GitHub Community
The new colorblind scheme is inaccessible for people with tritanopia #6385
Replies: 8 comments 40 replies
-
|
Hi @MilesBHuff! Thank you for the detailed feedback and I apologize for the delay in getting back to you. We have been listening to all the feedback closely since the beta launch of our colorblind themes. We have quite a few people at GitHub who are colorblind and helped us develop the color variables pre-beta launch, but as colorblindness affects people so differently we definitely went into the beta knowing we still had a lot to learn and improve. We are currently working on a new iteration of the themes that is exploring the following improvements across GitHub:
We will continue to improve these new themes and look forward to additional feedback from the community so we can make them as inclusive as possible! |
Beta Was this translation helpful? Give feedback.
-
This statement isn't very true -- while Protanopia and Deuteranopia do pattern together, Tritanopia has a completely different bias on the color wheel. It is easy to create a color scheme that prots and deuts can see, but trits cannot. Magenta/green is literally the only one that works for all three. |
Beta Was this translation helpful? Give feedback.
-
|
@mark-wiemer https://en.wikipedia.org/wiki/Monochromacy#Humans https://www.daxx.com/blog/development-trends/number-software-developers-world Assuming cone-monochromacy does not make someone more or less likely to become a programmer (and vice-versa), then multiplying the two figures together gets us roughly 594 programmers with cone-monochromacy. It is likely that some of those users do not use GitHub; but let's assume they do. So there are probably around 600 people who might benefit from monochromacy accessibility at GitHub, assuming GitHub is only used by programmers. EDIT: @gaknoll |
Beta Was this translation helpful? Give feedback.
-
|
Thanks @MilesBHuff for the continued detailed feedback & number crunching! We briefly paused additional work on the colorblind themes to ship a beta of our new light high contrast theme yesterday, but are returning to this work now that this new theme out in the wild for feedback. Based on our design explorations and your feedback we are now looking to separate out our colorblind themes into two groups:
I admit we were a bit overly optimistic about being able to find color scales that made GitHub more accessible to all three groups, but as you called out that was pretty much impossible without introducing compromises to all 3. These compromises also don't align with the guiding principles that led us to ship these themes in the first place: create themes designed to make GitHub a more inclusive home for all developers. We will first focus on updating the current colorblind themes to be solely Protanopia & Deuteranopia, but when we move into testing for the Tritanopia themes (either later this year or in early January) would you be willing to jump on a call and share your feedback? That would be much appreciated 🙏🏽 |
Beta Was this translation helpful? Give feedback.
-
|
Oh. Sure, lol; just know I'm not a credentialled expert on the subject -- I'm just an Internet person with a colorblindness simulator. I'm not even colorblind, myself. |
Beta Was this translation helpful? Give feedback.
-
Hi @mark-wiemer Yes and actually there are more people with achromatopsia than tritanopia. Color Blind vs Color DeficientBut achromatopsia and tritanopia have completely different use or need profiles. Deutan, Protan, and tritan (-opia and -anomalous) are color vision deficient (CVD) they are not really "color blind". Other than the hue insensitivity, they have equivalent or better visual function overall, including good luminance contrast sensitivity. Achromatopsia (rod only) and blue cone monochromatism (BCM) are not only actually colorblind they are low-vision or legally blind as well. Brief ExplainerS cones are few and scattered in the periphery only, and while rods out number cones by orders of magnitude, many rods are joined together in groups, each group to a single neural signal. As a result, the acuity even with correction is 20/70 or worse. In addition, rods are the most sensitive photoreceptors, used only for night vision, so lighter environments max them out (bleaching) causing severe photophobia—many wear sunglasses indoors. Congenital achromatopsia (1 in 33,000) and BCM (1 in 100,000) require assistive technology, at the very least setting their screen to mono and adjusting the brightness to no more than 8 cd/m². This is essentially out of scope for content authors. Cerebral achromatopsia and red or green mono are exceedingly rare. TritanTritan (-opia and -anomalous) (blue/violet insensitive) are about 1 of 50,000.... but even standard vision is tritonopic in our central foveal vision (here are no S cones in the fovea). Tritanopia, though missing all S cones, does not have any luminance or reading deficit, as S cones do not contribute to luminance nor detail. Their main issue is distinguishing between colors that contain blue primary, such as cyan versus green or purple/magenta versus red. The Myndex CVD Sim includes an experimental blue cone mono sim, which includes poor acuity and photophobia simulation. Thank you for reading |
Beta Was this translation helpful? Give feedback.
-
|
Hey Miles, this is a great trick! Just curious, do you consider Coblis to be a decent colorblindness simulator? If not, what would you recommend? |
Beta Was this translation helpful? Give feedback.
-
|
@VulumeCode I've definitely noticed that many "colorblind" themes use blue and yellow in a way that is counterintuitive for dichromats; and by that, I mean directly opposite what the normal associations would be. |
Beta Was this translation helpful? Give feedback.
-
|
As a green-weak person, red/green to yellow/blue just makes me do a double take each time. Yellow feels like the replacement for green, but surely blue can't be the replacement for red, so the yellow must have been the red. The distinction is clear to my eyes but my brain has to do more work. Green and magenta contrast very clearly, but it's garish. |
Beta Was this translation helpful? Give feedback.
-
|
Hi Mark @mark-wiemer
While Coblis has been around for a long time, it's rather poor in terms of accuracy or quality. I think simulators based on Brettel/Viénot are more correct. And here's a recent review that agrees with my assessment: Myndex CVD SimA few years ago I built the Myndex CVD simulator based around Brettel/Viénot 97/99, as part of the research that we are doing in accessible readability. It's free to use and you can process your own images on your local machine in JS. It displays a simulation of all of the major kinds of CVD all at once. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks @Myndex! I've bookmarked the simulator you sent. |
Beta Was this translation helpful? Give feedback.
-
As I showed in a different post, red Other things being equal, Green should Typically be the lightest color, red or blue should Generally be the darkest color. Full Blue should never be substituted for full green and full yellow should never be substituted for full red, but outside of that, color language is so imprecise it's not possible to make blanket statements about which color should be switched into some other color for any given purpose. Do you have an example where someone is trying to make a CVD accessible pallet by doing that?
Magenta is a very problematic color, just because it is distinguishable next to green does not mean that It's a good pair with green, And generally it's not. Green and magenta Is a nearly unreadable combination as an example, do to glare, chromatic aberration, and stereopsis. |
Beta Was this translation helpful? Give feedback.
-
Hi @VulumeCode Yes, this is why lately I prefer the term "color insensitive" as opposed to color blind, and I try to use that most of my writing, using colorblind only to describe those with achromatopsia or blue cone mono. Welcome to hear your opinions regarding descriptive terminology. The Myndex CVD simulator does simulate deuteranopia, protanopia, tritanopia, and not the anomalous levels for a couple reasons. First, it's based on the Brettel model, which is the most well researched in terms of matching clinically to actual perception for the -opia types. Other models seem to be comparing themselves to Brettel et alia. Though we are looking at other models to potentially adapt. But the second reason is in researching for accessibility, looking for the "maximum" cases to find fail points, and currently the primary focus is readability. The midrange cases, the anomalous types, are a part of live studies and live research that we're also conducting. Please let me know if you have any interest in that and/or commenting on anything that we're working on.
The Myndex sim does not use orange for deutan/protan—so I should ask, does the Myndex Sim appear orange to you? Being based on the Brettle models, the deutan/protan are shown blue/yellow, this is based in part on the functioning of the opponent color process of human vision, but also including studies with unilateral color insensitivity, where somebody is color insensitive in one eye but not the other, allowing them to give feedback in terms of the accuracy of colors chosen. So again this is related to the -opia not -anomalous types.
I'm not quite sure what do you mean? Tritanopia is missing the S cone, which is the blue/violet related cone. That does not mean they don't see blue. Here is a chart of normalized cone functions:
Notice that the cone responses for L, M, & S all overlap very substantially. L & M cones both actually peak in green, the L cones peak in a yellowish green, but extends far into red, M cones peak in a more teal green. S cones peak toward violet.
Something we are working on are best ways to shift a color palette for Daltonization to accommodate various levels of color insensitivities. One of the issues with Computer monitors, is that they only display three colors, red, green, and blue. Yellow on a computer monitor for instance is not yellow, it's still just red and green at the same time, very close together. It doesn't mix in the air like paint or something it mixes in your neurology. There are implications here for how color insensitive vision responds to independent primaries as they exist on different kinds of monitors, an area of continuing research. |
Beta Was this translation helpful? Give feedback.
-
GitHub uses orange for - in the deutan/protan scheme. As a green-weak person, the Myndex Sim default looks yellow to me. The Myndex Sim used on the default GitHub diff color scheme looks like a gross brown (+ and -).
Just saying that as a weak-green person I prefer the Tritanopia scheme for diffs. It respects that I can still see and understand red as -. Then I can reason by negation that blue means +. Also if you take a screenshot of a diff with the Tritanopia scheme and put it through the Myndex Sims, it looks just like the color scheme for deutan/protan. So I'm guessing even green-blind people could work with the Tritanopia scheme. |
Beta Was this translation helpful? Give feedback.
-
Hi @VulumeCode Ah, yes… Daltonizan is sometimes useful for complex things like images, and it can work for user your components ... As far as reading is concerned, nobody needs a special scheme, everyone benefits by never putting red on black and never putting blue on black. I see some attempts creating color palettes that don't align with normal expectations in an effort to make sure that things are "visible" as if colors were just going to vanish or disappear. They never really "vanish" on an sRGB display. In the worst case, red becomes about half-as-bright for somebody with protanopia. This means that red should not be used against black, particularly for text. Pure blue or pure red should never be up against black, if it is something that needs to be discerned or read. Pure blue has low resolution in the human vision system, S cones are scattered in the periphery and do not contribute to luminance nor to detail information. Red has very little contribution to luminance and for protanopia the contribution is cut in half. But also people hear the phrase "can't distinguish red/green" it's misinterpreted, red has such a lower luminance it is distinguishable:
But notice that for protanopia, Green and yellow are not distinguishable. But because of the luminous difference you can definitely see the difference between the red and the green. And here's where we start to run into a lot of these problems of understanding. And the next important question is, does it need to be a color that by itself is recognizable? |
Beta Was this translation helpful? Give feedback.
-
|
Hi Miles @MilesBHuff
The short answer: it is essentially the same as with other aspects of vision—you want a vision model that is close to perceptually uniform, or at least within the range you need is relative to actual perception. When modeling aspects of color (hue/chroma) you typically want to have your model doing the "Work" in (or close to) LMS cone space, in other words using reference cone data from empirical studies. DaltonLens has a good review of simulators, and I agree with his conclusions. We found pretty much the same from our research a few years ago building the Myndex color vision deficiency sim. mainly that Coblis and HCIRN & the other related sims were very inaccurate. We built ours because we needed an accurate sim to support our research into accessible readability for web content. Next post will have a brief encapsulation of important CVD considerations, stay tuned... |
Beta Was this translation helpful? Give feedback.
-
|
I'd like to comment on some of the thoughts that were brought up regarding accommodating user needs for those with CVD. I'm hoping this is the appropriate thread. There's additional background in each of the "spoilers", which are used to hide the bulk of the writing to make the post easier to read over all. The discussion below refers to the web standard color space sRGB, the narrowest standard color space, that requires the least heroic accommodation. Color Vision Deficiency and Colorblind Fast Facts
CLICK TO EXPAND CVD FACTS### Population and rarity As far as most common congenital (genetic) color insensitivities, deutan and protan (green or red insensitive respectively) affect 1 out of 12 men of European ancestry, but those numbers change very rapidly with different genetic backgrounds. Most reasonable estimate I've seen place the International population for all deutan and protan CVD types combined at about 5%, so a very common condition at 1 out of 20 people overall.The common deutan / protan CVD types have equivalent or better visual function as standard vision including equivalent or better luminance contrast sensitivity. Deutan have an enhanced ability to recognize camouflage compared to standard vision. The only significant functional luminance deficit is with the protan, in the case of red against black because they see red darker. Tritan: a rare CVD with anomalous or missing S cones (missing blue violet), estimates are about 1 out of 50,000. However it is useful to know that even standard vision is tritanopic in the central foveal vision, because the S cones are loosely scattered only in the periphery, and make up only about 2% to 7% of the total cones in normal vision. Because triton are only missing the S cones, which offer very little in terms of functional resolution, and are not a part of luminance, triton have no deficit in terms of readability, and their only deficit is in a reduced color gamut for discerning color-coded data. The following rare cases are not CVD, they are actually colorblind.Congenital rod-only Achromatopsia: 1 out of 30k-40k. US population is believed to be a little less than 10,000 for achromatopsia acquired through genetic background. Cerebral Achromatopsia: is exceedingly rare, Bartolomeo, 2013 discussed the very few known cases totaling perhaps two dozen patients over 40 years or so. the condition is caused by Brain trauma such as a stroke. Generally they are not co-morbid with low vision or photophobia as the standard cones are present on the retina. Blue Cone Monochromatism (BCM): presenting very similarly to congenital Achromatopsia, is co-morbid with low vision and photophobia. As we know from standard vision there are no S cones in central foveal vision, S cones are not part of luminance, and S cones carry very little detail as they are at a very low resolution. Red or Green Cone Monochromatism: there is some controversy whether or not this condition actually exists, and if it does it is so extremely rare, as in one out of 1 million to 2 million, there is very little information about the condition. Nevertheless since the L M cones make up luminance, which carries all fine detail, we can assume that they don't suffer the severe low vision or photophobia of BCM, and do see things close to a black and white / grayscale image. Green Cone Mono should be expected to have a response similar to Protanopia but virtually no blue. Readability and Accommodating CVD User NeedsUser needs follows two broad categories
CLICK TO EXPAND USER NEEDSReading All vision types need very good luminance contrast because the high spatial frequency of text, especially body text, has a much lower threshold of contrast sensitivity. Color contrasts, as in hue or chroma, do not play into readability, except if they interfere with luminance contrast or cars chromatic aberration. As such, the content author need only to make sure that all sighted users have ample luminance contrast for readability, based on the importance of the use-case of the particular text. Discerning For this reason, the single most important thing to do is to ensure that there is something other than pure color communicating the information. Color, as in Hue and colorfulness, should never be the sole means of communicating information. This is important for all forms of vision, and all visual users. for instance cultural differences may result in different color interpretations, so rely on non-color dependent indicators to accommodate all visual users. Accommodating color-coding and non-text largely requires that all color-coded information is discernible and understandable without the color (hue) being present, in other words, make it a functional design in black-and-white first. sRGB Low Fidelity AdvantageThe red, green, and blue color primaries in sRGB Color space are close enough together and create a small enough gamut that even a severe form of CVD such ias protanpia will still see the color red, but about 50% darker. For UHD or rec2020, the high fidelity spectrally-pure red primary can disappear for protanopia. sRGB is a better color space for accessibility, and it is the web default standard. Accommodating readingThe content author need only to make sure that all sighted users have ample luminance contrast for readability. Accommodating differentiationAll color-coded information must be understandable without the color being present. Color as in hue is a useful tool to enhance information, but not to be the sole source of information.DISCUSSION OF THE VARIOUS GITHUB CVD COLOR PALETTESCLICK TO EXPAND DISCUSSIONI actually wanted to talk about this last year or two and unfortunately have been too busy, but I thought I would just take a break from my other work and add some discussion.Daltonization is the process of changing a color palette of an image or other content to improve the user experience for somebody with a color vision deficiency. This could be very useful on an image to give it more apparent depth, and it can be used on a user interface in some cases. But it's not always necessary, the first question to ask is what makes it necessary? If the normal color scheme has color conflicts in it where are certain meta-states can easily be confused when viewed by different forms of CVD, that would be one useful example—but it's less common than you might think. As for readability, good luminance for standard vision is also good luminance for all forms of CVD with the exception of red and black, in the case of a protan. But accessibilityred and black is bad for all forms of vision as far as readability is concerned. This then brings up the question, was there an accessibility problem related to CVD with the standard default color scheme for GitHub? Let's take a peek at the color palettes. Here are examples of the GitHub Color schemes. The left most slice for lightmode or darkmode is the normal default color scheme, and then towards the right we have some of the other color schemes, for different contrasts or for different kinds of CVD. These samples are processed by the Myndex CVD simulator, top left is standard vision then clockwise: deuteranopia, protanopia, tritanopia. Large buttonsLight mode: default, high, deu/pro, tri. DarkMode: default, high, deu/pro, tri, dim click to enlarge Looking at these critically, starting with the large buttons. First, for standard vision, examine the first slice on the left of either light mode or dark mode, and look at how it was processed for the three CVD types. Are there color conflicts that didn't exist for standard vision? Is there anything that became difficult to read? And then look at the slices that are processed for different contrast or CVD types. Is there any real improvement? Or do they simply look the same? DiscussionCLICK TO EXPAND DISCUSSIONIf the human vision system is going to interpret colors a certain way due to CVD, that does not mean that it's incumbent on the designer to design with those interpreted colors, it's only necessary to be aware of them to insure that there are no conflicts. I may be missing something and I don't have the whole picture, but it appears to me here that the colors for the various CVD types are mostly the colors that would be predicted or close to them, if the normal colors were seen by CVD. This in itself does not help CVD individuals—what helps CvD individuals is increasing the distance between colors. And that was done on one button, the new issue button was made blue. But why is it not blue for regular vision? Shouldn't regular vision have that distinction other than it's rectangular shape? Having said those things, the deutan/protan/tritan versions aren't necessarily worse than the normal version, except the code highlighting deutan/protan is better in the normal scheme then the deutan/protan. Remember that CVD has equivalent or better visual function (including luminance contrast) as standard vision. The only issue for CVD (other things being equal) is discrimination of hue, and then red against a very dark color in the case of the protan. The bigger concern from a UX point of you is the large variation of color for the same button across the themes. For the default scheme, "open" is green for another scheme "open" is red for another scheme it's orange... though all of the close ovals are lavender… To me this is confusion without providing a real or articulable benefit. But then I don't understand or know what the motivation was behind this, and so I may be missing something and I grant you that. But that's also why I'm starting a conversation, for a better understanding of the goals. Next, the very thin icons Thin iconsLight mode: default, tri, deu/pro. DarkMode: default, tri, deu/pro click to enlarge DiscussionCLICK TO EXPAND DISCUSSIONSo, the same comments from above apply here as well, not knowing why Colors were changed so substantially when they don't seem to need to be? There is nevertheless an accessibility issue here, in that all of these icons are composed of thin lines. That makes them a relatively high spatial frequency compared to the solid buttons on the other example. The thing is that both color and luminance contrast are very sensitive to spatial frequency. And in the case of color (hue) our ability to distinguish between hue or chroma is sharply curtailed as the spatial frequency goes up, meaning as lines get thinner we rapidly lose our ability to distinguish between hues. Notice also that these thin lines against the white background lose much of their colorfulness vs the ones against the black background. And finally take a look at the tritanopia processed example you'll notice that all of the colorfulness is gone from all versions of the checkmark next to the word Tony and the switch icon next to the word test at the bottom than the other schemes. Accessible or NotIn closing, while I don't see a serious problem with the color schemes, I also don't see them as being particularly useful given the context and the very limited use of color. But I am very glad to see the exploration of visual accessible issues here, and I hope you take my comments in the spirit they were intended as constructive and helpful. I want to be clear that I don't see these specific example colors as being "not accessible", however I have to admit I don't consider Github particularly accessible overall. And here's why:
Thank you for reading, Andy Andrew Somers |
Beta Was this translation helpful? Give feedback.
-
|
Very good points all-around. Thank you for contributing your knowledge here. I'm afraid GitHub may not be particularly interested at this point in time, but I really hope they read what you've sent. |
Beta Was this translation helpful? Give feedback.
-
|
Myndex wrote:
Andy, thank you for actually including the longer blocks of information; however, my comprehension is noticeably slowed by various mistakes, and I am not fond of automatically replacing unexpected lexemes with whatever pops first off the heap of autocorruption. I do not discriminate between various reasons for semantic failure, nor do I judge harshly if the response is simply that the author did not have sufficient time to painstakingly correct each sentence, word, letter, and punctuation mark -- yes, that's four separate rounds of editing! -- however, I do frequently wish I could send pull requests for such complaints, instead of, well, complaining. Would it help for me to list in comments such errors, and then either delete the complaint comment after the author -- in this case, you, although this question is more generally applicable -- has updated the original comment, or leave the complaint in the cases where the response is clarification of my own failure to comprehend? |
Beta Was this translation helpful? Give feedback.
-
|
Hi @adlai
Sorry, I'm not quite sure what you're trying to say here other than I think you're saying that I might have some spelling or other errors in some of what I write? Just so you were aware, I am low vision and I rely on speech to text for most of what I write. It is very difficult for me to actually see the screen, and as a result I do frequently miss errors that others might take for granted. It is a function of the fact that I am disabled with low vision, and this has a lot to do with why I am currently focused on my current research. I always welcome corrections, especially the kind that I miss, despite my going over multiple times and despite using spellcheckers when possible, which actually is a big problem with GitHub because GitHub does not allow the use of spellchecking when you're using their forms here... (!!!) Come to think of it, this has caused me so many problems in the past, I really should raise an issue with GitHub. It's their JavaScript in some of the user interface is here that interferes with the spellchecker in browsers such a Safari. |
Beta Was this translation helpful? Give feedback.
-
|
Myndex wrote:
That is the answer to the main question; some folks get offended when people complain about typos. Years ago, I was quite interested in the kinds of information you could learn from somebody's typographical error profile -- e.g., whether text is typed letter-by-letter, or dictation -- although my curiosity about this has fatigued as the world has gone mobile and now the supermajority of text is from QWERTY layouts muddled by touchscreens. I have no silver bullet for exterminating typos; dreaming about spell checkers smart enough to catch not only single-letter errors, but also homophones and various other cases of grammar, is like claiming that ChatGPT can be both copywriter and editor. I revisited your longer post and made some notes. I ignored capitalisation issues, for brevity.
In closing, I must say that I found quite helpful your comments about spatial frequency, and the importance of distance-across-color-space between the semantically important points of the ultimately chosen palette. I don't have the necessary qualifications to preach best practices, nor the academic clout to drive research, although such descriptions do give me hope that eventually mathematical rigor could make the topic less of a niche interest, and part of what is included in statements about how literacy is evolving to include computer programming. |
Beta Was this translation helpful? Give feedback.
-
|
Hey guys,
The reason for this, as I mentioned earlier, is that magenta is not a "real" color: it always has to be composed of equal parts red and blue. People who can see one but not the other will still see something that is different from green, because red and blue are both maximally distant from green. |
Beta Was this translation helpful? Give feedback.
-
So, in your proposal, stop becomes blue for protanopia? I consider this a bad idea on a number of fronts. Above in your examples this means the warning color red becomes blue, and the non-warning green becomes yellow. This is the opposite of what we would want for a traffic signal. Also, traffic signals and stoplights are not trichromatic Displays. The "red" in traffic lights & vehicle brake lights is nominally a red/orange. NOTE: And so is the red in a sRGB display, for protanopia, the sRGB red/orange primary is reduced in luminance by 55%, but is at maximum saturation "yellow-ish". The green in traffic signals is closer to a turquoise, and it is out-of-gamut for an sRGB monitor. The salient points is that stop light red stimulates L & M cones, traffic signal green mostly stimulates M & S. One way to look at this is it keeps the warning color the most saturated for dichromats. As I previously said, there is much more nuance and complication to human vision, which is not served well by a brute-force "maximum discrimination", which is not generally needed and can have unintended consequences. Here's an article on the subject by somebody with color insensitive vision: https://dleppik.wordpress.com/2013/07/23/are-green-traffic-lights-really-turquoise/ You may find this interesting regarding language and color: |
Beta Was this translation helpful? Give feedback.
-
Thank you for taking time to engage in this educational discussion, including the images and links to simulators. I personally do not currently suffer from any such major impairments, nor am I an optician, neurologist, or quantum optics worker, so these posts are significantly more helpful than links to LMGTFY and quips about quickly skimming Wikipedia. However, I am slightly ticked off by such a strong statement as the quoted one. That kind of statement leads to the design philosophy where tweaking the palette will magically make chromatic presentations legible, which I consider simply not the most helpful approach when your highest purpose is for maximal human accessibility of the information presented by, and functionality implemented in, a webpage. I much preferred your earlier statement encouraging reduction of reliance on color differences for conveying any semantically important information. I think part of the problem with one-scheme-fits-all fixes to color impairments is that vision in general, and color vision specifically, is quantization of continuous signals, thus hopeless from the get go; yet such conversation is truly waffling and way off topic. |
Beta Was this translation helpful? Give feedback.
-
|
@adlai Aye, all fair points. I guess I've become a bit enamoured of the green/magenta thing as a quick fix (again, not as a panacæa) just because you can add a little blue to the red in your typical green/red distinction and instantly get a huge increase in accessiblity. I of course don't hold that it's a perfect end-all solution, but it can get you pretty far for very little investment. I see it as a really good starting place, and then building out from there, essentially. Most accessibility improvements take a lot more time and effort than a simple palette swap. But switching red to magenta is something you can do at the snap of your fingers, and it fixes most of the main issues for most people instantly, at least as far as I'm aware. Sort of your quick 70% solution while you work on the trickier 95% one. The ideal end-situation is of course to find ways other than just color to convey information. I apologize for not expressing things as well as I should have in the admittedly cheeky statement you quoted above. I just wanted to remind people that there's a quick "hack" that'll get you a lot of gain for very little effort, but again I don't intend for that palette swap to be the only thing done in the end. |
Beta Was this translation helpful? Give feedback.
-
|
The quick hack is wonderful while the range of communications media under consideration always ends at people reading off liquid-crystal displays. Most discussions of palette issues become nightmares once the question of legibility after printing is considered, although at that point the conversation is arguably no longer about accessibility. It would be incredible if all websites presenting informational graphics and syntax highlighting adhered to the standard occasionally seen in research publications, where the data values themselves can be obtained separately from some watermarked document. |
Beta Was this translation helpful? Give feedback.
-
|
Indeed. Worth noting, though, that the green/magenta hack works off of displays, too. Even when not working in sRGB, magenta is -- fundamentally -- a color that reflects equal parts red and blue light; there is no "magenta" wavelength of light. So even in print, this hack should still work. |
Beta Was this translation helpful? Give feedback.
-
|
__Hi @adlai thank you very much for commenting, it reminded me that I was going to demonstrate to @MilesBHuff why "green and magenta" is no only not the answer, it creates severe problems and is in fact very far from the answer. Human vision is far too nuanced for something as simplistic as a conjecture such as "just make it green and magenta". Green and magenta is a terrible terrible combination. There are examples of green and magenta, processed using a clinically accurate simulation of color vision deficiency. For each slide in the upper left-hand corner is standard vision and then several forms of CVD. All the slides are green and magenta, at various levels of brightness, and showing that the assertion that these are all equally readable for all vision types is quite simply not the case. #fg vs #bg#090 #f0f
#0c0 #f0f
#f0f #0c0
#f0f #090
#c0c #090
#c0c #0a0
#090 #c0c
#0a0 #c0c
These should be enough to demonstrate that this color combination is pretty much bad for everybody all the time, but I provided some examples of when it's best for one it's also worst for another. While you can see the giant ampersand, that's only because it's huge and therefore a very low spatial frequency, the remainder of the text is pretty much unreadable. In other words, this combination doesn't even work, much less be the panacea for all people.Human vision is far too nuanced to try and say "it needs to just be this and then everything is fine" because that's simply, absolutely, not true. Thank you for reading |
Beta Was this translation helpful? Give feedback.
-
Oh yeah, no, agree there. Green and magenta placed on-top of each other is not pleasant. But used as colors for contrasting UI elements, they work well. |
Beta Was this translation helpful? Give feedback.
-
|
MilesBHuff wrote:
Honestly, right here is where I begin wondering what disability to plead! Impatience, stubbornness, and redundancy are not disabilities, so I guess I remain simply obnoxious; yet the difference that you illustrate with the example images is not trivial for folks hampered by the two-dimensional visual projection to immediately notice. My recommendation, in the interest of reducing future requests for clarification, is that people open the two images in separate tabs, adjacent within one browser window, and then use some keyboard shortcut for alternating between the tabs. The difference that is being pointed out is one that is not made more obvious by spreading examples across larger and larger screen space, and possibly even more difficult to notice that way. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @adlai
I agree, and it's a hideous combination in part because of the significant artifacts it creates including chromatic aberration and chromostereopsis.
No, it is very much the responsibility of the content author.
Miles' examples are not noticeable across large screen space because they're not useful and they're not substantial. Blue against black is pretty much unreadable for everybody. And since this discussion has now moved over to readability, it's critical to understand that color as an hue is not relevant for readability, only luminance is. Pure blue on Black is bad for all vision types. So is pure red on black, and pure red on black for protanopia is the one hue that is or should be prohibited. Adding blue to it is not helpful for readability. Adding green for the purpose of increasing luminance is the one thing you can do to any color that is against a dark color, this is because "green" makes up the vast majority of luminance, over 70%, depending on the color model. The blue and the red/orange primary in an sRGB monitor are not related to the peak response of the S or L cones, but instead are located at their particular positions as a matter of economics relating to display technology, and in attempt to stimulate the S or L cone with some level of independence. Regardless, there is substantial crosstalk or overlap of all three cone types, particularly with the low fidelity primaries of an sRGB display. I cover this and other related understandings of how color vision works in "The Realities And Myths Of Contrast And Color |
Beta Was this translation helpful? Give feedback.
-
|
Myndex wrote:
The first word is probably a mistake; you probably meant "blue"? Further down, you've got "the one who that is", where the correct word is probably "hue". |
Beta Was this translation helpful? Give feedback.
-
|
@adlai wrote
Thank you, corrected. |
Beta Was this translation helpful? Give feedback.
-
(Originally posted here: https://github.community/t/the-new-colorblind-scheme-is-inaccessible-for-people-with-tritanopia.)
Orange/blue is not accessible for people with tritanopia. The only two-color scheme that works for people with all three types of single-cone-deficient color-blindnesses, is magenta/green. You can easily verify this with any decent colorblindness simulator.
The reason magenta/green works for all three, is that green is in the middle of the visual field, while magenta is a “fake” color (As in, there is no such thing as a “magenta” wavelength of light.) that is composed of equal parts red and blue, each which are at opposite edges of our visual spectrum. Someone with a deficient red cone (protanopia) will still see magenta as blue; someone with a deficient blue cone will still see magenta as red (tritanopia); and someone with a deficient green cone (deuteranopia) will still see the two as different hues, since green and red/blue are maximally distant.
As well, magenta/green is much easier for non-colorblind people to adjust to than orange/blue, as magenta/green is a strict superset of red/green (since magenta is by definition equal parts red and blue).
I figured this out a couple years ago by playing around with a color-blindness simulator until I found a color combination that worked for all 3 single-cone-deficient color-blindnesses. The explanation above as to why this is the case is original theory.
Beta Was this translation helpful? Give feedback.
All reactions