telephone game/memes

[kevinquinnyo]

I've been thinking about the concept of the telephone game/evolution, etc a lot lately.

Consider this scenario:

You very briefly flash a color to someone from a color chart of some kind with thousands of colors, ranging the full spectrum. Then you give them the wheel and tell them to find that color based on memory.

That person in turn performs the same test to another random person.

If this test were performed millions of times, would we expect to see the color, say it started at a pale shade of pink, go all the way to jet black and back around before the game ended?


I'm asking because basic logic would tell me that it would theoretically cycle through the entire color wheel, but for some reason it's hard to believe.

[quarky]

interesting. I'd like to know too.

[Dymanic]

Yes, very intriguing question. The next thought I had was that conversely, if you flashed the same color to a large number of people at the same time, then (with the application of a little statistics) it seems reasonable to expect that the results would converge on the original color with a very high degree of accuracy. To make it extra fun, I'd want a really big color wheel with really fine graduations.

I seemed to recall that the human eye (or is it the brain?) can typically distinguish between something like a million different colors, and when I went off to Wikipedia to freshen up on that, I got refreshed on something else I probably also had heard but sort of forgot, which is that the human eye is more sensitive to some colors than to others. Based on that, I think I'd predict that once the color in your experiment had drifted into a certain range, it would tend to get stuck there. Not sure about that though. This stuff gets pretty hairy.

I wouldn't be surprised if someone has actually done an experiment very much like the one you describe.

[Denver]

Maybe someone with some programming inclinations could set this up on a web site...

[Foolmewunz]

Costly, as you'd have to figure out a way to exclude persons with color blindness (particularly monochromatic color blindness which would give you a bounce to black/white/gray), otherwise you'd just get these anomalous bounces - depending on which of the various types of color blindness the subjects have. Since more than 10% of the population has some form of it, you'd have anomalous bounces 100,000 times in every million subjects, so it wouldn't matter how finely graded your blues are if the person can't tell them from reds.

Unless you exclude the color blind, every tenth person would bounce the response into completely unexpected territory. That might be okay with millions and millions of sujbects, but I doubt it. The mathematical chance of going through, say, 50 subjects without hitting a color blind individual (and thus getting "true" responses within the realm of errors of perception) would be rather high.

[drkitten]

I've been thinking about the concept of the telephone game/evolution, etc a lot lately.

Consider this scenario:

You very briefly flash a color to someone from a color chart of some kind with thousands of colors, ranging the full spectrum. Then you give them the wheel and tell them to find that color based on memory.

That person in turn performs the same test to another random person.

If this test were performed millions of times, would we expect to see the color, say it started at a pale shade of pink, go all the way to jet black and back around before the game ended?

My bet is that it would not. Based not on color perception, but upon how human memory seems to work.

People seem to store not sensory information, but "prototypes" and "schemas." For example, if you show people a sentence and ask them to repeat it to you, they'll tend to make errors that preserve the core meaning of the sentence but have minor differences in wording (such as replacing one preposition or pronoun with another.) They'll even believe that they actually heard "novel" sentences if the sentences are similar enough in content. (http://www.alleydog.com/cognotes/schemas.html).

We know that human color perception tends to get filtered through these sorts of prototypes as well -- see the Berlin/Kay work on "basic color terms." So what I suspect would happen is that the color would drift to a "prototypical" example of one of the basic colors; if you showed someone taupe, it would drift to tan, and thence to light brown, and then to brown and stay more or less fixed.

[Denver]

Costly, as you'd have to figure out a way to exclude persons with color blindness..

I suppose they could make 2 tests, one for color blind, one for 'normal', and let the incoming user choose which test to take.

[progressquest]

I suppose they could make 2 tests, one for color blind, one for 'normal', and let the incoming user choose which test to take.

But, most of the color-blind people are happily unaware that they are color-blind. I would show the same color to a small group of people, then take the average of what they thought they saw (discarding obvious outliers). It will move more slowly, but it should still move.

[Denver]

But, most of the color-blind people are happily unaware that they are color-blind. I would show the same color to a small group of people, then take the average of what they thought they saw (discarding obvious outliers). It will move more slowly, but it should still move.

I've never heard that most people who are color blind being unaware of it (not saying its not true: just have never seen stats on that). I just looked up the prevalence of color blindness, and was surprised at how common it really is.

But, it might also be possible to program in some 'answer reasonableness' into the tests. Or, even present the user with an initial test to determine if their answers indicate color blindness, before incorporating their main test answers into the main test results.

[malaka]

I haven't seen anyone post this (http://www.xrite.com/custom_page.aspx?PageID=77) link from this (http://forums.randi.org/showthread.php?t=170597&highlight=color+score) thread yet.

Perhaps have the test subject take the test above before taking the "telephone-color" test? Only test subject that score X can participate in the follow-up test?

[jasonpatterson]

If the color gradations are sufficiently close, like in the test from malaka's site, I'd expect the result to be a random walk, the correct color would be the most frequently chosen, but a shade or two in either direction would also be highly likely. Eventually it would have to go everywhere.

[Third Eye Open]

I haven't seen anyone post this (http://www.xrite.com/custom_page.aspx?PageID=77) link from this (http://forums.randi.org/showthread.php?t=170597&highlight=color+score) thread yet.

Perhaps have the test subject take the test above before taking the "telephone-color" test? Only test subject that score X can participate in the follow-up test?

That was a fun test! I got zero, yay me! /brag

[Ririon]

When playing the telephone game, someone along the line will usually just make stuff up to be funny. The most common method is to whisper something rude about the person who started the line of whispers. Which the poor person on the end of the line will have to say out loud. So in this case, the colors would jump around to bright yellow or purple at random points. :p

[kevinquinnyo]

When playing the telephone game, someone along the line will usually just make stuff up to be funny. The most common method is to whisper something rude about the person who started the line of whispers. Which the poor person on the end of the line will have to say out loud. So in this case, the colors would jump around to bright yellow or purple at random points. :p

Yeah I should have mentioned, that let's assume that, somehow, no one intentionally defects.

I guess a real-life way to do this would be to have a referee who observes each test, and gives the players some sort of incentive, maybe cash? the closer they are to the shade of the last test...

[kevinquinnyo]

My bet is that it would not. Based not on color perception, but upon how human memory seems to work.

People seem to store not sensory information, but "prototypes" and "schemas." For example, if you show people a sentence and ask them to repeat it to you, they'll tend to make errors that preserve the core meaning of the sentence but have minor differences in wording (such as replacing one preposition or pronoun with another.) They'll even believe that they actually heard "novel" sentences if the sentences are similar enough in content. (http://www.alleydog.com/cognotes/schemas.html).

We know that human color perception tends to get filtered through these sorts of prototypes as well -- see the Berlin/Kay work on "basic color terms." So what I suspect would happen is that the color would drift to a "prototypical" example of one of the basic colors; if you showed someone taupe, it would drift to tan, and thence to light brown, and then to brown and stay more or less fixed.


That's interesting.

Do you really think that it would level off into a prototypical shade?

My intuition tells me that even if it got to tan or brown, if it was some sort of taupe to begin with, it would only take one person to remember the color a bit "cooler," (less warm) by a few degrees, which could shift the whole game into the cooler colors and maybe out of the browns, and into a blueish brown, and then into a purple and so on..

[kevinquinnyo]

If the color gradations are sufficiently close, like in the test from malaka's site, I'd expect the result to be a random walk, the correct color would be the most frequently chosen, but a shade or two in either direction would also be highly likely. Eventually it would have to go everywhere.

I wish I could see the results of this test, because I'd like to see if it "walks" up and down for a long period of time, or if it goes in one direction for long periods of time, or both?

You know what I mean?

[shadron]

I've never heard that most people who are color blind being unaware of it (not saying its not true: just have never seen stats on that). I just looked up the prevalence of color blindness, and was surprised at how common it really is.

But, it might also be possible to program in some 'answer reasonableness' into the tests. Or, even present the user with an initial test to determine if their answers indicate color blindness, before incorporating their main test answers into the main test results.

There are also different sorts of color blindness, and degree. It depends on which receptors don't work, and by how much. Then there's night blindness, where the rodws, rather than some of the cones, don't work properly.

[kevinquinnyo]

Someone, with the ability to code, should make a website for this, and we'll see what happens.

There are two ways to prevent intentional defection.

One is to incentivize degrees of correctness, with actual money.

A cheaper way would be to just discard blatantly incorrect answers. So we would have to determine how many shades off from the flashed color is considered a BS answer, and just throw that one out, and send the prior color to the next person.

[Foolmewunz]

Speaking of disqualifying wildly off responses....

If you want an easy method to "disqualify" responses from the various forms of colour blindness, you could have the first two rows as a "sample" that they can warm up on. But they could consist of the traditional pallettes used to test for the various forms of colour blindness. Anyone who bounces to colours that indicate that he/she has colour blindness, you just push to a holding pen and don't let it count.

If you wanted to be of some public service, you could post a message explaining that their answers indicate whichever type of colour blindness, and were they aware of that, please get tested, etc.... Provide links to whatever organizations deal with that sort of thing. (You might even get someone to fund the idea for you if you did that, by the way.)

[drkitten]

That's interesting.

Do you really think that it would level off into a prototypical shade?

I do.


My intuition tells me that even if it got to tan or brown, if it was some sort of taupe to begin with, it would only take one person to remember the color a bit "cooler," (less warm) by a few degrees, which could shift the whole game into the cooler colors and maybe out of the browns, and into a blueish brown, and then into a purple and so on..

"A few degrees" cooler won't shift a prototypical brown to a blue; that's a fairly substantial shift we're talking about. (Part of what makes prototypes prototypes is that they're guarded by a large "picket fence" in all directions....) Prototypes are at the "bottoms" of local attractors, if you will, so you'd need a fairly large change to get out of the local basin and into another one. Since most changes will run downhill, you'd need to see several improbable consistently upward steps in a row.

[kevinquinnyo]

I do.



"A few degrees" cooler won't shift a prototypical brown to a blue; that's a fairly substantial shift we're talking about. (Part of what makes prototypes prototypes is that they're guarded by a large "picket fence" in all directions....) Prototypes are at the "bottoms" of local attractors, if you will, so you'd need a fairly large change to get out of the local basin and into another one. Since most changes will run downhill, you'd need to see several improbable consistently upward steps in a row.

Now I'm more interested in the possibility that the color would get stuck, and not cycle through the entire spectrum.

But I still don't understand your reasoning. Any given color is on a spectrum is arbitrary, so when I say "a few degrees" that's also arbitrary. This means that there is a point on the spectrum where you are as close as you can be to dead center in between red and blue. (warm and cool) It could either go warmer or cooler by some degree of error, and so on and so on.

Explain why you think there would be some bias in people's errors, instead of random error.

[kevinquinnyo]

Would it change your opinion if I said that instead of there being thousands of distinct colors, it was literally a wheel with a gradual shift through all colors, and they had to pick the spot on the wheel?

I guess another way to say that would be, instead of thousands of colors, what if it were millions?

[drkitten]

Explain why you think there would be some bias in people's errors, instead of random error.

Because people see and think in prototypes, not in continuous variables. And memory errors tend to move towards prototypes instead of away from them. So what they remember will probably not be an exact shade (since people don't do "exact" worth a damn), but some sort of linguistically chunked concept like "mostly red."

And when people try to reproduce "mostly red," they'll be more likely to pick something red than something "mostly blue."

[drkitten]

Would it change your opinion if I said that instead of there being thousands of distinct colors, it was literally a wheel with a gradual shift through all colors, and they had to pick the spot on the wheel?

No.

[Ririon]

Because people see and think in prototypes, not in continuous variables. And memory errors tend to move towards prototypes instead of away from them. So what they remember will probably not be an exact shade (since people don't do "exact" worth a damn), but some sort of linguistically chunked concept like "mostly red."

And when people try to reproduce "mostly red," they'll be more likely to pick something red than something "mostly blue."
Then you would need people from the same culture/language-group. If you include people from culture/language-groups where the most important colors are different shades of brown/green/yellow that are difficult to explain in English, you would get different results, right? At least if your hypothesis is right. :)

[kevinquinnyo]

Because people see and think in prototypes, not in continuous variables. And memory errors tend to move towards prototypes instead of away from them. So what they remember will probably not be an exact shade (since people don't do "exact" worth a damn), but some sort of linguistically chunked concept like "mostly red."

And when people try to reproduce "mostly red," they'll be more likely to pick something red than something "mostly blue."

Okay, I see what you're saying. I really want to find out if you're right or not.

Will someone please create this experiment online? Doesn't anyone else want to see what happens?

It won't be perfect, but if we can effectively eliminate the color-blind, or intentionally defective answers, it should be obvious if Dr K is right or not.

I would try it, but I'm so bad at coding, I would fail. And I guess the tricky part is in making sure that the back end collects all the data correctly.

Also, it would need to not allow you to press the back button and see the color again (to cheat).

[kevinquinnyo]

Okay, Dr K, you might actually be right.

I was talking to a friend, and she actually made a good point, along the same lines as you. She thinks the lightness/darkness might be more important than the color.

She said it's possible that it would go up and down in lightness and hue from that color, but once it gets close to solid black or solid white, it would likely get stuck there.

I tend to agree with this.


I'm going to make this website, and try to find out once and for all.

[drkitten]

Then you would need people from the same culture/language-group. If you include people from culture/language-groups where the most important colors are different shades of brown/green/yellow that are difficult to explain in English, you would get different results, right? At least if your hypothesis is right. :)

No, that was one of the findings of the Berlin/Kay experiments.

Prototypical colors seem to be universal. What is not universal are which colors are described by "basic color terms" and which ones are described by metaphors.

For example, "orange" is technically a metaphor (it's the color of -- you guessed it -- an orange). Similarly, avocado, grapefruit, and olive are all metaphors even if you can find them in the paint aisle.

Japanese doesn't have a basic term for "green" (it's techncially part of "blue"), but they are still aware of the distinction, and they use the metaphorical term "midori" (honeydew melon) to describe it. Similarly, I've heard phrases like "red like a banana" to describe what we in English would call "yellow" in languages with small inventories of color terms.

So, no, I don't think that the culture would have that big an effect. Even in Japan, people know the difference between "aoi" and "midori."

[kmortis]

That was a fun test! I got zero, yay me! /brag
Yeah, but you got three eyes to use...

I got a 79...crap onna stick. I think I'm going blind.

[kmortis]

No, that was one of the findings of the Berlin/Kay experiments.

Prototypical colors seem to be universal. What is not universal are which colors are described by "basic color terms" and which ones are described by metaphors.

For example, "orange" is technically a metaphor (it's the color of -- you guessed it -- an orange). Similarly, avocado, grapefruit, and olive are all metaphors even if you can find them in the paint aisle.

Japanese doesn't have a basic term for "green" (it's techncially part of "blue"), but they are still aware of the distinction, and they use the metaphorical term "midori" (honeydew melon) to describe it. Similarly, I've heard phrases like "red like a banana" to describe what we in English would call "yellow" in languages with small inventories of color terms.

So, no, I don't think that the culture would have that big an effect. Even in Japan, people know the difference between "aoi" and "midori."
Add to that, the OP conflated how the brain processes verbal and visual information. In one of his earlier posts, dr. kitten mentioned that humans remember the jist of a verbal input (my summary here just goes to prove this point as I can't be bothered to go and look at his actual post :)). Due to that phenomenon, we're more likely to add noise into the signal path, thus creating the game we all know an love.

The same is not true of colors. While we all might see varying shades of color, if there was a color wheel to compare against, no one is likely to choose a yellow when a blue was shown (ok, we can discuss the "boundaries" between color later).

[Ririon]

No, that was one of the findings of the Berlin/Kay experiments.

Prototypical colors seem to be universal. What is not universal are which colors are described by "basic color terms" and which ones are described by metaphors. (...)
OK, so what are the prototypical colors? Red/Green/Blue? Red/Yellow/Blue? Red/Orange/Yellow/Green/Blue/Violet? Indigo? Cyan/Magenta/Yellow/Black? White? Hot pink? Beige? And why?

[drkitten]

OK, so what are the prototypical colors? Red/Green/Blue? Red/Yellow/Blue? Red/Orange/Yellow/Green/Blue/Violet? Indigo? Cyan/Magenta/Yellow/Black? White? Hot pink? Beige? And why?

From Wikipedia :


Berlin and Kay also found that, in languages with less than the maximum eleven color categories, the colors found in these languages followed a specific evolutionary pattern. This pattern is as follows:

1. All languages contain terms for black and white.
2. If a language contains three terms, then it contains a term for red.
3. If a language contains four terms, then it contains a term for either green or yellow (but not both).
4. If a language contains five terms, then it contains terms for both green and yellow.
5. If a language contains six terms, then it contains a term for blue.
6. If a language contains seven terms, then it contains a term for brown.
7. If a language contains eight or more terms, then it contains a term for purple, pink, orange, grey, or some combination of these.

In addition to following this evolutionary pattern absolutely, each of the languages studied also selected virtually identical focal hues for each color category present. For example, the term for "red" in each of the languages corresponded to roughly the same shade in the Munsell color system. Consequently, they posited that the cognition, or perception, of each color category is also universal.

The last paragraph is the key one in this context. Even in a language where black and brown are named as the same color, the "focal hue" will still be black, not brown or brownish-black. A language that merges red, orange, and pink will still pick what English speakers call "red" as the focal color.