h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

Right. Gonna read it now.

You sound like quite a clever and thoughtful guy and I am a clever and frivolous woman (not that that should matter). I am also a physics teacher and don't like to sound stupid but I am always open to learning so maybe you can teach me.

Hmmm. a few things...

I'm not sure what you mean by 'mental primary colours'. Light: Red, Green and Blue. Paper: Magenta, Yellow and Cyan (not sure mysalf about this one and if I could be arsed to lok it up, no doubt I'd find out).

I go with a few points:
1. that we just call it violet but even so it is just a darker blue but that's the name we give it.
2. it's something to do with the fact that there's blood in us and that mixes with the blue to give us violet.
3. not sure about the fact that blue has more green in it and less has a redder effect.

Thinking.....
trying to think laterally now....

What about the blue fading into the ultraviolet (ignore the name 'ultraviolet'). Writing and thinking at the same time now.....fading into ultraviolet must mean fewer and fewer long wavelengths...hmm ... are we not confusing a purply colour with the fact that we mix red paint with blue paint and get purple... no.. becouse we get magenta when we mix light....

Why have I never thought about this before. I'm asked these kind of questions in class but not this one yet.


I shall get a prism out tomorrow and have another look. Also I will ask my colleagues.

I'm tempted to go with no. 2 at the mo.

Bl00dy hell. I have never written such a long post. It seems to be catching.

What I meant by 'mental primaries' is a bit complicated.

For simplicity's sake, let's pretend for the moment that all optical input to the eye comes in the form of variable intensity monochromatic light of any frequency, such that each small area of the retina isn't dealing with mixtures of multple wavelengths.

Assume a section of the visual brain is processing information from the colour signals coming from the eyes, correcting for ambient light colours, etc, but the information is still in the form of signals rather like the RGB values used to form pixels, without at that stage having any actual feeling of redness, etc.

Now, what I mean by a 'mental primary' is the (possibly personal) sensation of redness, greenness, etc. that accomanies each point in a scene, and enables us to consider one object as being redder than another object, or one particular colour as having more blue than green, or as being lacking in red.
The colours I consisder as fundamentally primary (RGBY) are primary in the sense that
a) they are the only ones needed to mentally label any point in the percieved visual spectrum,
b) it wouldn't be possible to label all spectral colours without all four, and
c) they seem to divide the spectrum into distinct regions. For instance, moving from the red end, the mental red decreases, and mental yellow increases until pure yellow is reached. Beyond yellow, it's all yellow-greens, with no red at all. Beyond green are green-blues, which contain no yellow, and beyond blue are the violets, which (to me) contain an increasing mental red component, but no green.


The mental primaries appear to be internally generated properties to allow us to easily process colours, and as such they aren't bound to map in any simple way to the external RGB information.
Looking at
http://hyperphysics.phy-astr.gsu.edu/hbase/vision/colcon.html#c1
it can be seen that although the RGB response curves are unevenly spaced in one way, the mental red/yellow/green/blue points are, if anything, spaced in the reverse manner, with the red->yellow spread being wider than yellow-green, and yellow->green wider than green->blue.

If the internal ideas of red, blue, etc aren't forced to tie in with how the eye actually receives colour, it does allow the brain to position them where they do the most good in terms of allowing discrimination between commonly-experienced colours. Having 'redness' active at both ends of the spectrum is actually very easy to achieve in terms of mental processing, and does allow for the use of one primary to help give distinction to the colour experience in two areas of the spectrum.

There is the issue of 'why *four* primaries?', but I think the evolutionary approach of successive colour systems (blue/yellow, then green/red) is at least a plausible one to explain that.

*Slight aside* - Looking at the diagram, I can't help wondering if it would be better to think of the optical red receptor as actually being largely a mental yellow one, with mental red being more a function of relative lack of green at either end of the spectrum than of actually being a positive colour like the other three.
However, I suppose having anti-green, yellow, green, and blue as colours would be even more confusing?

**Much* further aside*
I wonder if it is possible for someone to have a perfect retina, but a vision-processing system that actually *does* work with only three mental primaries?
Though it seems hard to square with the 2-pair evolutionary model, if it was possible to lose half of one system, than technically, I suppose someone lacking 'mental red' could still have full sensing capacity across most or all of the spectrum , especially if yellow's set point shifted towards the red end, and blue drifted up to the violet end. I suppose sensitivity might suffer, but possibly not enough to be easily detectable in terms of colour matching.
However, I suppose they couldn't have yellow wrap-around to the violet end of the spectrum the way red does, as it would then clash with its opposite colour (blue), so though they *may* have full discrimination across the spectrum, they may be less sensitive to subtle changes (having wider-spaced primaries), not have a primary common to both spectral ends, and they would also only see three primaries when looking at a full spectrum. It would be really interesting to find someone like that.

1. that we just call it violet but even so it is just a darker blue but that's the name we give it.
2. it's something to do with the fact that there's blood in us and that mixes with the blue to give us violet.
3. not sure about the fact that blue has more green in it and less has a redder effect.


1) To me, there is definitely the sensation of red in pure violet light, not just in magenta/purple mixed light or inks. If people didn't generally see violet light as having a red component, I assume it wouldn't get printed in books or modelled in RGB values as being purple.
2) Blood (or other retinal tissue in the light path), unless actively red-fluorescent in violet light would presumably have a filtering effect across a range of shorter wavelengths. It seems unlikely to actually *increase* the red signals being recieved at wavelength where the red receptor is largely blind.
3) It depends which blue you are referring to. One understandable misconception is that 'blue' light is the same thing as computer-screen blue. By my reckoning, computer-blue is actually the blue-green mix that causes *mental* blue to be experienced, and is far from the central frequency for blue receptors in the eye.
While in computer RGB, a perfect blue may be (0,0,255), it may still end up with a phosphor that produces light that stimulates blue and green receptors pretty equally.



>> "Why have I never thought about this before. I'm asked these kind of questions in class but not this one yet."

I do tend towards odd thoughts at times, and *I'd* not really thought enough about it before the rainbow conversation that started me off, *despite* having in my mind virtually all the pieces I needed to work out some plausible, if hard-to-explain explanation.
I had wondered about the red-in-violet issue on various occasions, but not enough to really get into the question. It was seeing the diagram linked to in the previous post that helped put all the pieces together, since until then I was still somehow hung up on some kind of RGB model for *conceptual* colours mapping fairly directly onto actual receptors, even though in other ways I was pretty convinced there were four basic colours that I *analyzed* the world in terms of, and the evolutionary reasons for that model.
I suppose I'd always just assumed that mental yellow was a basic red/green receptor average or suchlike, and left it at that.
I think I still thought that way when I started off the conversation - it's a very easy representation to picture, but a real progress barrier to correct thinking.

Still working on it.

<>

Explain please - having redness at both ends is easy to achieve. Why should one need to allow for one primary colour giving distinction at both ends of hte spectrum?

If you assumed that vision was largely like it is, but the sensation of red was confined to long-wavelengths, then there wouldn't be any sensation of the colour of light being different on any shorter wavelength than that which causes the sensation of blue.

That would either mean that there was a region of the spectrum beyond the mental 'blue-point' which would appear a uniform blue, or that the blue-point would have to shift to shorter wavelengths (corresponding to a much larger blue:green ratio in the input from the blue and green receptors).

If the mental blue-point did shift to shorter wavelengths, such that the blue and green points were further apart on the spectrum, then the sensation of blue-greens would cover a wider span, which could reduce the ability to discriminate between similar colours.

For example, let's pretend that the blue point is at 300nm, and the green point at 500nm, and assume that the blue and green sensations in the brain vary linearly with intermediate wavelengths, so that light at 350nm was experienced as being 3 times as blue as it was green.
If we could distinguish between two colours where the mental blue or green levels were 2% different in absolute terms, such that we could tell light whcih caused a 60/40 blue/green sensation from light which caused a 58/42 one, then our discrimination ability in wavelength terms would be 4nm.
However, using precisely the same neural circuitry to compare mental blue and mental green, if we alter the earlier signal processing circuits that actually generate the mental blue and green signals such that blue corresponds to light of 400nm wavelength, (such that light at 425nm was perceived as being 3 times as blue as it was green) then in the area between mental blue and green, we would have a much better discrimination ability (2nm) in terms of wavelength.

In practice, the quality of the input from the eye will be a limiting factor at some point - it wouldn't be possible to generate a stable mental blue and mental green that corresponded to very nearby wavelengths, since that would require each to be a function of receptor input that was very sensitive to small fluctuations.

However, it may well be the case that if the blue point was closer to the extreme violet end of the spectrum, then discrimination might suffer in areas between two colour points which had moved further apart.

Of course, it's possible that in nature, actual violet light is so rare that there wasn't any selective pressure to develop a system to analyze light in that area of the spectrum.

In that case, then it might well be predicted that the blue point would lie well below the peak sensitivity of the blue receptor (which it does), and that it largely marks the point above which there isn't much useful information for us to need to be able to make distinctions about.
The 'leaking' of the experience of redness into the very short wavelengths may just be an artifact of the way the 'redness' signal is generated.

If redness is calculated from some constant, with the addition of a term derived positively from red input, and negatively from green input, then it will be strong where red input is greater than green (or where there is red input but no green), weak where green is much stronger than red, but could also have a weak value where red input is absent, and green input is weak or absent, if the constant isn't completely cancelled out by the influence of green.

If redness was also completely inhibited not just by green receptor input, but by the presence of the greenness *property*, then it would be zero from yellow through blue, but could emerge again just beyond blue to help produce the violets.

Sorry to interrupt, but a skim read of this thread suggests you're leaving out from your thoughts a fairly critical element you need to consider - the physical/chemical process involved in the actual sensation of light. By that I mean, what exactly physically happens when a human retina is impinged by red light, yellow light, green light, blue light, etc.? It's quite a distinct and different process compared to what happens when the same light impinges on, say, a CCD in a video camera, or on a piece of cine film.

Light is ultimately fairly simple. It only gets complicated once it hits the back of the eye. Thus it would seem to me that if you're trying to make sense of why it's complicated, your starting point has to be the process it goes through at that point - what precisely happens to a cone cell when a light wave hits it?

H.

I'm not sure the retinal processes make a big difference to what I was thinking about, as long as I take into account that each different receptor has a fairly broad sensitivity curve, and that these curves do overlap, rather than each receptor responding to each own unique 'colour' of light.

I think it's reasonable to treat the retina as having three different kinds of sensor, with each sensor producing an output signal correlated with the total amount of light it is sensitive to, moderated by its particular sensitivities.

While there are certainly complexities in retinal light reception, like pigment bleaching giving an automatic-level-control effect, From the perspective of the output signals, I think it's a valid simplification to leave them out for the issue I was wondering about.

Similarly, there are various kinds of image processing which actually happen in the retina, rather than the brain proper, but counting those to be mental processes can help simplify the model.


Fundamentally, what I'm driving at is something like a two-step process which processes turns the three receptor signals from each area on the retina into intermediate values (which I'd think of as redness, yellowness, greenness and blueness).
These intermediate signals are then processed to give the full colour experience, allowing the experienceing of cyan, puce, orange, etc. However, I'd suggest that we also have some kind of conscious access to the intermediate signals, so that our tendency when comparing or describing colours is in terms of the intermediate values.

I'd also suggest that we don't have real conscious access to the raw (3-sensor) input data.

Possibly that's because much other processing happens in the first step, to compensate for the variable nature of lighting and generate some intermediates values that are much more stable than the raw data.

Possibly that's because the intermediate signals are more useful, since each of the four intermediates has a narrower range of activation than a raw receptor signal, and (at least, regarding monochromatic input) only overlaps with its two neighbours in a simple way, dropping from a peak at its central point to near zero at the central point of its neighbours.

(For light *mixtures*, things are more complicated, since all the intermediates might may be positive simultaneously. However, even then, it might be handy to have some simple signals that could only be all active when light has a white component. Possibly the whiteness (desaturation) of a colour could be calculated by looking at the strength of the smallest intermediate signal, or something like that?)