Join 3,430 readers in helping fund MetaFilter (Hide)

Tags:

How many greens are there?
February 12, 2007 8:10 AM   Subscribe

How many shades of the color green can the human eye differentiate?
posted by tamagosan to Science & Nature (24 answers total)
 
I would rather imagine it depends on the human in question.
posted by DoctorFedora at 8:21 AM on February 12, 2007


Not to mention what you define as "green." If you have a sort of blue-green, it will be considered by some to be a shade of blue, and by others to be a shade of green.
posted by DevilsAdvocate at 8:24 AM on February 12, 2007


In theory, there are an infinite number of shades of any color. A shade is just a particular combination of photoreceptors being activated. As far as we know, the perception of shade is more of a real-valued phenomenon than a discrete one.

However, differentiability is a different story altogether, and like many other sensory processes, there is a threshold below which people will judge shades that are only slightly different to be the same.
posted by ripple at 8:25 AM on February 12, 2007


From an Ask Yahoo column about the number of different colors the eye can perceive, I got to this article. It seems to be somewhere between tens of thousands and several million.
posted by Plutor at 8:33 AM on February 12, 2007


In theory, there are an infinite number of shades of any color. A shade is just a particular combination of photoreceptors being activated.

So we have an infinite number of photoreceptors?
posted by mdn at 8:36 AM on February 12, 2007


I think ripple's comment about differentiability cuts to the heart of this. A good practical answer for this would probably involve a defintion of the bounds of green-ness, and then AB tests of a lot of pairs of differing shades in those bounds, against enough subjects to get a good normative baseline. Figure out the average threshold (or thresholds—it may vary across the spectrum of greens) and approximate from there.

I imagine this sort of thing has been done a lot, in one way or another, in optical research, but whether it's been done in a way that will let you extract meaningful numbers for "greens" from publicly available results is another question.
posted by cortex at 9:27 AM on February 12, 2007


Also, armchair data: I recall as a kid reading (perhaps from the Charlie Brown Illustrated Encyclopedia—salt to taste) that the figure for human-distinguishable colors was about 7 million. In a BBC Science Journalism sense, you could say that there's about a million or so discernable greens, then.
posted by cortex at 9:30 AM on February 12, 2007


Also, different languages/cultures define the "frets" in the color spectrum differently.
posted by iamkimiam at 9:36 AM on February 12, 2007


So we have an infinite number of photoreceptors?
That's irrelevant; what's relevant is how many levels of excitation one can actually exhibit. If you think about the output as a continuous variable then there are infinitely many different levels possible. In practice, the output is discretized by the time we're making a perception, so there is a finite number - but it will vary from person to person. I would tend to agree that "a few million" greens is plausible for someone with good color acuity. You can experiment with this yourself by tinkering with color codes in 3-byte RGB form and shifting a component by just 1 unit. Can you perceive the difference?

A really curious thing is that our perception of a color is a combined profile from 3 different receptors, each of which is receptive to a different slice of the spectrum (together with a "brightness" sense from the rods). That should give rise to a 3-dimensional "cube" of colors - but then, not every profile can actually be caused by natural light, so much of the cube consists of "impossible" colors. You can, however, with some trickery cause those states to occur artifically - giving rise, for example, to "unnatural" greens that you would never experience normally. So add a few extra to your count if you like.
posted by Wolfdog at 9:47 AM on February 12, 2007


According to the excellent Ask A Color Scientist, we can discriminate a difference in color/wavelength of about 2nm at the middle of the color spectrum. However their answer for the total number of colors we can discriminate is a bit vague: "Millions".
posted by teg at 9:52 AM on February 12, 2007


I think the context you see them in could also be important; I might remember two block of green that I see at different times or in different places as the same color, but if you placed them next to each other then I might be able to discern the difference. Of course, that probably has more to do with color theory and I don't know if that is what you are looking for.
posted by rmless at 9:58 AM on February 12, 2007


My guess would be about a million, max. You could set up an experiment with three color tabs, two of which would be identical. The point at which the subject couldn't consistently identify those two would give an indication of the "fineness" of discrimination, and you could extrapolate from there. Bear in mind that color is in the light, not in the paint, so it should be done with full spectrum illumination, or on a monitor.
posted by weapons-grade pandemonium at 9:59 AM on February 12, 2007


Also, and I'm not a physicist, but since light is both a particle and a wave, and visible light is a specific wavelength, wouldn't that imply that there is indeed a finite number of colors since at some point, you will reach the smallest wavelength physically possible?

I thought of this while looking up at the incredibly rare (for this time of year) blue sky last week here in Seattle. I could see the subtle gradients from light to dark blue and thought of how many colors that must be, then realized that although the number might as well be infinite for me, it would surely have to be finite in the big picture.

anyone know if what I'm spouting makes sense?
posted by johnstein at 10:01 AM on February 12, 2007


*reads other posts and realizes he's approaching the problem from the wrong end*

Since our eyes have far less resolution than required to even see all the colors, the fact that there may or may not be a finite number of colors is implicitly irrelevant, eh?

interesting question, though.
posted by johnstein at 10:05 AM on February 12, 2007


Johnstein - What you're saying isn't quite the right use of the particle/wave duality. When dealing with photons as particles, you talk about their energy (which is inversely proportional to wavelength) and photons can take any nonzero energy. The particle-like behavior manifests itself when you start counting them, for instance looking at the photoelectric effect.
posted by Schismatic at 10:16 AM on February 12, 2007


yea, I knew I wasn't using it correctly (thanks for the link), but is there any discretization of the wavelength spectrum? or is it 100% continuous (truly infinite colors)?
posted by johnstein at 10:18 AM on February 12, 2007


I like this question.

I have no idea where to look this up, but a smart person once told me that humans are able to distinguish a greater number of shades of green than any other colour. The reason he gave had something to do with human evolution-- perhaps because early humans spent alot of time around those greeny, planty things.

If this is true, then you could *not* simply divide the total number of perceivable shades by the number of colours.

Sorry for the complete lack of data and citations.

Also, I would like to say that I agree with johnstein's comment about the perceivable number of colours versus the actual number of colours.
posted by blisterpack at 10:57 AM on February 12, 2007


Also, DoctorFedora has a point, especially when you consider than women can discriminate between more shades than men (in general).
posted by timepiece at 11:10 AM on February 12, 2007


As others have pointed out, this all depends on your definition of "green". If we (conservatively) define green as between 500 and 560 nm (the region centered around the 533 nm cone photoreceptor), and use the result described in teg's link of 1-2 nm of resolution, we can see about 30-60 shades of green. This accounts, however, only for differences in hue. We also need to account for differences in saturation and brightness, which we are also capable of perceiving. This is how we get from the less than 200 hues (pure wavelengths) we can see to the millions of "colors" we can perceive.

And at this point, it starts to get complicated, especially since you might no longer perceive some of the saturation values as "green" anymore. But let's try and do a back-of-the-envelope calculation. I can't find any good studies on the perception of saturation and brightness, but photoshop allows for 100 levels of each. If this is a good approximation to our perception, we would expect in the range of hundreds of thousands of greens to be distinguishable.

johnstein writes "yea, I knew I wasn't using it correctly (thanks for the link), but is there any discretization of the wavelength spectrum? or is it 100% continuous (truly infinite colors)?"

I'm not a physicist, but I think this is actually a very deep question that may or may not have a consensus answer. In practical terms, the wavelength of a photon can take on any value: the spectrum is continuous. However, at a very fine scale, there might be quantitization. This would probably be related to the quantitization of spacetime at the Planck scale, which is pretty damn complicated stuff. But it seems to me that if spacetime is quantitized, then the wavelengths available to photons would also be quantitized. This is at a scale much finer than perceivable by any available instruments, though.
posted by mr_roboto at 11:25 AM on February 12, 2007 [1 favorite]


A shade is just a particular combination of photoreceptors being activated.

This isn't accurate at all. Perception and perceptive "processing" takes place in all parts of the optic system, from the retina back to the occipital cortex. Further, in order to analyze and discriminate shades, and then talk about what you did, you need to use most of the brain's association cortex.
posted by ikkyu2 at 4:12 PM on February 12, 2007


Depends on how green a colour needs to be to be called green... consider that we have 3 bands of colour photo-receptors, one of which is for greenish colours. For a monochromatic light, we may well be able to differentiate between two colours 2nm apart but we do it not by measuring the wavelength but by noticing how the relative magnitude responses of the 3 receptors change. We then synthesize "colour" from those three magnitudes.

For this reason, certain mixes of spectra appear identical to us as other monochromatic light. That's how we get yellow from computer monitors: add red and green, cause the same response in the eye as occurs when illuminated by monochromatic yellow.

So are we talking truely green monochromatic light in the green part of the spectrum? Or arbitrary mixtures of spectrum that cause a response in human perceived to contain green; e.g. a pale (unsaturated) lime-milkshake colour contains green but also lots of other bits of spectrum.

Assuming we take the "any mix is valid" approach, we should think about the total dynamic range of the human eye (about 90dB... but that number applies to intensity - different receptor type - not colour). If I tag a wild-arsed-guess and say 80dB of total dynamic range for our green sensor and 70dB for the other two, we'll maybe be in the right ballpark. That's too hard, so I'm gonna say 80dB for all three since all these numbers are pulled from my fundament.

Now for differentiability... the usual consensus is that we can perceive a 1 in 128 difference in intensity; a ratio of 1.0078125. Wrong receptors in the eye again so lets fudge and say 1 in 64 for green, 1 in 32 for the other two.

Now the total size of the perceivable (NOT at one instant but over many hours as one's eyes adjusts) colour space is defined by the product of the total number of discrete intensity steps on each receptor; call it n.

ng65/64 = 108
log(ng) = 8*64/65 = 7.87
ng = 74.1e6
nother= 57.2e6

Total colour space = ng * nother2 = 4.2e15

Which is a lotta colours, but that's not surprising given that we're talking all the way from starlight to blinding brightness in all three colour channels. Now, what fraction of that space is green? How green does a green mix have to be to be called green? Lets set the definition at "more activity on the green receptor than the other two combined".

The colour space I defined is a cube but with log basis, which makes it more difficult to subdivide the space by colour. I'm going to take a really dodgy subdivision, cut one corner off and call it green, say 1/8. That leaves something on the order of 1014 shades (varying hue, intensity and saturation) of green.

Adjust assumption numbers (dynamic range, minimal level differentiability) according to a better knowledge of physiology than I have. Note that this is the total perceivable colour space, it cannot be observed at any one time. At one instant, we're limited by the 1-in-128 rule of thumb to approximately a couple of million colours.
posted by polyglot at 4:19 PM on February 12, 2007


Polyglot, your analysis is quite internally consistent, but it is not pertinent because the human vision system is not simply a digital photoreceptor that yields a discrete numeric value for color. Very basic perceptual research, for instance, reveals that the same color can be perceived very differently depending on the color context surrounding it.

People vary considerably in their ability to discriminate color, with the most skilled probably able to discriminate several million shades under ideal conditions.
posted by ikkyu2 at 4:31 PM on February 12, 2007


Oh I agree that things look quite different in context; there are plenty of demonstrations of this. I don't think that changes how one would analyse the differentiable shades.

My derivation is based on the concept that if you put 2 shades next to each other, they look different yet both green, they count towards the total.
posted by polyglot at 4:55 PM on February 12, 2007


Perhaps I should refine it and say "each looks green when placed next to white". That comes into the definition of how green is green and is completely fudged by my "1/8 of the total space" hack at the end, which is really quite invalid. Obviously it will differ from person-to-person and there's plenty of wiggle-room near the secondaries (eg cyan).
posted by polyglot at 5:28 PM on February 12, 2007


« Older WannabeEditorFilter: I hated m...   |  Tell me about being an Air Cou... Newer »
This thread is closed to new comments.