Closed sarasuri closed 6 months ago
detlevhfischer
commented
2 years ago @sarasuri I think, yes, as in-page conforming alternative version. The alternative text of the graphic might then say something like "Bar chart of Number of Enquiries and Thresholds, 2008-2011 (see table below)". The table would need to be correctly marked up to qualify as accessible alternative.
sarasuri
commented
2 years ago @detlevhfischer Thank you for feedback. Some people have a different opinion with this approach that sighted user who are colour blind shouldn't have to look at the table to perceive the information from the graph. One of the solutions would be to include patterns in the graph to convey the same information in a way it does not depend on colour to meet this the Success Criterion 1.4.1: Use of Color. Any thoughts? Thanks
GreggVan
commented
2 years ago If someone has vision - then there is information in a chart that is not in a table — that can be made available to them. So the chart should not convey information by color alone. And having a table is best we can do if you don’t have vision (until we have dynamic tactile) but it is not the same or equivalent to a chart — or we would just use tables instead of charts.
So the chart should not rely on color alone — and it also should not consider the table to be its equivalent. It is an alternative but it is not equivalent. It does not convey relationships or patterns or trends in the same way as the graphic can.
Best
gregg
On Sep 28, 2022, at 11:35 AM, sarasuri @.***> wrote:
@detlevhfischer https://github.com/detlevhfischer Thank you for feedback. Some people have a different opinion with this approach that sighted user who are colour blind shouldn't have to look at the table to perceive the information from the graph. One of the solutions would be to include patterns in the graph to convey the same information in a way it does not depend on colour to meet this the Success Criterion 1.4.1: Use of Color. What are you thoughts?
— Reply to this email directly, view it on GitHub https://github.com/w3c/wcag/issues/2690#issuecomment-1261316487, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACNGDXVCS2WJAWEO4C4P5LDWASFVZANCNFSM6AAAAAAQWD5GJI. You are receiving this because you are subscribed to this thread.
detlevhfischer
commented
2 years ago @sarasuri I agree that it would be better to use something like pattern (or a label on top of the bars) in addition to color but that won't always be possible. While I agree with @GreggVan that the chart should not convey information by color alone, I would emphazise the should in that sentence. I would still hold that for that type of data, a well-formed data table below can be considered a minimum equivalent conforming alternative if all data points can be unambiguously accessed visually and programmatically. Of course it is easier to see that turnover has doubled from one year to the next in a bar chart, but the same information is available if I can read turnover 2009: 4 million / turnover 2010 8 million.
So, in normative terms, the requirement of 1.4.1, "Color is not used as the only visual means of conveying information", is satisfied by the table.
An additional note regarding data visualisation: It really depends on quantity and context. Once you go beyond a few data points, differentiating by color, even when adding pattern, will get increasingly difficult, and at some point pattern may just add confusion. It all depends...
GreggVan
commented
2 years ago I will disagree on a table being a sufficient equivalent for a chart for anyone who is not blind. And even for people who are blind a description of patterns or trends can be important. (Except for very simple chart and very simple tables with not very many entries)
There is a reason that the charts are included or preferred for presenting many types of information. There are very few people (with our without disabilities) that can look at a table and see what is obvious from a chart.
My emails are not standards so I don’t use must — but I don’t see any reason to no create charts and graphs that do not rely on color alone. We did it for decades (centuries?) before we had color. Color is fantastic to add but not rely on. By the way remember that by color I mean HUE and LIGHTNESS.
So red and pink are two different COLORS when I say "not conveyed by color alone" you do not have to use patterns if you have just a couple colors and they are different darkness/lightness.
Best
gregg
On Sep 29, 2022, at 6:13 AM, Detlev Fischer @.***> wrote:
@sarasuri https://github.com/sarasuri I agree that it would be better to use somethng like pattern (or a label on top of the bars) in addition to color but that won't always be possible. While I agree with @GreggVan https://github.com/GreggVan that the chart should not convey information by color alone, I would emphazise the should in that sentence. I would still hold that for that type of data, a well-formed data table below can be considered a minimum equivalent alternative if all data points can be unambiguously accessed visually and programmatically. Of course it is easier to see that turnover has doubled from one year to the next in a bar, but the same information is available if I can read turnover 2009: 4 million / turnover 2010 8 million. Also, I think regarding data visualisation, it really depends on quantity. Once you go beyond a few data points, differentiating by color, even when adding pattern, will get increasingly difficult, and at some point pattern may just add confusion. It all depends...
— Reply to this email directly, view it on GitHub https://github.com/w3c/wcag/issues/2690#issuecomment-1262261089, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACNGDXVGQJK2W7Q6SYTL6YDWAWIWFANCNFSM6AAAAAAQWD5GJI. You are receiving this because you were mentioned.
Myndex
commented
2 years ago Hi Gregg @GreggVan (and also Hi @sarasuri @detlevhfischer and others:)
Gregg said:
I will disagree on a table being a sufficient equivalent for a chart for anyone who is not blind. And even for people who are blind a description of patterns or trends can be important. There is a reason that the charts are included or preferred for presenting many types of information. There are very few people (with or without disabilities) that can look at a table and see what is obvious from a chart.
I agree very much: graphical presentation tells the story much better than an abstract table of values. Using a table of values as a way to mitigate color perception issues with a chart is not a salient choice.
By the way remember that by color I mean HUE and LIGHTNESS. So red and pink are two different COLORS when I say "not conveyed by color alone" you do not have to use patterns if you have just a couple colors and they are different darkness/lightness.
I am a little confused by this statement, but it may be merely a confusion of terminology.
by color I mean HUE and LIGHTNESS
I would say that for clarity, color only means hue/chroma, and not luminance (which is perceived as lightness).
Lightness/darkness/brightness are context sensitive perceptions of light volume, or luminance, separate from color as in hue/chroma/saturation. The distinction is not merely one of semantics—neurologically, achromatic luminance is processed separately from hue/chroma.
Lightness and hue are very literally separated early in the human vision system, first in the opponent process of the post-photoreceptor ganglion cells, and then in the visual cortex. V1 processing the luminance channel, and V4/V8 processing hue/chroma downstream (and delayed by about 70ms).
Of principal importance is that luminance carries all of the high spatial frequency signals (edges, patterns, fine details) while hue/chroma are sensitive to very low spatial frequencies (large areas, no details).
As a result of the separated processing and paths, the importance of "not carried by color alone" refers not to luminance, only to hue and chroma, for the reasons of certain impairments of color insensitivity.
In reference to an sRGB display, the web default standard, and the appropriate display standard for accessibility purposes, luminance is not a factor with the exception of the sRGB red primary for the protan (and the very rare monochromat/achromatopsia). In the case of protanopia, the sRGB red primary has about half the luminance vs standard vision. Otherwise, deutan, protan, tritan have standard luminance perception, including standard or better luminance contrast sensitivity and standard or better visual function overall.
For tritan, missing the S cone is not a factor for luminance as the S cone does not contribute additively to the luminance channel. As far as luminance discrimination is concerned, with the exception of colors relying on the red primary due to the protan deficit, luminance can and should be considered separately from hue/chroma.
Unfortunately, luminance perception, meaning achromatic changes in lightness/darkness/brightness, is a generally poor way too encode information. The FAA has some standards that restrict information coding using grey levels to three: black, grey, white. The reason is that it is difficult distinguish a particular grey level without substantial context.
With standard vision there are many distinguishable hues, so hue is a more ideal method of information coding. Except of course for color insensitive vision. But even color insensitive vision distinguishes some colors. And here, with some reservations, hue/chroma and luminance together provide for coding information with "not just color".
And again, in the statement "not just color" color means only hue/chroma irrespective of luminance.
But there is another contrast at play in the chart example above, a simple bar chart, with a pair of related bars repeated three times for three different dates.
The unmentioned contrast is the positional contrast. Each pair of bars features a bar on the left that means "Enquiries" and a bar on the right that means "Threshold". The positional contrast of these two bars in each pair is also an information coding method.
In the above example, the legend has the following issues:
1) It is top to bottom, while the bar pairs are left to right.
2) The colors for the legend are very small squares.
3) The two chart colors noted above have substantial luminance differences, as we can see in the CVD processed samples below.
In the set of four below, top left is standard vision, followed clockwise by deuteranopia, protanopia, and tritanopia. Notice in this simple chart that all three of the most common forms of CVD have adequate discernibility here.
So, if we process that chart to reduce to plain luminance, as if it were displayed on a black-and-white display (hey, we had those when I was a kid, LOL) it would look like:
Many of the color vision deficiency simulation tools out on the Internet today use this simple as RGB luminance claiming that it represents achromatopsia or the monochromatic, and this is completely incorrect.
Someone with Rod only achromatopsia, or a blue cone mono chrome app, will have no perception of red and a reduced perception of green.
But more important: the loss of the L and M cones means that these very rare impairments are also affected by very low vision and extreme photophobia. An experimental simulation of BCM is:
Not only are no colors seen, but it is blurry and also "blown out" due to the photophobia effect exceeding the range of the rods.
Side note: rod only achromatopsia and BCM are the only "actually color blind" and are very rare in the group of color vision impairments. The others, Deu/Pro/Tri are color insensitive, but not color blind.
These were processed by The Myndex Color Vision Simulator which uses, and/or is based upon, the Brettel et alia methods of CVD simulation (see bibliography on the page).
In the case of a bar chart as in the example, other factors that are "more than color" include the position of the elements, and in the above case as repeating pairs, satisfies the more than color. But also, the luminance difference (lightness) is also part of the "more than color".
The one issue I'd raise is the way the legend is handled, as the legend has problems in related to the bars as clearly as possible, but for this simple example of two variations, probably sufficient.
The bigger issue, and the point wherein adding patterns and more significant luminance (or other) variations is needed, is in more complex dataViz: half a dozen lines criss-crossing in a line chart, a map of roads, or a pie with a lot of pieces, etc.
But in the end, keep in mind that information coding is as much information organization, and presenting in an ordered, and visually easy to perceive manner is the best practice.
Thank you for reading,
Andy
For additional reading on this subject, please see my article "The Realities And Myths Of Contrast And Color"
patrickhlauke
commented
2 years ago I am a little confused by this statement, but it may be merely a confusion of terminology.
by color I mean HUE and LIGHTNESS
I would say that for clarity, color only means hue/chroma, and not luminance (which is perceived as lightness).
see the second note on https://www.w3.org/WAI/WCAG21/Understanding/use-of-color.html
this was a long-standing fudge in WCAG 2.0, which was first hidden away in G183, but I forced it to bubble up into the understanding document https://github.com/w3c/wcag/pull/1500
while i personally don't agree with the retrospective redefinition, it apparently reflects more what AGWG intended at the time (or the addition of that loophole in G183)
Myndex
commented
2 years ago Thank you for the context Patrick @patrickhlauke
I'm familiar with #1500, I made this post in that thread, which conveys some of what I mention above, and other aspects as well.
while i personally don't agree with the retrospective redefinition, it apparently reflects more what AGWG intended at the time
I'm not clear on which redefinition you disagree with?
GreggVan
commented
2 years ago Hi Andrew.
I see what you mean. I think we have been thinking in terms of HUE and LIGHTNESS - and thinking those are the two variablea. That is Color = HUE and LIGHTNESS (e.g Red and Pink are two different colors with the same HUE but different LIGHTNESS).
CHROMA somehow got left out if it all together.
This clearly is not the full or correct story - and I have a great example below to help make (your) point to people.
(Here is a text description of the picture in advance for those who can’t see - or who have different color vision than most) The picture has two sides A and B. Left side (A) has grey background with 7 color patches on it as follows (This is MY description and comment about what I think this text on this background would look like to a person with typical vision) Kind of dark olive-green - looks like almost no contrast - would be very hard to read text in this color - but could be done slowly. Bright red - has great contrast and text would be very easily readable - though a bit eye straining to read it for long Dark turquoise maybe - would be readable but not great with normal vision A hot pink/purple - has great contrast and text would be very easily readable - though a bit garish and eye straining to read it for long A blue - again looks readable but not wonderful A burnt orange - pretty readable if it were text but not wonderful - ok for something without much text or with large text but wouldn’t want to read paragraphs of it Green - again readable but not great - ok for something without much text or with large text but wouldn’t want to read paragraphs of it On the right side (B) is a completely grey background (same exact grey) with no apparent blocks of color on it of any kind or contrast
The caption says Figure 8.1. Helmholtz–Kohlrausch Effect. (Panel) A: Various colours on a grey background, all measuring L = 50 in Photoshop. (Panel) B: same image converted to greyscale mode.
In fact all the colors are of equal CIE lightness calculated as follow L = 116 (Y/Yn)1/3 - 16; 0.008856 < Y/Yn
And indeed - if you use the Colour Contrast Analyzer — you will find that ALL of the colors come out with a contrast ratio of 1:1 with the background (identical to each other)
Now - this does not mean that the contrast defined in WCAG is not any good. It did / does move us in the right direction. But it DOES suggest that we could be finding something that is BETTER. (And that some outcomes of the current measure yield false-negative and false-positive results
I DON’T think we should define color as just hue. That would confuse many people and is not the common or plain language use of the term.
But we certainly should explore other methods for taking about contrast for readability.
Sooooo. Andrew what is your suggestion for how to define Color differently.
Gregg
On Oct 8, 2022, at 10:00 PM, Andrew Somers @.***> wrote:
Hi Gregg @GreggVan https://github.com/GreggVan (and also Hi @sarasuri https://github.com/sarasuri @detlevhfischer https://github.com/detlevhfischer and others:)
Gregg said:
I will disagree on a table being a sufficient equivalent for a chart for anyone who is not blind. And even for people who are blind a description of patterns or trends can be important. There is a reason that the charts are included or preferred for presenting many types of information. There are very few people (with or without disabilities) that can look at a table and see what is obvious from a chart.
I agree very much: graphical presentation tells the story much better than an abstract table of values. Using a table of values as a way to mitigate color perception issues with a chart is not a salient choice.
By the way remember that by color I mean HUE and LIGHTNESS. So red and pink are two different COLORS when I say "not conveyed by color alone" you do not have to use patterns if you have just a couple colors and they are different darkness/lightness.
I am a little confused by this statement, but it may be merely a confusion of terminology.
by color I mean HUE and LIGHTNESS
I would say that for clarity, color only means hue/chroma, and not luminance (which is perceived as lightness).
I know you know this Gregg, but I am going to recite for others who may be reading this thread to clarify:
Lightness/darkness/brightness are context sensitive perceptions of light volume, or luminance, separate from color as in hue/chroma/saturation. The distinction is not merely one of semantics—neurologically, achromatic luminance is processed separately from hue/chroma.
Lightness and hue are very literally separated early in the human vision system, first in the opponent process of the post-photoreceptor ganglion cells, and then in the visual cortex. V1 processing the luminance channel, and V4/V8 processing hue/chroma downstream (and delayed by about 70ms).
Of principal importance is that luminance carries all of the high spatial frequency signals (edges, patterns, fine details) while hue/chroma are sensitive to very low spatial frequencies (large areas, no details).
Not Only "Color"
As a result of the separated processing and paths, the importance of "not carried by color alone" refers not to luminance, only to hue and chroma, for the reasons of certain impairments of color insensitivity.
In reference to an sRGB display, the web default standard, and the appropriate display standard for accessibility purposes, luminance is not a factor with the exception of the sRGB red primary for the protan (and the very rare monochromat/achromatopsia). In the case of protanopia, the sRGB red primary has about half the luminance vs standard vision. Otherwise, deutan, protan, tritan have standard luminance perception, including standard or better luminance contrast sensitivity and standard or better visual function overall.
For tritan, missing the S cone is not a factor for luminance as the S cone does not contribute additively to the luminance channel. As far as luminance discrimination is concerned, with the exception of colors relying on the red primary due to the protan deficit, luminance can and should be considered separately from hue/chroma.
Unfortunately, luminance perception, meaning achromatic changes in lightness/darkness/brightness, is a generally poor way too encode information. The FAA has some standards that restrict information coding using grey levels to three: black, grey, white. The reason is that it is difficult distinguish a particular grey level without substantial context.
With standard vision there are many distinguishable hues, so hue is a more ideal method of information coding. Except of course for color insensitive vision. But even color insensitive vision distinguishes some colors. And here, with some reservations, hue/chroma and luminance together provide for coding information with "not just color".
And again, in the statement "not just color" color means only hue/chroma irrespective of luminance.
The Chart Example Above
But there is another contrast at play in the chart example above, a simple bar chart, with a pair of related bars repeated three times for three different dates.
The unmentioned contrast is the positional contrast. Each pair of bars features a bar on the left that means "Enquiries" and a bar on the right that means "Threshold". The positional contrast of these two bars in each pair is also an information coding method.
The "problem" that may arise is with the LEGEND.
In the above example, the legend has the following issues:
It is top to bottom, while the bar pairs are left to right.
In left to right reading languages, it may be intuitive that the top legend entry means the left bar in each pair. But this may not extend too right/left reading languages, nor to top/down reading languages. Other things being equal, in cases when possible, relating the spatial orientation of legend elements to the orientation of the chart elements can help improve understanding. The colors for the legend are very small squares.
As mentioned above, color (hue/chroma) are processed such that they work in very low spatial frequencies. An important result is that our hue discrimination is reduced for small stimuli. As such, a legend with tiny color samples will, especially on more complex charts, lead to misunderstandings and misreading of the color, even for standard vision. The two chart colors noted above have substantial luminance differences, as we can see in the CVD processed samples below.
Simulated CVD Examples
In the set of four below, top left is standard vision, followed clockwise by deuteranopia, protanopia, and tritanopia. Notice in this simple chart that all three of the most common forms of CVD have adequate discernibility here.
https://user-images.githubusercontent.com/42009457/194737991-6242ccca-68a3-40be-8067-9f70dbe81dc7.png Just The Luminance
So, if we process that chart to reduce to plain luminance, as if it were displayed on a black-and-white display (hey, we had those when I was a kid, LOL) it would look like:
https://user-images.githubusercontent.com/42009457/194738061-7327458c-f7f9-48c3-a095-3bdfc4f8bb94.png IMPORTANT: The above is NOT representative of achromatopsia!!!
Many of the color vision deficiency simulation tools out on the Internet today use this simple as RGB luminance claiming that it represents achromatopsia or the monochromatic, and this is completely incorrect.
Someone with Rod only achromatopsia, or a blue cone mono chrome app, will have no perception of red and a reduced perception of green.
But more important: the loss of the L and M cones means that these very rare impairments are also affected by very low vision and extreme photophobia. An experimental simulation of BCM is:
https://user-images.githubusercontent.com/42009457/194738147-bbd06220-161b-4450-95b6-ea892261c826.png Not only are no colors seen, but it is blurry and also "blown out" due to the photophobia effect exceeding the range of the rods.
Side note: rod only achromatopsia and BCM are the only "actually color blind" and are very rare in the group of color vision impairments. The others, Deu/Pro/Tri are color insensitive, but not color blind.
The Simulator Used
These were processed by The Myndex Color Vision Simulator https://www.myndex.com/CVD/ which uses, and/or is based upon, the Brettel et alia methods of CVD simulation (see bibliography on the page).
TL;DR
In the case of a bar chart as in the example, other factors that are "more than color" include the position of the elements, and in the above case as repeating pairs, satisfies the more than color. But also, the luminance difference (lightness) is also part of the "more than color".
The one issue I'd raise is the way the legend is handled, as the legend has problems in related to the bars as clearly as possible, but for this simple example of two variations, probably sufficient.
The bigger issue, and the point wherein adding patterns and more significant luminance (or other) variations is needed, is in more complex dataViz: half a dozen lines criss-crossing in a line chart, a map of roads, or a pie with a lot of pieces, etc.
But in the end, keep in mind that information coding is as much information organization, and presenting in an ordered, and visually easy to perceive manner is the best practice.
Thank you for reading,
Andy
For additional reading on this subject, please see my article "The Realities And Myths Of Contrast And Color" https://www.smashingmagazine.com/2022/09/realities-myths-contrast-color/ — Reply to this email directly, view it on GitHub https://github.com/w3c/wcag/issues/2690#issuecomment-1272455989, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACNGDXX5ABLBIPFR5HJ3W6TWCJGNFANCNFSM6AAAAAAQWD5GJI. You are receiving this because you were mentioned.
Myndex
commented
2 years ago Hi Gregg @GreggVan
I am going to respond somewhat out of order, and again, I will be covering some things I know you know, but for clarity and foundation, and as part of the record for anyone reading.
I DON’T think we should define color as just hue.
I agree, color should be at a minimum hue and the "colorfulness" of the color.
There is a specific reason that "lightness" may be over-broad here, which I am going to cover.
In art, pink is a tint of red. Pink and Red could both be at the same "lightness". In this example for red and pink, which of these is lighter?:
example of red and pink, both having approximately equivalent luminance
For our work in accessibly, there is a definite reason that it is useful to separate the understanding of color as in hue/colorfulness and achromatic lightness/darkness/brightness, and that reason is that, as I discussed in the previous post, hue/colorfulness (hue/chroma) are processed separately by the human vision system from lightness/darkness/brightness (achromatic luminance), and are subject to very different contextual relationships, particularly spatial frequency, but also, variations in perception due to the spectrum of impairments.
Color (hue/chroma) and lightness (luminance) are functionally separate, and impacted in different ways by various impairments. Some impairments affect color but not luminance, and vice versa.
This is in large part because the luminance channel is not just the "light/dark" channel—luminance is the "detail" channel. All the details, edge detection, and high spatial frequency signals, like text, exist in luminance.
Color (hue/chroma) has a third the resolution, and a fraction of the contrast sensitivity of the luminance channel for everything smaller than large detail-less patches of color. Color is also processed later, downstream, and not used in areas such as the VWFA leading to lexical processing. I.e. not useful for small thin text—on the other hand, color is very useful for object recognition.
The result is a very clear divide, particularly when considering impairments and accessibility, between hue/chroma and luminance.
Sooooo. Andrew what is your suggestion for how to define Color differently.
Well, isn't that the million dollar question... I have been working a lot in terms of writing in plain language.
I know I've been criticized here for writing too technically or pedantically, and I have been thinking a lot about plain language, or at least language aimed at a more general audience. Some examples:
"Why APCA" is a brief, plain language discussion of "lightness contrast for readability."
"The Realities and Myths of Color" is intended as a plain language overview of the "key points" of our perception, aimed as designers.
When common language used in these, it is intended to mirror that of color science, and hopefully retaining the important nuances.
We can reduce the number of terms needed to describe color (and contrasts of color differences) to three channels of visual data:
Our perception of luminance can be defined in plain language as one of, or the collection of "lightness/darkness/brightness." Of these, "lightness" appears ideal as the lead term, with darkness and brightness reserved as "additional descriptors".
For plain language, I think most understand "hue" to be how red and blue differ. Is there another way to describe hue difference other than with... "hue"? Here's an interesting article on the word "hue".
Is colorfulness the best term to use for "saturation" or "chroma"? The problem with using "saturation" that it is not universal—some color models use a chroma correlate instead of saturation. Ans some color models that call the color control "saturation" are not really "saturation" but some other non-orthogonal color control.
Other modifiers: Shade and Tint. Shade reduces colorfulness and reduces lightness (i.e. darkens). Tint reduces colorfulness and increases lightness (i.e. lightens or brightens).
However, some plain language sources indicate that tint can also mean shifting the hue in the direction of some other color. And shade could be a "darker shade" or a "lighter shade".
For the luminance channel:
For the color channel:
HUE for identity
COLORFULNESS for chroma/saturation,
NEUTRAL to define achromatic values.
Regarding Contrast:
In fact all the colors are of equal CIE lightness calculated as follow L = 116 (Y/Yn)1/3 - 16; 0.008856 < Y/Yn And indeed - if you use the Colour Contrast Analyzer — you will find that ALL of the colors come out with a contrast ratio of 1:1 with the background (identical to each other)
CIE lightness is not an absolute, it is a calculated prediction of perceptual lightness under a specific set of conditions, and does not apply well to self-illuminated displays.
Also, the 1931 CIE standard 2° observer is known to have inaccuracies, not only as a result of HK effects.
Something you might find interesting is that the "standard" CIE ∆L* as a contrast metric matches WCAG 2 contrast math in terms of indicating when to flip text from black or white for "best contrast". You can see this demonstrated at https://colorjs.io/apps/blackwhite/
Unfortunately, both are equally incorrect regarding the flip-point as it pertains to text (black/white). This is due to a major shift in the contrast center for high spatial frequency stimuli (text) as opposed to large flat patches of color.
CIE L is based on Munsell Value, which is spatially large patches of reflected colors in a defined ambient environment. As a result, L has the center point (L* 50) right at 18.4 Y, aka 18% grey. WCAG 2 math also has the "middle contrast" point at about 18% (#767676).
Depending on polarity (and proximal field), a perceptually uniform center point is about CIE 34Y to CIE 37Y for higher spatial frequencies such as small thin text and thin lines (2px and less).
But we certainly should explore other methods for taking about contrast for readability.
- for people with low vision
- for people with different color perception
- for people with no color perception
I discuss this in the article on Smashing Magazine. For all sighted users, the luminance channel is the only important channel for readability. In the HVS, most of the spatial detail information is literally discarded for the two opponent color channels (red/green and blue/yellow).
Hue/chroma are useful for discerning differences in data, such are for object recognition, or on graphical charts, but those are a separate processes from readability. Reading text requires good luminance contrast, hue/chroma are notwithstanding, and this is especially true of small thin text.
Luminance is a function solely of the L and M cones, and the L and M cones are the only cone types in the foveal central vision. In other words, we are all tritanopic in our fovea.
These are some of the reasons that when we talk about the needs for color (hue/chroma) we need to keep the luminance channel related processing tasks separate in our discussion.
Especially with the statement "don't use color as the only information coding" as this refers only to hue/chroma because luminance IS a second aspect/addition to information coding, (which you did mention), as is position, size, shape, and these other contrasts are processed independently of the hue/chroma signal.
The LMS cones lead to the 14 types of ganglion cells which "data compress" the 5 million cone cells to fit the 1.2 million optic nerve paths. Most color information is discarded in the process, what remains is mostly achromatic luminance (black/white), then two opponent color channels, red/green and blue/yellow¹, which are at a substantially reduced resolution. The luminance is processed first in V1, then the color data later in V4, V8 etc... luminance and hue/chroma are separate, we need to consider them as separate, and recognize that the separate processing and separate functions need separate considerations.
Footnote 1): there are some changes afoot regarding the understanding of the actual LMS cone to opponent process, see: https://link.springer.com/article/10.3758/s13414-020-02216-7
To make the above post more readable, the following section was moved here to the end—it provides some of the supporting background for the above conclusions.
There is much complexity and nuance here, and the frequent confusion in part stems from color science where we have a lot of terms with very narrow scope, and a subset of those same terms are used in a broader scope in plain language. Not to mention how classical art uses similar terms a little differently, while in film and broadcast engineering there are further differences, and when you add international cultural differences into the mix, the view is even murkier.
The fact that so much of the terminology we use sounds alike is a further hinderance to understanding. Luminance and luminosity and illuminance all mean very different things.
So, luminance is the brightness of the light bulb, illuminance is the light radiating into the environment as lux, until it bounces off something, and the light bouncing off a surface is again luminance, until it hits the retina, and then it's a Troland. 😳
So if all that seems easy and straight-forward (I know it doesn't, that's my idea of a joke), the fact still remains that, none of these light measurements describes the perception of lightness/darkness/brightness.
And then, how do you "measure" a human perception? Or better, predict a perception from a given set of variables? At this point we leap into the various vision/color models and that is far beyond the scope of this post, other than to say that light measures such as luminance are not perceptual.
The point I am getting at is, what can we eliminate from the discussion and still provide a good understanding and meaningful terminology?
Luminance, as an absolute measure of light (L) or a relative one (Y) is linear and additive. if you have 100 photons and triple it, you have 300 photons. But a human does not see a tripling, we see only a modest increase. HVS is non-linear in response, but that non-linearity is itself further non-linear based on multiple contextual factors.
Luminance is not a "plain language term", but it also does not describe what we most need to describe, that being "how we perceive light".
What is plain language? Language that is generally or intuitively understood? We certainly want to avoid language that is given "alternate meanings" as much as possible.
Let's start here and work our way to plain language.
We can look to the CIE's glossary for a small sample of some relevant terms:
CIE International Lighting Vocabulary
luminance Lv; L density of luminous intensity with respect to projected area in a specified direction at a specified point on a real or imaginary surface.
contrast,
brightness attribute of a visual perception according to which an area appears to emit, transmit or reflect, more or less light
lightness, <of a related colour> brightness of an area judged relative to the brightness of a similarly illuminated area that appears to be white or highly transmitting
hue attribute of a visual perception according to which an area appears to be similar to one of the colors red, yellow, green, and blue, or to a combination of adjacent pairs of these colours considered in a closed ring
colourfulness attribute of a visual perception according to which the perceived colour of an area appears to be more or less chromatic
saturation colourfulness of an area judged in proportion to its brightness
chroma colourfulness of an area judged as a proportion of the brightness of a similarly illuminated area that appears grey, white or highly transmitting
In color science we use "chroma" as the measure of the distance from an achromatic axis. Saturation is (depending on the model) chroma divided by luminance (or similar transform). As a matter of practical usefulness, saturation is a preferred correlate as (again depending on the model) more along the lines of an orthogonal property.
Though perhaps in a true plain-language, "colorfulness" is the best term?
Thinking in terms of consumer television sets—in the days of NTSC and PAL, terms related to the physical controls came into use.
For NTSC only (so, North America and Japan) there was the TINT control, sometimes called HUE.
Next, for "colorfulness" NTSC and PAL had the COLOR control, sometimes called SATURATION, and on professional gear sometimes called CHROMA.
A problem with some of the "old TV terms" is that in the old school TVs, "contrast" actually meant how bright the peak white was, this was also sometimes called "picture". And "brightness" actually meant the position of the black level or pedestal.
The need for some of these controls is in the distant past in the carefree days of unstable tube-type analog broadcast technology. In particular, I'd suggest we leave these meanings of contrast and brightness to history.
Value (luminance)
Hue
Chroma
Saturation is used a lot in art or by artists in general, and of course so is hue.
Shade/tone/tint are used in art to define adding achromatic black/grey/white (respectively) to a color, and achromatic (neutral) pigments are not usually referred to as "colors". I'd say most artists are trained to not call black or white "colors" per se, though "your mileage may vary" here, art school was a long, long time ago for me.
A tint is a mixture of a color with white, which increases lightness, while a shade is a mixture with black, which increases darkness. Both processes affect the resulting color mixture's relative saturation. A tone is produced either by mixing a color with gray, or by both tinting and shading.
Shade/tone/tint reduce saturation without affecting hue (in theory). However, perceived hue can be affected, just as perceived hue is affected by adjacent colors (context).
Note that "TINT" here has a very different meaning that "TINT" as in the hue control on an NTSC television set.
GreggVan
commented
2 years ago Like drinking from a fire hose !
Lots of good info here — but a couple questions or comments.
1) When talking about plain language — I don’t think any regular person would call the background of your image 1 "PINK". It looks mauve or something. So what I (and most people) think of a pink would never have the same lightness as red. But since we aren’t using PINK in our standard - what the color is defined as is probably not that important.
2) if I understand you correctly, for readability you said that what was most important was lightness. Yet with the examples that I provided, they all had the same lightness but vastly different readability. Or are we talking about two different things? What does the color contrast analyzer measure something besides lightness?
3) trauma is a very technical term and colorfulness sounds like a much friendlier term. However at least with chroma I don't know what it means. However with colorfulness I would think I would have some understanding and don't. If we mean chroma I would stick with that term and teach people what it means.
4) for readability contrast what exactly would you propose we use? Can you provide what you think the success criteria language should be and definitions for each of the more technical terms in the success criteria?
thanks
On Oct 9, 2022, at 10:33 PM, Andrew Somers @.***> wrote:
Hi Gregg @GreggVan https://github.com/GreggVan I am going to respond somewhat out of order, and again, I will be covering some things I know you know, but for clarity and foundation, and as part of the record for anyone reading.
I DON’T think we should define color as just hue. I agree, color should be at a minimum hue and the "colorfulness" of the color.
There is a specific reason that "lightness" may be over-broad here, which I am going to cover.
In art, pink is a tint of red. Pink and Red could both be at the same "lightness". In this example for red and pink, which of these is lighter?:
https://user-images.githubusercontent.com/42009457/194791956-ea15c3e4-c4a2-4f00-b7b3-0ad373cb8ae1.pngexample of red and pink, both having approximately equivalent luminance The importance of separating perception of luminance vs hue/chroma
For our work in accessibly, there is a definite reason that it is useful to separate the understanding of color as in hue/colorfulness and achromatic lightness/darkness/brightness, and that reason is that, as I discussed in the previous post, hue/colorfulness (hue/chroma) are processed separately by the human vision system from lightness/darkness/brightness (achromatic luminance), and are subject to very different contextual relationships, particularly spatial frequency, but also, variations in perception due to the spectrum of impairments.
Color (hue/chroma) and lightness (luminance) are functionally separate, and impacted in different ways by various impairments. Some impairments affect color but not luminance, and vice versa.
This is in large part because the luminance channel is not just the "light/dark" channel—luminance is the "detail" channel. All the details, edge detection, and high spatial frequency signals, like text, exist in luminance.
Color (hue/chroma) has a third the resolution, and a fraction of the contrast sensitivity of the luminance channel for everything smaller than large detail-less patches of color. Color is also processed later, downstream, and not used in areas such as the VWFA leading to lexical processing. I.e. not useful for small thin text—on the other hand, color is very useful for object recognition.
The result is a very clear divide, particularly when considering impairments and accessibility, between hue/chroma and luminance.
Plain Color Terms?
Sooooo. Andrew what is your suggestion for how to define Color differently.
Well, isn't that the million dollar question... I have been working a lot in terms of writing in plain language.
I know I've been criticized here for writing too technically or pedantically, and I have been thinking a lot about plain language, or at least language aimed at a more general audience. Some examples:
"Why APCA" https://git.apcacontrast.com/documentation/WhyAPCA is a brief, plain language discussion of "lightness contrast for readability."
"The Realities and Myths of Color" https://www.smashingmagazine.com/2022/09/realities-myths-contrast-color/ is intended as a plain language overview of the "key points" of our perception, aimed as designers.
When common language used in these, it is intended to mirror that of color science, and hopefully retaining the important nuances.
For the minutiae, see "Background" at the very end of this post.
Defining some plain terms
We can reduce the number of terms needed to describe color (and contrasts of color differences) to three channels of visual data:
Luminance (lightness/darkness/brightness) colorfulness (saturation or chroma) color identity (hue) Our perception of luminance can be defined in plain language as one of, or the collection of "lightness/darkness/brightness." Of these, "lightness" appears ideal as the lead term, with darkness and brightness reserved as "additional descriptors".
For plain language, I think most understand "hue" to be how red and blue differ. Is there another way to describe hue difference other than with... "hue"? Here's an interesting article on the word "hue". https://www.worldwidewords.org/articles/hue.htm Is colorfulness the best term to use for "saturation" or "chroma"? The problem with using "saturation" that it is not universal—some color models use a chroma correlate instead of saturation. Ans some color models that call the color control "saturation" are not really "saturation" but some other non-orthogonal color control.
Other modifiers: Shade and Tint. Shade reduces colorfulness and reduces lightness (i.e. darkens). Tint reduces colorfulness and increases lightness (i.e. lightens or brightens).
However, some plain language sources indicate that tint can also mean shifting the hue in the direction of some other color. And shade could be a "darker shade" or a "lighter shade".
TL;DR - The Terms:
For the luminance channel:
LIGHTNESS, and if describing certain contexts, DARKNESS and BRIGHTNESS. For the color channel:
HUE for identity
COLORFULNESS for chroma/saturation,
though chroma or saturation are appropriate when discussing a specific model that uses one or the other. NEUTRAL to define achromatic values.
Regarding Contrast:
The value generated by APCA is Lc which is "lightness contrast". I use the term lightness contrast as APCA is weighted for human perception of supra-threshold contrast of high-spatial-frequency stimuli on self-illuminated displays. Other Comments to Gregg:
In fact all the colors are of equal CIE lightness calculated as follow L = 116 (Y/Yn)1/3 - 16; 0.008856 < Y/Yn And indeed - if you use the Colour Contrast Analyzer — you will find that ALL of the colors come out with a contrast ratio of 1:1 with the background (identical to each other)
CIE lightness is not an absolute, it is a calculated prediction of perceptual lightness under a specific set of conditions, and does not apply well to self-illuminated displays.
Also, the 1931 CIE standard 2° observer is known to have inaccuracies, not only as a result of HK effects.
Something you might find interesting is that the "standard" CIE ∆L* as a contrast metric matches WCAG 2 contrast math in terms of indicating when to flip text from black or white for "best contrast". You can see this demonstrated at https://colorjs.io/apps/blackwhite/ https://colorjs.io/apps/blackwhite/ Unfortunately, both are equally incorrect regarding the flip-point as it pertains to text (black/white). This is due to a major shift in the contrast center for high spatial frequency stimuli (text) as opposed to large flat patches of color.
CIE L is based on Munsell Value, which is spatially large patches of reflected colors in a defined ambient environment. As a result, L has the center point (L* 50) right at 18.4 Y, aka 18% grey. WCAG 2 math also has the "middle contrast" point at about 18% (#767676).
Depending on polarity (and proximal field), a perceptually uniform center point is about CIE 34Y to CIE 37Y for higher spatial frequencies such as small thin text and thin lines (2px and less).
Readability
But we certainly should explore other methods for taking about contrast for readability.
for people with low vision for people with different color perception for people with no color perception I discuss this in the article on Smashing Magazine https://www.smashingmagazine.com/2022/09/realities-myths-contrast-color/. For all sighted users, the luminance channel is the only important channel for readability. In the HVS, most of the spatial detail information is literally discarded for the two opponent color channels (red/green and blue/yellow).
Hue/chroma are useful for discerning differences in data, such are for object recognition, or on graphical charts, but those are a separate processes from readability. Reading text requires good luminance contrast, hue/chroma are notwithstanding, and this is especially true of small thin text.
Luminance is a function solely of the L and M cones, and the L and M cones are the only cone types in the foveal central vision. In other words, we are all tritanopic in our fovea.
These are some of the reasons that when we talk about the needs for color (hue/chroma) we need to keep the luminance channel related processing tasks separate in our discussion.
Especially with the statement "don't use color as the only information coding" as this refers only to hue/chroma because luminance IS a second aspect/addition to information coding, (which you did mention), as is position, size, shape, and these other contrasts are processed independently of the hue/chroma signal.
Closing note
The LMS cones lead to the 14 types of ganglion cells which "data compress" the 5 million cone cells to fit the 1.2 million optic nerve paths. Most color information is discarded in the process, what remains is mostly achromatic luminance (black/white), then two opponent color channels, red/green and blue/yellow¹, which are at a substantially reduced resolution. The luminance is processed first in V1, then the color data later in V4, V8 etc... luminance and hue/chroma are separate, we need to consider them as separate, and recognize that the separate processing and separate functions need separate considerations.
Footnote 1): there are some changes afoot regarding the understanding of the actual LMS cone to opponent process, see: https://link.springer.com/article/10.3758/s13414-020-02216-7 https://link.springer.com/article/10.3758/s13414-020-02216-7 POSTSCRIPT: The Minutiae
To make the above post more readable, the following section was moved here to the end—it provides some of the supporting background for the above conclusions.
Background
There is much complexity and nuance here, and the frequent confusion in part stems from color science where we have a lot of terms with very narrow scope, and a subset of those same terms are used in a broader scope in plain language. Not to mention how classical art uses similar terms a little differently, while in film and broadcast engineering there are further differences, and when you add international cultural differences into the mix, the view is even murkier.
The fact that so much of the terminology we use sounds alike is a further hinderance to understanding. Luminance and luminosity and illuminance all mean very different things.
Luminance (L) is a physical measure of light, measured per the surface of the emitter, spectrally weighted for the human visual system (HVS), using a spot meter for example. the unit is "nit" or cd/m² Luminance (Y) is a relative measure of light, relating to an abstracted physical measurement where 0 is black and 1 (or 100) is white of a perfect Lambertian reflector. Illuminance a physical measure of light as it is radiating from the emitter, but measured a distance from the emitter, and not of the surface of the emitter, using a meter with an integrating sphere for example, or put another way, Illuminance is the density of incident luminous flux with respect to area. The unit is "Lux". Luminosity is point-light over time, used in astronomy. BUT... But the "Luminosity Function" is the curve that describes the HVS weighting of luminance, though "Luminosity Function" is generally phased out as archaic and replaced with "luminous efficiency function" due to the obvious confusion ("luminosity" being used relative to vision still exists sometimes in some texts relating to the field of CVD due to historical reasons). luminous color is perceived to belong to an area that appears to be emitting light as a primary light source. Luminous flux is a quantity derived from the radiant flux https://cie.co.at/eilvterm/17-21-038, Φe, by evaluating the radiation according to its action upon the CIE standard photometric observer https://cie.co.at/eilvterm/17-21-036. So, luminance is the brightness of the light bulb, illuminance is the light radiating into the environment as lux, until it bounces off something, and the light bouncing off a surface is again luminance, until it hits the retina, and then it's a Troland. 😳
So if all that seems easy and straight-forward (I know it doesn't, that's my idea of a joke), the fact still remains that, none of these light measurements describes the perception of lightness/darkness/brightness.
And then, how do you "measure" a human perception? Or better, predict a perception from a given set of variables? At this point we leap into the various vision/color models and that is far beyond the scope of this post, other than to say that light measures such as luminance are not perceptual.
The Perceptual Point
The point I am getting at is, what can we eliminate from the discussion and still provide a good understanding and meaningful terminology?
Luminance, as an absolute measure of light (L) or a relative one (Y) is linear and additive. if you have 100 photons and triple it, you have 300 photons. But a human does not see a tripling, we see only a modest increase. HVS is non-linear in response, but that non-linearity is itself further non-linear based on multiple contextual factors.
Luminance is not a "plain language term", but it also does not describe what we most need to describe, that being "how we perceive light".
Plain Talk About Language
What is plain language? Language that is generally or intuitively understood? We certainly want to avoid language that is given "alternate meanings" as much as possible.
Color Science
Let's start here and work our way to plain language.
We can look to the CIE's glossary for a small sample of some relevant terms:
CIE International Lighting Vocabulary https://cie.co.at/e-ilv luminance Lv; L density of luminous intensity https://cie.co.at/eilvterm/17-21-045 with respect to projected area in a specified direction at a specified point on a real or imaginary surface.
contrast, perceived contrast assessment of the difference in appearance of two or more parts of a field seen simultaneously or successively. EXAMPLE: Brightness contrast, lightness contrast, color contrast, simultaneous contrast, successive contrast.
brightness attribute of a visual perception https://cie.co.at/eilvterm/17-22-082 according to which an area appears to emit, transmit or reflect, more or less light
lightness, <of a related colour https://cie.co.at/eilvterm/17-22-047> brightness https://cie.co.at/eilvterm/17-22-059 of an area judged relative to the brightness of a similarly illuminated area that appears to be white or highly transmitting
hue attribute of a visual perception https://cie.co.at/eilvterm/17-22-082 according to which an area appears to be similar to one of the color https://cie.co.at/eilvterm/17-22-040s red, yellow, green, and blue, or to a combination of adjacent pairs of these colour https://cie.co.at/eilvterm/17-22-040s considered in a closed ring
colourfulness attribute of a visual perception https://cie.co.at/eilvterm/17-22-082 according to which the perceived colour https://cie.co.at/eilvterm/17-22-040 of an area appears to be more or less chromatic
saturation colourfulness https://cie.co.at/eilvterm/17-22-072 of an area judged in proportion to its brightness https://cie.co.at/eilvterm/17-22-059 chroma colourfulness https://cie.co.at/eilvterm/17-22-072 of an area judged as a proportion of the brightness https://cie.co.at/eilvterm/17-22-059 of a similarly illuminated area that appears grey, white or highly transmitting
Regarding Colorfulness
In color science we use "chroma" as the measure of the distance from an achromatic axis. Saturation is (depending on the model) chroma divided by luminance (or similar transform). As a matter of practical usefulness, saturation is a preferred correlate as (again depending on the model) more along the lines of an orthogonal property.
Though perhaps in a true plain-language, "colorfulness" is the best term?
Next, let's look at some common language examples.
From Consumer Electronics:
Thinking in terms of consumer television sets—in the days of NTSC and PAL, terms related to the physical controls came into use.
For NTSC only (so, North America and Japan) there was the TINT control, sometimes called HUE.
Next, for "colorfulness" NTSC and PAL had the COLOR control, sometimes called SATURATION, and on professional gear sometimes called CHROMA.
A problem with some of the "old TV terms" is that in the old school TVs, "contrast" actually meant how bright the peak white was, this was also sometimes called "picture". And "brightness" actually meant the position of the black level or pedestal.
The need for some of these controls is in the distant past in the carefree days of unstable tube-type analog broadcast technology. In particular, I'd suggest we leave these meanings of contrast and brightness to history.
Munsell
Value (luminance) Hue Chroma
Some Terms in Art
Saturation is used a lot in art or by artists in general, and of course so is hue.
Shade/tone/tint are used in art to define adding achromatic black/grey/white (respectively) to a color, and achromatic (neutral) pigments are not usually referred to as "colors". I'd say most artists are trained to not call black or white "colors" per se, though "your mileage may vary" here, art school was a long, long time ago for me.
A tint is a mixture of a color with white, which increases lightness, while a shade is a mixture with black, which increases darkness. Both processes affect the resulting color mixture's relative saturation. A tone is produced either by mixing a color with gray, or by both tinting and shading.
Shade/tone/tint reduce saturation without affecting hue (in theory). However, perceived hue can be affected, just as perceived hue is affected by adjacent colors (context).
Note that "TINT" here has a very different meaning that "TINT" as in the hue control on an NTSC television set.
— Reply to this email directly, view it on GitHub https://github.com/w3c/wcag/issues/2690#issuecomment-1272806831, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACNGDXTLQXURSNXB7UNM4DLWCOTEFANCNFSM6AAAAAAQWD5GJI. You are receiving this because you were mentioned.
Myndex
commented
2 years ago Hi Gregg @GreggVan
Like drinking from a fire hose!
Yes, I know I suffer from an egregious prolixity.
But also, this is a complicated subject and not well described outside of some very thick books... 😎
1) When talking about plain language — I don’t think any regular person would call the background of your image 1 "PINK"....
Well, color naming is an entire discipline unto itself, and I wasn't trying to open that can of worms, so much as indicate the ambiguous nature of the subject. But better to look at the color in isolation:
Wikipedia indicates this as between "Solid Pink" and "New York Pink" and lighter than "Mauve Taupe", with Mauve itself being more purple. Of course, shown against the red/orange of the sRGB primary, #a87376 may indeed seem a bit purple.
Other names: this is close to Pantone 7523 U, and is a lighter tint of Pantone's 2015 color of the year "Marsala 18-1438" and some other color names are "Carley's Rose" (Sherwin Williams), "Handmade Red", or Munsell 5.9R 5.2/4.3
And we could bash away at "what color is this really" all day and not get to the point I'm trying to make, so instead let's consider alternate examples, again a red and a pink with a "calculated" equivalent luminance:
So to demonstrate contrast as it applies to each of the three most common axis¹ of perceptual color models: lightness, Hue, colorfulness which are the human perceptions of luminance, wavelength, and colorimetric purity, here are some examples with a single contrast axis isolated, the other axes being null or neutral.
The measurement model used is CIELUV, using the SAPC tool in research mode. Luminance using SAPC $YS$ (estimated screen luminance, which is calculated slightly differently as it is based on calibrated physical monitors in a reference environment)._ Some insignificant deviations from null are a result of 8-bit precision and/or practical issues with color picker interfaces.
Note [1] Some color appearance models are more elaborate, with separate correlates for brightness ( $Q$ ) in addition to lightness ( $J$ ), but for our purposes here we will focus on the simpler perceptual lightness ( $L^*$ or Lstar) and some derivations ( $L^S$ and $L^C$ ).
These samples are maximum hue contrast (180°) but zero chroma contrast and zero luminance contrast. Here, chroma is nulled out about $LUV \ 44 \ C$, luminance is nulled about $Y_S \ 22$. Colors are b77877 vs 00988d.
Color data click to enlarge:
These samples are near maximum chroma contrast, that is, if the maximum chroma for a given hue is 120, then the most saturated is set to 120, and here the lower chroma color is set to 20, so the chroma contrast (distance) is about five times the chroma for these samples. Chroma is isolated, meaning zero hue contrast and zero luminance contrast. Here hue is about 315° and $Y_S$ about 22.5. Colors are 9e8099 vs ed00d5.
Color data click to enlarge:
These samples are at zero chroma contrast and zero hue contrast. We then have four pairs of samples at increasing luminance contrasts, all while maintaining zero hue and zero chroma contrast. Consequentially, this means that the saturation contrast DOES change, as saturation is chroma/luminance in the LUV model—but for this demo, chroma is what we want to hold static.
The luminance contrasts will be defined as perceptual lightness contrast ( $L^C$ ), approximately $L^C \ 15, L^C \ 30, L^C \ 45$ , and $L^C \ 60$ . The equivalent WCAG 2 contrast will be listed for each sample set. For this group, hue is about 320° and chroma is about 50.
Per APCA guidelines, $L^C \ 15$ is the minimum lightness contrast for large/thick non-semantic non-text elements. No text permitted at this level (except decorative). Colors are 944a85 on 690059, which is ~2:1 under WCAG_2. 2:1 incidentally is the FAA minimum for text (using slightly different math).
Nevertheless, under the current APCA guidelines, text is not permitted at the Lc15 level, as this is considered the "level of invisibility" for impairments with significant contrast sensitivity loss.
Color data click to enlarge:
Per APCA guidelines, Lc 30 is the minimum lightness contrast for any text (except decorative), mainly addressing "spot readable" text, and also the minimum for large/solid semantic non-text elements. Colors are b76fa7 on 690059, which is ~3.3:1 under WCAG_2.
Color data click to enlarge:
Per APCA guidelines, Lc 45 is the minimum lightness contrast for large/bold content text, and smaller "spot readable" text. Colors are d68fc6 on 690059, which is ~4.9:1 under WCAG_2.
Color data click to enlarge:
Per APCA guidelines, Lc 60 is the minimum lightness contrast for most normal content content text, and smaller "sub-fluent" text. However, body-text minimum is Lc 75. Colors are ffadea on 750064, which is ~6.4:1 under WCAG_2.
Color data click to enlarge:
As we can see from the above demonstration, even maximum hue or chroma difference/contrast is not helpful to our goals of readability, and not helpful for perceivability of small details.
On the other hand, even small differences in luminance/lightness have a substantial effect on readability and perceivability of small details.
When I use the term "lightness" I and making a plain language simplification of "the human perception of luminance in a given context". Lightness is the perception that results from viewing a given luminance at a given light-adaptation state.
Lightness (luminance) is processed differently by our neurology than color (hue/chroma), and is subject to very different parameters, and is used differently by our neurological and cognitive processes.
Therefore, everything I have been writing has held this as an important understanding, and it is why I bring the subject up: the lightness, as in luminance, of a given color is a separate property, and not a part of the "color" as recited in the "must use more than just color" axiom.
2) if I understand you correctly, for readability you said that what was most important was lightness. Yet with the examples that I provided, they all had the same lightness but vastly different readability. Or are we talking about two different things? What does the color contrast analyzer measure something besides lightness?
I'm not sure I understand your question? If the text and the background have the same lightness, i.e. no luminance contrast, they will be essentially unreadable to everyone.
It is not possible to claim readability based on large (low spatial frequency) patches of color. It is not possible to ignore spatal frequency, and spatal frequency is the primary driver of contrast perception, more than color, and especially for readability. It is incorrect to assume that color distance equals readability without considering the effect of spatial frequency. SEE:
When you said earlier:
Bright red - has great contrast and text would be very easily readable - though a bit eye straining to read it for long
Depends very much on the spatial frequency—are we talking large bold text or small thin text? Here is the full red primary on an achromatic grey background with equivalent luminance and equivalent perceptual lightness (HK effects notwithstanding). Do you find this readable?
I can see the gigantic ampersand, but I can not read the three lines of sample text. There is no luminance contrast, though there is maximum chroma contrast.
And also, keep in mind that the "assumed luminance" of red vs achromatic grey, as calculated using the IEC standard which is derived from the 1931 CIE standard observer is not accurate enough for the kinds of hair-splitting we are doing here. On this point, I am presently considering moving to the Judd/Vos standard observer instead for these and related reasons. At the moment, APCA is using a modified luminance calculation with a very mild protan compensation that also works to alleviate some of these discrepancies.
My plan though has been to move to Judd/Vos (or one of the others) and handle the protan compensation with the separate (and as yet unreleased) color module.
3) [chroma] is a very technical term and colorfulness sounds like a much friendlier term. However at least with chroma I don't know what it means. However with colorfulness I would think I would have some understanding and don't. If we mean chroma I would stick with that term and teach people what it means.
Well, when talking about the "amount of color" may mean either chroma OR saturation, depending on the color model. So I am trying to think of a way to be encompassing and descriptive, and not too narrow. Chroma is has a very specific meaning and is not always related to a physical color control. Saturation is a better and more intuitive term, and used in art. "Color level" is another, or "color volume" ...
Colorfulness encompasses the variations through different color models in terms of the correlate that adjusts chroma or saturation etc.
4) for readability contrast what exactly would you propose we use? Can you provide what you think the success criteria language should be and definitions for each of the more technical terms in the success criteria? thanks
I have written about this extensively, actually, and this was also a topic of discussion in the Visual Contrast subgroup, and has been given much thought and exploration—particularly in terms of conveying a challenging subject to the public, and artists/designers/developers in particular.
I introduce the concept of perceptual lightness contrast and readability contrast in the brief overview Why APCA:
"Why APCA" is a link I post often, as I encourage everyone with questions regarding perceptually uniform contrast to start there. It's also listed at the top of the linktree, which is a short curated list of resources.
The general APCA documentation page with more technical information, links, formulas, and glossary is in this readme file.
And the main catalog of contrast and color resources, including white papers, plain language articles, comparative studies, third-party reviews, and extended documentation down to the minutiae is:
Also, there is the APCA Discussion Forum where I post documentation works-in-progress and answer questions, list tools & adopters, etc.
And yes, I am aware that there is a MASS of materials to go through. This is a complicated subject, and I am working on distilling to a manageable set of documents. Part of the start to this is the recent article at Smashing Magazine which is in-depth but also intended as simplified language to the extent possible.
Please let me know if I can answer further questions.
Thank you for reading,
Andy
Myndex
commented
2 years ago @GreggVan Just a quick add regarding this last point
Can you provide what you think the success criteria language should be
Some of this is/was already in drafts of material I wrote for WCAG_3.
My most recent pull request in the Silver repo is PR #630 and incorporates a lot of the current state of the art here. The content of the PR 630 breaks things down into actionable steps.
fstrr
commented
6 months ago This issue is labelled as a discussion, so we’re moving this to Discussions. There doesn’t seem to be an update to make to the documentation, but if that changes, we can move it back to the issues list.
Hi
Have a question on Success Criterion 1.4.1: Use of Color and graphs.
When a web page has graphs for example a bar chart that uses color differences to convey information. In addition, it has a data table with same information below the graph. Is providing a data table with the same information considered as sufficient to meet the Success Criterion 1.4.1: Use of Color?