So let's use this image as an example, and we're taking little patches from it.

And those patches, well they're square with a central target.

Before we used circular patches, but it's the same idea.

The template that we're using to sample those images are effectively this.

A surround that can be either dark or light, and

we're asking what's the luminance that appears in the center of those surrounds

when we sample again and again from natural images with little templates

that match the phenomenon that we're trying to explain.

So here's the dark surround, here's the light surround and

the question mark indicates what we're asking.

We' re asking how often when there's a dark surround,

is there a luminance of some value in the central patch?

And how often in a light surround,

is there a value of some luminance in a central patch?

And what's the difference between those statistics

of a luminance in a dark surround in the central patch and

a luminance in a light surround in the central patch?

And you might expect there to be a difference, and

in fact, there is a difference, and why is that?

Well, look at these templates here.

Again these are templates that we're going to sample the images from.

Not once or twice, but millions and

millions of times in thousands and thousands of different images

to get an idea of what human experience has always been.

Well, if you look at this or

any scene, you'll be aware as in the two little template examples,

that very often the center patch, the luminance

of the center patch is going to be similar to the luminance of the surround.

The luminance of the center patch is going to be similar to the luminance of

the surround.

Why is that?

It's obvious why it is.

Of course that won't always be the case.

There will be many examples when that's not the case.

But on average, it's the case that when you have a little template like

this that the stuff, surrounding stuff in the middle, is going to be the same.

The stuff surrounding the stuff in the middle, and

by stuff I mean the luminance, is going to be the same.

Why, that is simply because that's the way images come from the world onto

our retinas, or that's the way the physical world is constructed.

That nearby stuff in any image tends to be the same stuff.

And therefore generating the same luminance at any moment in time.

So there's a simple explanation of why you would expect a difference

between the luminances that we extract from this kind of analysis for

the center patch in a dark surround, and the center patch in a lighter surround.

Again, this is the object of the exercise.

We want to explain why empirically, this discrepancy might come to be.

And here's the result of that kind of analysis, and

it gives you just what you would have expected based on what I just said.

Here is the arcs around, and the question mark indicates the distribution,

the frequency of the currents of luminances in a dark surround,

versus the frequency of occurrence of luminances in a light surround.

And it's obvious from these probability distributions of the target,

of the question mark luminance, that when the dark surround is analyzed,

the frequency of occurrence tends to be a darker

luminance, a less intense luminance, I shouldn't say darker luminance,

that's perceptual, but a less intense luminance than

the same analysis for the central patch in a light surround.

Again, the reason's obvious.

Generally speaking, statistically there's going to be similar stuff in the center.

And if the surround is dark,

the stuff in the center is going to be darker, less intense in its luminance.

And so conversely with a more intense luminance surround, the center patch is

generally going to be more illuminated as you see here in this statistical analysis.

Well remember, what we're trying to do here, is explain

why it is that we see the center patch

in the dark surround as different from the center patch in light surround.

So here, we come back to the phenomenon that we're trying to explain

this discrepancy between equiluminant patches in the dark and the lights around.

And these are equiluminant patches.

Again, now they don't look like the same to you because we're

being influenced by the same statistical

empirical evidence that explains the stimulus that we started with.

Remember the object of the exercise is to explain this and

we're now taking these templates which were effectively that

stimulus and asking why it is that they look different.

So they do look different and that's what we see in this,

these again are equiluminant patches.

And what we've done now is just take that same evidence of the frequency

of occurrence of luminance values in the center of these patches,

and we've now re-plotted that as cumulative probability distributions.

Asking, over the course of human history, what has been the frequency

of occurrence of luminance values in a dark surround and a light surround?

And when you put a patch that's equiluminant, so

this T is this patch, and this T is this patch,

and these Ts are equal to each other.

We put these same luminance patches in a dark surround or

a light surround, we see them as different.

And this explains why that is.

You can see that the same luminance value is going to have a different

empirical rank in a dark surround, Than it has in the light surround.

So in the light surround,

the empirical rank of this center patch is much lower.

It's occurred much less often than the same patch in a dark surround.

So your experience over the eons of human evolution is that when a patch of

this particular luminance is stuck in a dark surround versus the light surround,

the rank of your experience has been very different for those two patches.

You've much more often seen the same luminance value

in a dark surround than that value in a light surround.

And the idea is that the explanation of this difference

is that the two identical patches in these two contexts,

dark and light, occur because the perceived value or

subjective value, doesn't track physical movements,

it tracks the significance of our perception for behavioral success.

We can't see the physical luminance.

If we track their significance of these patches for

behavioral success based not on their physical properties, but

on their perceptual ranks, well we could get around the inverse problem.

And use that information in a world in which luminances are hidden from us,

to see stuff in a way, lightness and brightness,

that serves our biological need to succeed,

without having to know the actual luminances that are in these patches.

Which, as I said many times now, are inaccessible to us.

So this makes an interesting point about the physical world,

which I think we wouldn't have thought about otherwise.

And that is that the role of the physical world is really an arena that

provides feedback by virtue of the success or failure of our behavior.

Ultimately of course, that's in Darwanian terms of evolution by natural selection.

But the relative success that we need to get by in the world with luminances

that are hidden from us, and yet nonetheless can provide us empirical

information that allows us to succeed is remarkable.

The physical world is acting as an arena in which we're testing

empirically by trial and error.

Do we succeed or

not if we rank the perceptual lightness in relation to the physical luminance

in terms of its frequency of occurrence over the eons of human evolution?

So, a hard argument, and I'm going to come back to it again and

again in other context.

And I think it begins to sort of get across the idea

that you can explain discrepancies, and there are many of them that

we've talked about already, the discrepancies between the real world and

the subjective world of perception in this way.

An Empirical Explanation Based on Accumulated Experience (part 2)

Visual Perception and the Brain

Conheça os instrutores