to monopolize the concept for a specific set of proposals.

They reinforce this apparent monopoly by implying that a lack of correlation between the percept and the physical world is a necessary consequence of the link between perceptual processes and behavior: “...the biological feedback loop...would progressively order the basic visual qualities we perceive (apparent length, lightness, etc) according to their impact on biological (reproductive) success, rather than with the generative properties in the world (actual physical lengths, the reflective and illumination values underlying luminance, etc) [On what basis do the authors refer to reflectance and illumination as objective properties of the world (see below)?]” The “rather than” term is a verbal trick, implying that a perceptual representation that is correlated with (selected) physical properties would be less adaptive than one that is not. But such an assumption is not warranted.

The “evidence” for the position is also contingent on looseness in the supporting arguments. We are told, for example, that “since luminance measures the number of photons falling on the retina, common sense suggests that measurements of light intensity and perceived lightness should be proportional. Thus, if two surfaces return the same amount of light to the eye, they “should” be perceived as equally light.” This is a straw man. In the most extreme cases of same-luminance surfaces producing different lightness, the different surfaces, in addition to producing lightness percepts, also produce illumination percepts. Higher perceived lightness correlates with lower perceived illumination. The visual system is attempting to provide relative estimates for reflectance and illumination. This could not be achieved by directly representing luminance, whatever “common sense” might say. It is not correct to say that the visual system never correctly (or to a good approximations) represents relative reflectance/illumination values across surfaces. When it does, it is not by accident, it is what the system is designed (so to speak) to achieve. The idea that all images that elicit reflectance/illumination percepts do so on the basis of the frequency of evolutionary experience - behavioral responses to each particular stimulus, and their consequences for the reproductive success of the individual - is not credible. Even if we take apparent illumination out of the picture, the problem is just as big. Are we supposed to explain e.g. every Kanisza figure, on the basis of how frequently it occurred?

And, for that matter, the notion that “sensory stimuli [are] associated with biologically useful responses” does not justify the claim that stimulus frequency determines perceptual values, since a stimulus pattern may be very frequent but have little biological consequence, or rare but have a larger biological impact. So, in another sleight-of-hand, the authors have slipped an unwarranted frequency-biologically-relevant link into their argument.

In addition to the lightness example, the other piece of “evidence” offered is equally problematic on a number of counts. First, we are told that psychophysical experiments indicate that lines oriented at about thirty degrees from the vertical appear longer than lines at any other orientation. But the perceived length of a line is entirely contingent on the figure in which it is incorporated. So even if the authors claim that, over evolutionary time, line segments with the thirty degree orientation occurred (and were reacted to in an evolutionary dispositive way), more often than any other orientation, this would not explain why an oblique with a greater deviation from the vertical will reliably yield a longer percept than a vertical when incorporated in, e.g. a Shepard box. Even if we take the authors' natural scene statistics at face value, they are not relevant, since edges always occur in objects, and the shapes of these objects mediate the perception of length of an edge. Assumptions regarding the shapes of objects even determine whether or not a physically present or absent edge will or will not be seen.

The natural scene statistics offered are also not credible. Human ancestors varied differed greatly in size form the present, individual humans differ in height based on age, the eyes and body are constantly in motion, the distance to the object being viewed is constantly changing, and all of these things affect the orientation of the projection of a physical edge. Most of our time is arguably spent looking at close range, and at people. A statistical distribution developed on the basis of measurements from a fixed height at a relatively great distance of groomed gardens is arguably not a good approximation of human experience. Finally, are the authors really claiming that an isolated line segment of thirty degrees looks longer than others because it (supposedly) projected more frequently on the retinas of humans and their ancestors? If all percepts are drawn from frequency distributions, then how do we account for their qualitative differences? Are colors also labelled on the basis of the frequency that each collection of wavelengths occurs?

There is, finally, a fundamental contradiction in the argument that perceptual values don't ever recover or estimate properties of objects and conditions in the environment (I discuss this contradiction in Maniatis (in press). The problem is that if we choose to adopt this view, then we are not entitled to refer to the physical world as though we did, in fact, have knowledge about it. Thus, statements such as “consider the objective length of a line, e.g. the edge of a ruler...” are paradoxical because they imply that the authors have access to the very objective facts that they claim are inaccessible. It does no good to argue that we detect objective facts using objects called instruments, since the properties of these objects, too, are only accessible through perception. Given their position, the authors are not entitled to make any reference to the objective world and its properties, since such references directly contradict it.

On 2015 Aug 28, Lydia Maniatis commented:

There seems to be a chicken and egg problem going on here as well. Here: https://www.youtube.com/watch?v=QORWM3Pl760, one of the authors seems to be defining "stimulus" as something that is constructed by the visual system, as opposed to a light -reflecting object in the world, and at the same time he is saying that the nature of this stimulus is determined in some way by its frequency of occurrence. But the latter has no existence prior to the former. There has to be some reference at some point to an objective situation, a stimulus in that sense, for the conversation to work.

Do the authors consider there to be a link between the topography of the light energy on the retina and the topography of the light energy in the world one moment before it strikes the retina?