Visual illusion, control of brightness

https://twitter.com/AkiyoshiKitaoka/status/1028473566193315841

Generally, we can assume that the perception signal (p) is highly correlated
to the input quantity (qi), and we can assume a simple input function.

In this case, the researcher in the video controls the brightness of the
moving square by varying its background brightness. So, the physical
brightness of the square is NOT controlled (it is not changing), while the
perception of the brightness is controlled.

[From Bruce Nevin (20190906.11:43 ET)]

Individual sensor cells in the retina detect brightness at the intensity level. That would be a simple input function generating p “highly correlated to the input quantity” qi,** as expected. I doubt that subjective awareness of intensity is able to drill down to individual sensors in the retina, and consequently I am not surprised that reported brightness is affected by contrast with context. Subjective report of brightness is a function of the retina as a whole, or anyway regions of it, where there is extensive cross-connection among the sensor cells. We get edge detection in the retina, which is a function of relative brightness. https://www.jneurosci.org/content/26/51/13250

···

/Bruce

On Fri, Sep 6, 2019 at 9:59 AM adam.matic@gmail.com csgnet@lists.illinois.edu wrote:

https://twitter.com/AkiyoshiKitaoka/status/1028473566193315841

Generally, we can assume that the perception signal § is highly correlated

to the input quantity (qi), and we can assume a simple input function.

In this case, the researcher in the video controls the brightness of the

moving square by varying its background brightness. So, the physical

brightness of the square is NOT controlled (it is not changing), while the

perception of the brightness is controlled.

[Rick Marken 2019-09-06_10:15:10]

https://twitter.com/AkiyoshiKitaoka/status/1028473566193315841

Generally, we can assume that the perception signal § is highly correlated

to the input quantity (qi), and we can assume a simple input function.

In this case, the researcher in the video controls the brightness of the

moving square by varying its background brightness. So, the physical

brightness of the square is NOT controlled (it is not changing), while the

perception of the brightness is controlled.

RM: Great find Adam!

RM: This is a great demonstration of the control of a different perceptual variable than the one that is assumed to be controlled. In this demo, when a person controls what is verbally described as “brightness”, what they are actually controlling is something like the ratio of the luminance level of the moving square to the luminance level of the current background, not the absolute luminance level of the moving square. Â

RM: I have a demo on the net that shows how this would affect what an observer sees as what a person is controlling:

https://www.mindreadings.com/ControlDemo/Instruct.html Â

RM: In this demo the person is asked to control the size of a circle, keeping it constant. It looks like the person is, indeed, following instructions and controlling the size of the circle as long as the context circles remain a constant size. But when the context starts to vary, suddenly it looks like the person is no longer controlling the size of the circle because the size of the circle now starts varying considerably, as can be seen in the graphs plotted at the end of the demo.Â

RM: If one thinks of the size of the circle as the controlled variable, qi, then it looks like p goes from being a good to a poor representation of qi. But, of course, what is actually going on is that the size of the circle was never qi; qi was always a perception that was a function of the relation of the physical size of the circle, v1, to the physical size of the context, v2. So when the subject is controlling the size of the circle they are controlling something like v1/v2, not just v1. But it looks like they are controlling v1 when v2 is constant.Â

RM: I think this is a good demonstration of why we have to identify qi using the test for the controlled variable. When the context is constant it looks like qi = v1. But when we vary the context we see that qi is more like v1/v2. And this is what PCT research is all about; figuring out what variables organisms are controlling when they are behaving in various ways.

RM: Thanks for posting this!

BestÂ

Rick

···

On Fri, Sep 6, 2019 at 6:59 AM adam.matic@gmail.com csgnet@lists.illinois.edu wrote:


Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

[Martin Taylor 2019.09.06.11.46]

https://twitter.com/AkiyoshiKitaoka/status/1028473566193315841

Generally, we can assume that the perception signal (p) is highly correlated
to the input quantity (qi), and we can assume a simple input function.

I'd make a slight edit: "we can assume" -> "we assume" if you use the word "correlated". I don't think the assumption is justified in general, given the ordinary behaviour of neurons, which is to respond strongly to edge contrast in space or time, but to settle toward a neutral firing rate away from the edges between areas of different consciously perceived values. The illusion depicted is a very old one. Somewhere some decades ago when I was working in visual perception I saw a picture of a 12th or 13th century Korean celadon cup or vase or something. This cup had a double version of the effect, using a a lightness modulation down on the sky side of the edge of the moon and up on the moon side. The moon looked very bright against a dark sky although the extended areas of moon and sky were the same lightness.

Even individual rods and cones respond strongly to onsets and offsets followed by a return toward a neutral firing rate. And you can perceive the effect at levels high enough that you cannot control the perception, such as your perception of the political climate, which always seems to return to "this is just normal" a year or three after an election in which everything changed.

In this case, the researcher in the video controls the brightness of the
moving square by varying its background brightness.

I didn't see any video. perhaps because I refuse to get involved with any social media like Facebook and Twitter. But the still images make very clear what the experiment must be.

So, the physical
brightness of the square is NOT controlled (it is not changing), while the
perception of the brightness is controlled.

Correct. And what do you derive from that observation, given that by definition (at least in the often copied square control loop diagram) qi is whatever the environment provides to the sensors that eventually reaches the perceptual function to create the controlled perception?

I actually programmed something very similar a few years ago when I was trying to set up an experiment on tolerance levels when controlling more than two dimensions with a two-dimensional mouse. My display was a disk of one colour and brightness within a wide ring of a different colour and brightness. There were other things to control apart from the match between the disk colour and brightness to that of a disk inside a mid-grey ring, such as (I forget what they actually were, so these are just ones I might have used) the pitch and loudness of a tone to be kept steady when bracketed by tone bursts of different intensities and frequencies, and things like that. The upshot is that I find it hard to believe that "qi" is EVER independent of context, whether you are talking about the lightness of a surface or the level of democracy in the current political climate.

But I can't think of a clean way to track the effects of temporal context, which are sometimes called "adaptation". The grey patch would seem dark after that patch had been brightly lit, or bright after it had been for a while kept dark. One thing we do know is that our vision absolutely requires that the input to each retinal sensor keeps changing. If the eye is artificially inhibited from gross movements and from continuing its normal microtremor at around 60Hz, object edges and then entire objects disappear and the visual field begins to take on a uniform whitish mid-brightness aspect, as though you were in the middle of a dense cloud. I find i can sometimes hold my eye still enough for a while to lose some of the edges of objects in my visual field.

Martin
PS. My own answer to the question I posed is that qi is not the photon density across the square. As Rick so often insists without using the word, qi is a vector that describes all the inputs from the environment that influence the controlled perception.

···

On 2019/09/6 9:58 AM, adam.matic@gmail.com (via csgnet Mailing List) wrote: