Color Meaning in East Asian Cultures: A Design Guide

By IWantFreeColorTools.com · 27 min read

Color meaning in East Asian cultures is real, but it's much narrower than most design guides suggest. The strongest research finds that colour-emotion associations are largely shared across humanity, with culture adjusting them at the margins rather than reversing them. The associations that genuinely matter are tied to specific contexts, like a wedding invitation or a gift, not to every element in an interface.

Which makes most "red means luck, white means death" advice a poor guide to actual design decisions. Here's what the evidence supports, and the one linguistic difference that should genuinely change how you name and pick colors.

How different is color meaning across cultures, really?

Less different than the design industry assumes, and there's a large study that settles a lot of this.

Jonauskaite and colleagues surveyed 4,598 participants across 30 nations speaking 22 native languages, asking them to associate 20 emotion concepts with 12 colour terms. The results were published in Psychological Science in 2020.

The headline finding is that colour-emotion associations were substantially universal, with an average pattern-similarity coefficient of r = .88. Participants in all 30 nations agreed that black and red are the most emotional colours, and that brown is the least. Men and women produced almost identical pairings, as did different age groups.

But the study didn't find pure universality either. A machine-learning analysis showed that nation predicted colour-emotion associations beyond the shared baseline, and similarity was greater between nations that were linguistically or geographically close. So culture matters. It just modulates a strong common signal rather than replacing it.

A second study is worth adding here, because it tested a specific association in both regions at once rather than surveying broadly. Lei Wu, Chenxi Lu, Wijnand van Dijk, Hong Li and Simone Schnall ran seven experiments with 357 university students in the UK and China, published as The Color Red Is Implicitly Associated With Social Status in the United Kingdom and China in Frontiers in Psychology (2018, vol. 9). They used implicit association tests pairing red against high-status logos, luxury car marques in the British samples and Chinese car and university brands in the Chinese ones.

Red tracked with high status in both places. The pooled effect across all seven experiments was medium, at d = 0.54, running as high as d = 0.92 and d = 0.99 in individual Chinese experiments. The finding that matters for this article is the comparison they ran between the two samples: a mixed ANOVA suggested the strength of the red and status association was equal among the British and Chinese participants.

So here's a colour association loaded with centuries of specific imperial and ceremonial history in China, and it turns up at the same strength in a country with none of that history. That's the pattern this whole page keeps running into. The shared signal is doing most of the work, and the cultural layer sits on top of it rather than replacing it.

The older marketing literature points the same way. Madden, Hewett and Roth ran an eight-country study of colour meanings and preferences, published in the Journal of International Marketing in 2000, and found a mix of cross-cultural similarity and difference, with blue associated with high quality fairly consistently across the countries studied.

How settled is this research, and does it shift over time?

Settled enough that you can plan around it. One large study is one study, so the fair question is whether the whole field points the same way. It does, and somebody has now checked properly.

Jonauskaite and Mohr at the University of Lausanne published a systematic review in Psychonomic Bulletin and Review on 13 January 2025. Following PRISMA 2020 guidelines, they pulled together 132 peer-reviewed empirical studies covering 42,266 participants across 64 countries, published between 1895 and 2022. That's 128 years of research on whether colours carry emotional meaning.

Their conclusion: colour-emotion correspondences are systematic and, in their words, highly congruent across cultures and populations. The correspondences held across countries, across age groups, across historical time periods, and across different ways of asking the question.

Some of the specific pairs they list will look familiar: red with anger, love and passion, yellow and orange with happiness, blue and green with calm and sadness, black with sadness and fear. Light colours skewed positive, dark colours negative, more or less everywhere.

Two limits on what that proves, and both matter for this page.

The review is about colour and emotion, which is not the same thing as colour and social convention. It tells you that red feels activating almost everywhere. It says nothing about whether handing someone a white envelope at a wedding is a mistake. Those conventions are real, learned, and local, and no amount of emotion research makes them go away. Keep the two categories separate and most of the confusion in this topic disappears.

The authors also flag that some of the many-to-many pairings they found could have been driven by cultural differences they couldn't isolate, and that the causal question, whether colours actually change how you feel rather than what you associate, is still open. Which is a more honest position than most design writing on colour psychology takes.

Which cultural color associations actually hold up?

Some do, and it's worth separating them from the ones that get repeated because they sound good.

Red at celebrations in Chinese culture. This is not folklore. Red envelopes at Lunar New Year and weddings are an active, everyday practice across Chinese communities including Singapore and Malaysia. If you are designing a Lunar New Year campaign, red is the expected register, and using muted greys instead will read as tone-deaf.

White and mourning. Also real. White is worn at funerals across much of the region, and giving white flowers as a gift can land badly. This has genuine consequences for gifting, packaging of ceremonial products, and event design.

Gold and status. Gold reads as premium and auspicious in Chinese contexts, which is why you see it heavily in festive packaging and luxury positioning.

Now the honest caveat, and it's the important part of this section. These associations are context-bound. They activate around ceremony, gifts, festivals and life events. They do not follow the colour into every setting. A Chinese user does not read a white app background as funereal, because white backgrounds are the norm on the sites and apps they use all day.

You'll also find widely repeated claims about the five elements mapping colours to directions and seasons. That framework is genuine as a historical and philosophical system. But most of what circulates in design articles about it traces to other design articles rather than to scholarship, and it's rarely load-bearing for a real design decision. Treat it as background, not a rule.

If you do want the actual scholarship rather than the design-blog version of it, it exists and it's worth knowing where. Xin Zhang, Muhammad Faizal Abdul Rani and Yizhe Zhang examined how wuxing and yin-yang colour theory actually showed up in built work in the Journal of Chinese Architecture and Urbanism (2026), working through Ming and Qing dynasty architecture and Forbidden City furniture. Their conclusion is a useful corrective: colour choices there came out of philosophical doctrine, what materials were actually available, and sociopolitical function all at once, not from a clean lookup table of colour to element.

Which is rather the point. Read the real thing and the framework turns out to be messier and more contingent than the tidy five-way diagrams that circulate online, and a lot less like something you can apply to a landing page. Interesting history. Still not a design rule.

Which colors do people in the region actually prefer?

Worth separating two questions that get mashed together. What a colour means is one thing. Which colours people actually like is another, and for palette work the second one is more useful.

Saito studied exactly this, publishing in Perceptual and Motor Skills in 1996 with 490 subjects: 175 Japanese, 158 Chinese and 157 Indonesian. The finding that matters most here is blunt. A high preference for white was common to all three countries.

Sit with that next to the "never use white in Asia" advice. The people that advice is about like white. Differences did show up, and they were statistically significant, but they appeared in the frequency of specific hues and tones rather than in broad strokes. Saito's conclusion was that environmental and cultural associations shape preference at that finer level.

So the useful mental model isn't "different cultures, different colours." It's "same broad preferences, different fine tuning."

And here's the study that shows where those finer associations come from, which is more interesting than any symbolism list.

Zhang and Han published "Experience Reverses the Red Effect among Chinese Stockbrokers" in PLoS One in 2014. There's a well-known finding in psychology that seeing red before a test hurts performance. They reproduced it in Chinese college students, who scored 22.17 in the red condition against 25.25 in green.

Then they ran the same thing on Chinese stockbrokers, and it flipped. Red 22.50, green 18.92. Red helped.

The explanation is the whole point. As the paper puts it, in the Chinese stock market red stands for a rise in price and green for a decrease. Brokers spend their working lives watching red mean money arriving. That exposure rewired a robust psychological effect in the opposite direction.

What a designer should take from this: colour associations are learned from what people are repeatedly exposed to, not inherited from a culture's history. Which is exactly why looking at the products a market actually uses beats reading about what that market supposedly believes.

Do men and women in East Asia prefer different colors?

A bit, but not in the tidy pink-and-blue way packaging assumes, and the best evidence from the region says the pattern is messier than the famous studies suggest.

The study everyone quotes is Hurlbert and Ling's "Biological components of sex differences in color preference", published in Current Biology in 2007. It tested 208 young adults aged 20 to 26, British and Han Chinese, and found women leaned toward reddish-purple hues while men peaked around blue-green. The part that made headlines was that the gap between women and men looked much the same in both groups, which the authors read as a possible biological component.

But it's one study of people in their early twenties, and a short research letter at that. It can't tell you much about a 50-year-old shopper in Taipei, or whether the pattern holds in Seoul or Osaka. For a finding that gets repeated as settled biology, it's carrying a lot of weight.

The later research pushes back. Yokosawa, Schloss, Asano and Palmer compared Japanese and American color preferences in Cognitive Science in 2016. The two groups shared a lot, including a peak around blue, a dip around dark yellow and a liking for saturated colors, but Japanese participants liked darker colors less than Americans did. On gender, they found "complex gender differences" that "did not conform to previously reported effects." In other words, the neat female-reddish, male-bluish split didn't show up cleanly in Japan.

Their second experiment is the useful one for designers. They tested bilingual Japanese-American participants and found their preferences sat in between the Japanese and US results. That fits the idea that preferences are shaped by what you've lived with, the same lesson the stockbroker study above teaches with red and green.

And a study further afield makes the same point more bluntly. Taylor, Clifford and Franklin compared British adults with Himba adults in rural Namibia, a community relatively untouched by global consumer culture, in the Journal of Experimental Psychology: General in 2013. The Himba showed none of the so-called universal patterns or sex differences at all. Their title says it: color preferences are not universal.

What should that change in your designs?

• Don't gender-code by default. A pink version for women and a blue one for men is a marketing convention, not a finding, and the evidence from Japan specifically doesn't support it as a rule.
• Check the category, not the gender. Look at what's already selling in that product category in that market. Local shelves reflect what people have actually been exposed to, which is what shapes preference.
• Watch lightness as much as hue. The clearest Japan-US difference Yokosawa's team found was about dark colors, not about red or blue. If your palette leans heavily dark, that's worth testing in Japan before launch.
• Test with real users. The research section below walks through how, and our brand color palette guide covers building a palette flexible enough to adjust once the results come in.

Does language change which colors people see as separate?

Yes, and this is the finding that should actually change your work.

Colour vocabulary differs between languages in ways that affect how people group shades. Kuriki and colleagues at Tohoku University tested this properly in modern Japanese, publishing The Modern Japanese Color Lexicon in the peer-reviewed Journal of Vision in 2017. They showed 52 native Japanese speakers a set of 320 colour samples plus achromatic ones and asked for single-word names.

They found 19 commonly used Japanese colour categories. Eleven correspond to the familiar basic terms: red, green, blue, yellow, purple, pink, brown, orange, white, grey and black. But eight more appeared as distinct categories in their own right, including mizu (water, meaning light blue), hada (skin tone), matcha (yellow-green), enji (maroon) and yamabuki (gold).

The number worth remembering: Tohoku University reported that mizu was used by 98 percent of informants. An earlier study by Uchikawa and Boynton in 1987 did not identify mizu as a basic category, so this looks like a vocabulary that has shifted within living memory.

Two design consequences follow directly:

• Light blue is not "a shade of blue" to a Japanese speaker. It's closer to its own colour. If your palette uses light blue and mid blue as two steps of one ramp, users may read them as two different colours rather than a tonal pair.
• Colour naming needs localising, not translating. Calling a swatch "light blue" in a Japanese interface loses information that "mizu" carries. Our colour naming guide covers why names matter more than designers expect.

One related point that gets misreported constantly. Japanese traffic lights are often described as proof that Japanese speakers cannot distinguish blue from green. That's wrong. The Kuriki study found modern Japanese is not a "grue" language, since blue and green are clearly separate categories. What's happening with traffic signals is a historical naming convention, because ao once covered a broader range and the old term stuck to the signal. The words lagged the categories.

Do Korean and Chinese speakers split colors the same way?

No, and the two languages fail differently, which is more useful than it sounds. Japanese gets all the attention here because of mizu, but it isn't the only lexicon in the region that cuts the spectrum somewhere English doesn't.

Korean splits green. Where an English speaker has one word, Korean has two obligatory ones: yeondu for yellow-green and chorok for green. There's no single Korean term that covers both, so a Korean speaker naming a green ramp has to pick a side at some point in it.

Debi Roberson, Hyensou Pak and J. Richard Hanley tested whether that changes anything perceptual, publishing Categorical perception of colour in the left and right visual field is verbally mediated: evidence from Korean in Cognition in 2008 (vol. 107, pp. 752 to 762). They ran a visual search task on English and Korean speakers using colours straddling the yeondu and chorok boundary.

Korean speakers discriminated faster across that boundary than within either category. English speakers showed nothing at the same boundary, because for them it's all green. The effect also showed up as lateralised, appearing in the right visual field for fast responders and in both fields for slower ones, and it weakened under verbal interference. Which is the part that names the mechanism: the boundary is doing its work through the language, not underneath it.

That last detail matters more than the headline, and the same team went and proved it the hard way.

Roberson, Hanley and Pak followed up with Thresholds for color discrimination in English and Korean speakers, also in Cognition (2009, vol. 112, issue 3, pp. 482 to 487). This time they measured just-noticeable differences rather than search speed. And they found no difference in discrimination thresholds between the two language groups, despite the groups having different category boundaries in exactly that region.

Read those two papers together and you get the honest version: Korean speakers aren't seeing more greens than you are. Their eyes resolve the same differences yours do. What's different is what happens after, when a name gets attached and the two shades land in separate mental bins. It's a labelling effect sitting on top of identical sensory hardware, which is precisely the shape of every other finding on this page.

Chinese breaks in the opposite direction, and this one is genuinely awkward for interface work.

Vincent C. Sun and Chien-Chung Chen published Basic color categories and Mandarin Chinese color terms in PLOS One in 2018. They showed 59 native Mandarin speakers in Taiwan 280 Natural Color System samples under D65 illumination and asked them to match 32 single-character colour terms to samples, then ran principal component analysis over the results.

The categories themselves came out looking normal. As the authors put it, the basic colour categories among Mandarin speakers are similar to those found in the World Color Survey. No exotic carve-ups, no missing hues. The problem is the layer above.

Those categories are, in the paper's words, represented by wide-spread and inconsistent colour terms among speakers. Terms like Mo (ink), Tie (iron) and Cai (vegetable) scattered badly across samples, with different people pointing them at different things. White, blue and purple didn't even load as major factors in the analysis, not because Mandarin speakers lack those categories but because the synonyms split the vote between them.

So the two languages create opposite problems for you:

• Korean adds a boundary your ramp doesn't know about. A five-step green scale that reads as one family in English may cross a hard naming line partway down for a Korean user. If those steps are supposed to signal "same thing, more of it," check where yeondu becomes chorok before you assume the ramp reads as continuous.
• Mandarin makes the colour name a weak identifier. Two Chinese-speaking users can use different words for the same swatch, or the same word for different swatches, more readily than two English speakers will. So a filter labelled only with a colour word, a support doc that says "click the zong button," or a chart legend leaning on names rather than swatches is doing less work than you think.
• Show the colour, don't only name it. This is the fix for both at once. A swatch next to the label costs nothing and survives every version of this problem, including the Japanese one from the section above.

Worth flagging the sample bounds honestly, since this page has been strict about that elsewhere. The Mandarin study ran in Taiwan, and term usage does vary between Taiwan and the mainland. The Korean work used university-age participants in a lab task. Neither is a national survey. But both point the same way the rest of the evidence does: the perception is shared, the vocabulary isn't, and the vocabulary is the part that reaches your interface. Our colour naming guide goes further into why naming carries more weight than most teams give it.

Which design decisions does color meaning genuinely affect?

Four situations, and they're narrower than the topic suggests.

1. Anything tied to a festival or life event. Lunar New Year, weddings, funerals, gift packaging. Here the conventions are strong, current, and worth respecting.
2. Physical products and packaging. A box sits in someone's home and gets given to other people, so it carries social meaning an app screen does not.
3. Colour naming in the interface. As above, this is where language differences bite hardest and where a straight translation loses meaning.
4. Charts where colour carries meaning. Red and green for gain and loss is the honest example. In Western markets red usually means a loss. In mainland China and Taiwan, stock charts conventionally use red for rising and green for falling, which is the reverse. Zhang and Han state it plainly in their PLoS One paper: in the Chinese stock market, red is a rise and green is a decrease. Getting this backwards on a finance product is a real, concrete error.

That last one is the single most useful item on this page for most designers, because it's specific, verifiable by looking at any Chinese brokerage app, and it genuinely inverts.

When does it not matter as much as you think?

Most of the rest of your interface.

• Backgrounds. White and near-white are the global default and read as neutral everywhere.
• Error and warning states. Red for error and amber for warning are conventions people learn from software, not from culture. The Jonauskaite data supports this, since black and red rated as the most emotional colours in all 30 nations.
• Brand colour on its own. Plenty of successful brands in the region use blue, green or black without cultural friction.
• Neutral greys. Nobody is reading meaning into your border colours.

There's a harder version of this point, and it comes from the one place you'd most expect cultural colour to pay off: getting someone in China to actually buy something. Zhao and Guan ran three experiments for Frontiers in Psychology (2023, vol. 13, article 1028425), an implicit association test plus bank and restaurant scenarios, with 268, 188 and 299 participants analysed. Their title says it: Chinese consumers do not always respond to red.

What moved purchase intention was blue, and not for any reason a symbolism list would predict. Blue walls made the space read as more spacious than red ones, and that perceived spaciousness was what carried the effect through. The mechanism is spatial perception, not cultural meaning. So even in the commercial setting where "red is lucky in China" should be doing its most obvious work, the colour that won did so through something with no cultural content in it at all.

Accessibility, by contrast, matters identically everywhere. Contrast requirements do not change by market, and a palette that fails WCAG contrast in one country fails in all of them. If you have limited time to spend on your palette, spend it there rather than on symbolism.

Which specific mistakes actually cause problems?

A short list, because the honest answer is that there aren't many. The ones that exist tend to be linguistic or ceremonial rather than perceptual, and they show up on physical things more than on screens.

Green headwear on anything aimed at Chinese markets. This is the one that catches people out, and it isn't really about colour at all. In Mandarin, the phrase for wearing a green hat, dai lu maozi, is a common idiom meaning a man whose partner has been unfaithful. The World of Chinese traces competing accounts of where it came from, including Tang dynasty rules putting low-ranking officials in green and a Yuan dynasty edict on green headwear, with the modern insulting sense settled by the Ming period. The origin is genuinely disputed. The current usage is not.

Practically, that means a green cap on a mascot, a green hat in a product photo, or green headwear as promotional merchandise is a real problem, and it's the kind of thing that gets screenshotted. Note the narrowness though: green itself is fine, green packaging is fine, a green logo is fine. It's specifically the hat.

Flagging the evidence quality, since this page has been fussy about that throughout: this is a documented linguistic convention rather than an experimental finding. There's no controlled study measuring it. But it's the sort of claim any Mandarin speaker will confirm in about four seconds, which is a different kind of reliable from a peer-reviewed effect size.

White in gift and ceremonial contexts. Covered above, and it stays on this list because it's concrete. White flowers as a gift, white ceremonial packaging, white in anything wedding-adjacent. Not white interfaces, not white space.

Red and green inverted on financial charts. Also covered above and still the single most consequential item for most product teams, because it silently reverses the meaning of your data rather than just looking odd.

And that's roughly it. Notice what isn't here: no rule about avoiding particular hues in branding, no palette that offends, no colour that means the opposite. The failures are specific, checkable, and mostly avoidable by showing the work to one native speaker before it ships, which is the same advice the research section lands on for less interesting reasons.

Does color vision deficiency change the picture?

Yes, and this is the one place in the whole topic where a genuine population-level difference across these markets is measurable. It's also the place designers are most likely to draw the wrong conclusion from it.

The best data comes from the Multi-Ethnic Pediatric Eye Disease Study. Xie, Tarczy-Hornoch, Lin, Cotter, Torres and Varma tested around 4,005 preschool children aged 3 to 6 in Southern California and published in Ophthalmology (2014, vol. 121, pages 1469 to 1474). Because it screened four groups with the same protocol in the same place, it sidesteps the usual problem of comparing studies that used different tests in different decades.

Prevalence in boys came out like this:

In girls it ran between 0 and 0.5 percent across every group, low enough that the researchers said they couldn't meaningfully compare the four. That lopsidedness isn't cultural. The common red-green forms are X-linked, which is why this is overwhelmingly a male pattern everywhere.

So the raw fact is real: colour vision deficiency is roughly half as common among Asian boys as among non-Hispanic white boys in that sample. Now the tempting and wrong inference.

You cannot conclude that red and green are safer to rely on in East Asian markets. Three percent of men is about one in 32. On a consumer product with a hundred thousand male users, that's a few thousand people who can't reliably separate your two most meaningful colours. A smaller number is still a large number, and none of those users care about the comparison with another country.

And here's where it collides with the most concrete finding on this page. The section above notes that stock charts in mainland China and Taiwan use red for a rise and green for a fall, inverting the Western convention. Red-green deficiency is exactly the deficiency that makes those two hues hard to tell apart. So the one context in this entire article where colour genuinely carries meaning rather than mood is also the context where hue alone is least safe to depend on.

It gets slightly worse for anyone building across both markets. A user who has learned to cope with a red-green chart in one convention has that workaround inverted in the other. Whatever mental shortcut they built now produces the opposite answer.

The fix is the same one that applies everywhere, which is rather the theme of this guide:

• Never let hue be the only carrier. Pair it with a sign, an arrow, a position, or a label. A number that reads +2.4% with an up arrow works regardless of which colour convention it's painted in and regardless of who's looking at it.
• Test the palette in greyscale. If two states become indistinguishable when you drop the colour, they're relying on hue alone.
• Keep the contrast bar where it already is. The requirements don't shift by market, and our WCAG colour accessibility guide covers the thresholds.

Which lands on the same conclusion as the rest of this page, from a different direction. The differences across these markets that you can actually design against are not differences in what colours mean. They're a chart convention and a population statistic, and both of them point at the same discipline: encode the meaning somewhere other than the hue, and the cultural question stops being load-bearing.

Does the script itself change your contrast decisions?

Yes, and this one is hiding in the accessibility spec you already follow rather than in anything about symbolism. If you're setting Chinese, Japanese or Korean text, the contrast thresholds you learned for Latin type don't transfer cleanly, and the standard says so itself.

Start with what WCAG actually requires. Normal text needs a contrast ratio of at least 4.5:1, and large-scale text drops to 3:1. Nothing surprising there. The interesting part is how large-scale gets defined.

Per the W3C's Understanding Success Criterion 1.4.3, large-scale text means "at least 18 point or 14 point bold or font size that would yield equivalent size for Chinese, Japanese and Korean (CJK) fonts." So the spec explicitly declines to give you a number for CJK.

It explains why, too. The 18pt and 14pt figures come from print conventions for roman text: 14pt is the minimum large-print size and 18pt the next standard size up. For CJK, W3C says the equivalent would be "the minimum large print size used for those languages and the next larger standard large print size." Different typographic tradition, different numbers, and the spec leaves you to find them rather than pretending 18pt is universal.

That matters because the same document gives the reason the 3:1 relaxation exists at all: text that is larger and has wider character strokes is easier to read at lower contrast. Turn that around and the CJK problem becomes obvious. A character built from 20 strokes in the same em box as a Latin letter built from two has much finer strokes and much tighter counters at identical point size. The thing the relaxation depends on is exactly the thing that's in shorter supply.

W3C makes the related point directly: fonts with extraordinarily thin strokes are harder to read, especially at lower contrast levels.

What the legibility research adds

There's measured work on this. Liu, Yu and Zhang ran a within-subjects experiment with 36 participants doing character-search tasks on pseudo-texts, published in Human Factors and Ergonomics in Manufacturing and Service Industries (2016, vol. 26, pp. 381 to 392). They varied font size, stroke width and character complexity, and measured search time and accuracy.

Font size and character complexity both had significant effects. Stroke width on its own did not, which is worth stating plainly rather than glossing. But every two-way interaction between the three was significant, and that's where the useful finding sits:

• For high-complexity characters, stroke width was the critical factor. Size alone didn't rescue them.
• For low-complexity characters, font size was the critical factor. The usual advice applies.
• Best overall in their setup was 12pt with a 1:10 stroke width ratio, which held across both complexity levels.

Read those two rows together and you get something a single global rule can't express. Your interface probably mixes both kinds of character in the same string, so whichever variable you fix, some of the text is being served by the wrong one.

What to actually do

• Don't reuse your Latin type scale. A size that reads comfortably in English is frequently too small for the same information in CJK. This is the single most common version of the mistake.
• Treat 3:1 as unavailable unless you've checked. The large-text exemption is defined by an equivalence you haven't established. Holding CJK body text to 4.5:1 costs you very little and removes the question.
• Be wary of light weights. Thin CJK faces lose stroke definition faster than thin Latin faces, and the loss compounds at lower contrast. If you want a light look, buy it with size rather than weight.
• Test with dense characters, not easy ones. Sample text full of simple glyphs will pass when your real content doesn't. Use the most stroke-heavy strings your product actually renders.
• Check it on a real device at real brightness. Fine strokes are exactly what a dim phone screen outdoors destroys first.

And notice where this lands relative to everything else on this page. It's not a claim about what colours mean to anyone. It's a measurable interaction between script, size and contrast, and it's the kind of difference across these markets that genuinely changes what you build. Our colour accessibility guide covers the ratios themselves in more detail.

Does dark mode change any of this?

Yes, and it lands hardest on exactly the thing the section above just established. If CJK characters already have finer strokes than Latin type at the same point size, then inverting the polarity attacks the detail that was already in short supply.

Start with the general finding, because it's the part most design teams have backwards. Research on display polarity has repeatedly found a positive polarity advantage, meaning dark text on a light background is more legible than light text on a dark one. The proposed mechanism is unglamorous and physical: a brighter screen makes your pupil contract, a smaller pupil reduces optical aberrations, and less aberration means finer detail survives the trip to your retina.

Now put that next to a 20-stroke character. The strokes and the gaps between them are the fine detail. That's not an argument against dark mode, but it does mean the safety margin you had in Latin type is thinner here, and small CJK text on a dark background is where it runs out first.

Then there's the part that connects straight back to the cultural material earlier in this guide, and it's a genuinely awkward collision.

Fan, Xie, Dong and Wang ran a controlled study on text colour under negative polarity and published it in Sensors (2024). Fifty participants, a black background at 28 candela per square metre, three ambient light levels from 0 to 10 lux, and five text colours. They ran a 20 minute reading task plus 50 symbol recognition questions, and measured fatigue with eye tracking rather than only asking people how they felt.

The fatigue ranking came out red, then green, then blue, then white, then yellow, worst to best. Red text produced the lowest blink rate and the lowest pupil accommodation of the five. Cognitive performance told the same story: best with yellow, blue and white, and significantly worse with red.

Red. The colour this entire region reaches for when something matters.

That's the collision. Everything earlier in this guide about red carrying celebration, luck and emphasis is about what red means. This is about what red does to someone's eyes over twenty minutes on a dark screen. Both are true, and they pull in opposite directions the moment you ship a dark theme.

What to actually do with that, without overcorrecting.

• Keep red for accents, not for reading. A red badge, a red price, a red seasonal banner: fine, and culturally right. A red paragraph on a dark background is where the evidence says stop.
• Raise the floor on small CJK text in dark themes. If you were already going up a size or a weight for CJK, go a little further in dark mode rather than assuming the light-mode setting transfers.
• Test in the light people actually use. The study found ambient illumination mattered: at 10 lux the fatigue markers improved significantly against 0 lux. Phones in a dark bedroom are the hard case, not the office.

Two honest limits, because the polarity literature is not as tidy as the summaries suggest. The advantage narrows or reverses for some groups, with age effects running the other way, and dark mode does reduce reported fatigue in genuinely dim conditions. And the Sensors study tested symbol recognition on a black background at low brightness, which is a specific setup rather than a verdict on every dark theme ever shipped.

Our dark mode colour guide covers the general palette questions, elevation, surfaces and whether to use pure black. This section is only the part that changes because of the script and the cultural weight on red. So the takeaway isn't that dark mode is wrong. It's that dark mode plus dense scripts plus red text is a stack of three small costs that people tend to evaluate one at a time, and the region where that stack is most likely to occur is the one this guide is about.

Is any of this actually a legal requirement?

In one of these three markets, yes, and it reaches private companies. In the other two, mostly not, and the gap between them is bigger than most teams assume.

Start with Korea, because it's the strict one. The W3C's policy record for the Republic of Korea lists the Act on Anti-Discrimination against and Remedies for Persons with Disabilities, enacted in 2008 and last updated 30 June 2017, and records its scope as public sector and private sector. The technical standard sitting under it is the Korean Web Content Accessibility Guidelines 2.1, catalogued as TTAK.OT-10.0003/R2. So if you're shipping a commercial product into Korea, accessibility isn't a best practice you've chosen. It's the backdrop you're operating against.

Japan works differently. W3C's Japan record lists JIS X 8341-3:2016, developed by the Web Accessibility Infrastructure Committee, and states plainly that compliance with the standard is voluntary. Government sites get pulled in from a different direction, through the Ministry of Internal Affairs and Communications guidelines dated 20 April 2016, which are a procurement recommendation rather than a rule with teeth.

But something did change for private businesses recently, and it's easy to miss because it isn't a web standard at all. The Act for Eliminating Discrimination against Persons with Disabilities was amended with effect from 1 April 2024. Japan's Cabinet Office, in its FY2023 Annual Report on Government Measures for Persons with Disabilities, describes the change as moving provision of reasonable accommodation by companies from best endeavors to a legal obligation. That's not a contrast ratio. It's a duty to accommodate when someone asks.

China is the loosest of the three on paper. W3C's China record lists the Law on the Protection of Persons with Disabilities with a scope covering both public and private sector, but it isn't web-specific. The web standard that does exist, YD/T 1761 from the Ministry of Industry and Information Technology, is recorded with a scope of government only, and W3C describes it as based on an early, non-Recommendation version of WCAG 2.0 published in 2008.

Which market's rules should you actually build to?

Build to current WCAG and stop tracking the three national numbers separately. That sounds like a dodge. It isn't, and the reason is in the dates.

Look at what those standards derive from. Japan's JIS X 8341-3:2016 is a WCAG 2.0 derivative. Korea's KWCAG 2.1 is built on WCAG 2.0 success criteria. China's YD/T 1761 traces to a draft of WCAG 2.0 from before it was even a Recommendation. WCAG 2.0 was published in 2008. Every regional standard here is anchored to a document that's now well over fifteen years old, and none of them will move as fast as your product does.

Which gives you a clean shortcut. Meeting current WCAG AA clears all three by construction, because they're derivatives of an earlier version of the same thing. You don't need three compliance workstreams. You need one, set at the level that swallows the others.

Two things this doesn't let you skip

The first is the derivative problem. A standard built on WCAG 2.0 "considering domestic condition," which is how W3C describes the Korean guidelines, is not the same document as WCAG. Local criteria get added and reworded. So if you're in a procurement process or answering a formal accessibility question in Korea, read the actual KWCAG text rather than handing over a WCAG audit and assuming it maps line for line. Building to a superset protects your users. It doesn't automatically fill in someone's compliance form.

The second is that Japan's 2024 change has a different shape from the others. A conformance standard is something you test against on a schedule. A duty to provide reasonable accommodation is triggered by a person asking. You can pass every automated scan and still be unprepared, because the thing being asked of you is a response, not a score. Worth knowing who in your organisation fields that request before someone sends it.

And notice where this leaves the argument on this page. Almost everything about colour symbolism in these three countries turned out to be weaker and more shared than the design blogs claim. This is the exception that runs the other way. The rules genuinely do differ by country, they differ in a way that changes what you're obliged to do rather than what looks nice, and Korea binding private companies while Japan's standard stays voluntary is a real distinction you can act on. If you were looking for the part of this topic where the border actually matters, it's here, and it has nothing to do with what red means.

What does have a measured effect in the region?

Fair question after several sections of "less than you'd think." Symbolism is weakly evidenced. Colour itself isn't, and there's regional research to point at.

Yu, Liu, Kumar and colleagues reviewed 73 peer-reviewed studies of urban colour in Chinese and South Korean cities for Frontiers in Psychology (2026, vol. 17), using structural equation modelling to link measurable colour properties to how people actually felt and behaved. Three pathways came out consistently:

• Mid-range colour complexity raised perceived safety and comfort, with a median effect of β around 0.43. Not the brightest scheme and not the plainest. The middle.
• Dominant hue and saturation improved wayfinding and sense of place, β around 0.30. People navigate partly by colour, and a scheme with no dominant hue gives them less to work with.
• Cool and reflective palettes cut heat stress and annual building energy use by 2 to 4 percent. That one is physics rather than psychology, and it shows up on a bill.

The most useful finding is the shape rather than the size. Street-level colour complexity followed an inverted U, with emotional pleasure peaking around 0.86 on their complexity measure and coordination peaking at about 0.84, falling away on both sides of each. Too plain reads as bleak. Too busy reads as chaotic. Colour harmony separately predicted well-being (β = 0.52) and place identity (β = 0.37).

Two things about that peak are worth knowing before you treat 0.86 as a target. It came from street data in Xi'an specifically, not from the region as a whole, which is a neat illustration of this article's own closing point about researching the actual city rather than the continent. And the complexity figure isn't a designer's judgement call. The review draws on studies that quantified colour from street-view imagery and drone footage, so it's a measurement of what is physically there rather than an impression of it.

Which is the reason to trust the shape more than the number. You are unlikely to compute a complexity score for your own work, and you don't need to. The transferable finding is that there is a middle, that both ends of the range cost you, and that somebody has now measured where the middle sits in at least one real city.

Notice what none of that is about. Not one of those pathways depends on red meaning luck or white meaning mourning. They're about how much colour, how varied, and how well it fits together. So if you want your palette to do measurable work in a Chinese or Korean context, complexity and harmony are better levers than symbolism, and they're the ones with numbers attached.

Worth saying this is environmental and urban research, so it transfers to physical spaces, signage and wayfinding more directly than to a mobile app. But the underlying point travels: the properties of your palette are doing more work than the meanings you assigned to it.

And there's a second regional difference that's genuinely measured, sits in the medical literature rather than the design literature, and almost never makes it into cultural colour guides. How many of your users can tell red from green at all.

Jennifer Birch reviewed the population surveys for the Journal of the Optical Society of America A (2012, vol. 29, no. 3) and found the rates aren't the same everywhere. Red-green deficiency runs at about 8 percent of men and about 0.4 percent of women among European Caucasians, but between 4 and 6.5 percent of men of Chinese and Japanese ethnicity. Birch also notes the male to female prevalence ratio is markedly different between Europeans and Asians, so this isn't a single number scaled up or down.

A large single-population study lands in the same range. Gao and Tian screened 17,303 Chinese college students for the International Journal of Ophthalmology (2023, vol. 16, no. 9) and found colour vision deficiency in 4.11 percent of the men and 0.16 percent of the women, 2.20 percent overall. They also found the effect showed up most in work, and noted that Chinese universities restrict students with CVD from majors including medicine, public security technology, chemical engineering, agriculture and forestry, art and astronomy.

Two things follow, and they pull in opposite directions. The rate among men is somewhat lower than the European figure most Western accessibility guidance is built on, so if you've been designing to an 8 percent assumption you're not under-preparing. But 4 percent of men is still roughly one in 25, which on any product with real traffic is a large number of people, and the second study suggests they're more likely to have been steered away from visual professions than to have been accommodated by them.

So the practical rule doesn't change, it just stops being a Western import. Never carry meaning on hue alone. Pair colour with a label, an icon, a position or a shape, and check your palette against a deficiency simulator before it ships. Our colour accessibility and WCAG guide covers the contrast side of that in detail. This is the part of "designing colour for East Asia" that has actual population numbers behind it, which is more than the symbolism can say.

Does the colour itself change how people behave?

In one specific case, yes, and it's measured. It also happens to be the best evidence on this whole page for the argument the article has been making, which is that culture bends a real effect rather than inventing one.

Go back to the charts. Everything above treats red and green on a financial display as a convention problem: get it backwards and you've reversed your own data. That's true, and it's the practical point. But there's a second layer underneath it, because the colour isn't only labelling the number. It's changing what people do with it.

William Bazley, Henrik Cronqvist and Milica Mormann ran eight experiments with 1,451 people for a paper called "Visual Finance: The Pervasive Effects of Red on Investor Behavior", published in Management Science in 2021. They showed people identical financial information and varied only the colour it was rendered in.

Red moved three things: risk preferences, expectations about future returns, and actual trading decisions. The detail worth holding onto is what it did to pessimism. When losses were shown in red, pessimistic expectations persisted. The same numbers in black or blue produced reversal beliefs instead, meaning people expected a bounce. Same data, different colour, different forecast.

Then the finding that makes this a section rather than a footnote. The effects were muted in China, where red carries prosperity rather than danger.

Sit with what that shows. This isn't a case of one culture having an effect and another not having it. It's a genuine, replicated behavioural effect whose strength depends on what red has been trained to mean where you are. Western red draws on the teacher's pen and the overdrawn account. Chinese red draws on the envelope and the wedding. The mechanism is shared. The learned content is local, and it turns the dial.

One control in that study is worth flagging, because it also connects to the colour vision section above. The effects vanished among colourblind participants, and other colours didn't reproduce what red did. So this is specifically about perceiving red, not about visual emphasis in general. That's a tighter result than most colour research manages.

What to actually do with it, if you build anything that shows people numbers they care about:

• Treat red as a load-bearing choice on financial displays, not styling. It affects what users predict will happen next, which is a strange amount of power for a stylesheet to hold.
• Don't assume porting the palette ports the effect. A design tuned to soften red-driven panic in a US product is solving a problem that is measurably weaker in a Chinese one.
• Never let colour carry the meaning alone. Arrows, signs and labels survive both the convention flip and colour vision deficiency. The article makes this point for accessibility reasons. This is the same fix for a second reason.
• Keep it in proportion. This is finance, where colour maps directly onto gain and loss. It is not a licence to assume red is doing similar work on a marketing page.

And that last line is the through-line of the whole page. The effects that survive scrutiny are narrow and situational. This one earns its place because somebody measured it, found where it weakened, and could say why.

How should you research color for a specific market?

Look at what people in that market already use, not at what articles say they believe.

1. Open the top five local competitors. Not their marketing site, the actual product. Sample the real colours with an image colour picker and note the palette.
2. Check the category, not just the country. Finance, gaming and beauty look completely different from each other within the same market.
3. Look for festival variants. Does the brand shift its palette for Lunar New Year? That tells you where symbolism is live and where it isn't.
4. Ask actual users. Five people from the market will tell you more than any guide. Show them the palette without explanation and ask what it feels like.
5. Then build the palette properly. Contrast, consistency, and a sensible number of steps. Our brand colour palette guide covers the mechanics, and the palette generator handles the construction.

The reason this order matters: symbolism lists tell you what a culture supposedly believed. Competitor products tell you what its users are actually looking at this week. When the two disagree, the products are right.

And a closing caution worth stating plainly. "East Asian culture" covers China, Japan, Korea, Taiwan, Hong Kong and diaspora communities across Southeast Asia, with different languages, histories and design conventions. Treating them as one colour system is the deepest error in this genre, and it's usually the first one made. If your market is Seoul, research Seoul.

What else do people ask?

Does red really mean luck in Chinese culture?

In specific contexts, yes. Red is genuinely tied to celebration, weddings and Lunar New Year, and red envelopes are a real practice, not a design myth. What is overstated is the idea that red carries that meaning everywhere it appears. On a dashboard, a red error state reads as an error to a Shanghai user the same as anywhere else.

Are color meanings actually different across cultures?

Less than design blogs suggest. Jonauskaite and colleagues surveyed 4,598 people across 30 nations speaking 22 languages, publishing in Psychological Science in 2020, and found colour-emotion associations were substantially universal, with an average similarity coefficient of r = .88. Nation still predicted variation beyond that, but the shared baseline is large.

Should you avoid white in designs for East Asian markets?

No. White is associated with mourning in parts of the region, which is real, but it is also the background of nearly every website, app and product page used there daily. The association attaches to specific contexts like funerals and gifts, not to interface backgrounds. Blanket avoidance is the wrong lesson.

Does Japanese have a different word for light blue?

Yes, and this one genuinely affects design work. Tohoku University research published in the Journal of Vision in 2017 found mizu, meaning water, functions as its own colour category in modern Japanese, used by 98 percent of the study's informants. English treats light blue as a shade of blue. Japanese speakers may treat it as a separate colour.

How should you actually research color for an Asian market?

Look at what local products already do rather than reading symbolism lists. Open the top apps and retail sites in that market and note the actual palettes in use. Then test with local users. A twenty-minute look at real competitors beats any generic guide, including this one, because it reflects current practice rather than folklore.

Sources: Hurlbert, A.C. and Ling, Y., "Biological components of sex differences in color preference," Current Biology 17(16), R623 to R625, 2007, PMID 17714645, for the 208 British and Han Chinese participants aged 20 to 26 and the reddish versus blue-green split. Yokosawa, K., Schloss, K.B., Asano, M. and Palmer, S.E., "Ecological Effects in Cross-Cultural Differences Between U.S. and Japanese Color Preferences," Cognitive Science 40, 1590 to 1616, 2016, PMID 26400420, for the shared and differing preferences, the complex gender differences and the intermediate multicultural results. Taylor, C., Clifford, A. and Franklin, A., "Color preferences are not universal," Journal of Experimental Psychology: General 142, 1015 to 1027, 2013, PMID 23148465, for the Himba comparison. Bazley, W., Cronqvist, H. and Mormann, M., "Visual Finance: The Pervasive Effects of Red on Investor Behavior," Management Science, 2021, eight experiments with 1,451 participants, for the effects of red on risk preferences, return expectations and trading decisions, the persistence of pessimism under red against reversal beliefs under black or blue, the muting of these effects in China, and the absence of any effect among colourblind participants. W3C Web Accessibility Initiative, Understanding Success Criterion 1.4.3: Contrast (Minimum), for the 4.5:1 and 3:1 ratios, the definition of large-scale text including the explicit CJK equivalence carve-out, the derivation of the 18pt and 14pt figures from roman large-print conventions, and the statements that larger text with wider character strokes is easier to read at lower contrast and that extraordinarily thin strokes are harder to read at lower contrast levels. Liu, N., Yu, R. and Zhang, Y. (2016), "Effects of Font Size, Stroke Width, and Character Complexity on the Legibility of Chinese Characters," Human Factors and Ergonomics in Manufacturing and Service Industries 26(3):381 to 392, a within-subjects character-search experiment with 36 participants, for the significant effects of font size and character complexity, the non-significant main effect of stroke width alongside significant two-way interactions, stroke width being critical at high character complexity and font size at low complexity, and the 12pt with 1:10 stroke width combination performing best across both. Jonauskaite, D. et al. (2020), "Universal Patterns in Color-Emotion Associations Are Further Shaped by Linguistic and Geographic Proximity," Psychological Science 31(10), for the 4,598 participants across 30 nations and the r = .88 similarity coefficient. Jonauskaite, D. and Mohr, C. (2025), "Do we feel colours? A systematic review of 128 years of psychological research linking colours and emotions," Psychonomic Bulletin and Review, published 13 January 2025, PMC12325498, for the 132 studies, 42,266 participants, 64 countries, the 1895 to 2022 range, and the consistency across countries, age groups and time periods. That review covers colour-emotion association, not social convention, so it does not speak to practices like gift or ceremonial colour. Kuriki, I., Lange, R., Muto, Y., Brown, A.M., Fukuda, K., Tokunaga, R., Lindsey, D.T., Uchikawa, K. and Shioiri, S. (2017), "The modern Japanese color lexicon," Journal of Vision 17(3), reported by Tohoku University, for the 52 informants, 320 samples, 19 categories and the 98 percent figure for mizu. Madden, T.J., Hewett, K. and Roth, M.S. (2000), "Managing Images in Different Cultures: A Cross-National Study of Color Meanings and Preferences," Journal of International Marketing 8(4), 90 to 107. Saito, M. (1996), "A comparative study of color preferences in Japan, China and Indonesia, with emphasis on the preference for white," Perceptual and Motor Skills 83(1), PubMed 8873184, for the 490 subjects across three countries and the shared high preference for white. Zhang, T. and Han, B. (2014), "Experience Reverses the Red Effect among Chinese Stockbrokers," PLoS One 9(2), PMC3933460, for the reversed red effect and the Chinese stock market colour convention. Note that study used 24 students and 24 stockbrokers, so its effect sizes are provisional. Birch, J. (2012), "Worldwide prevalence of red-green color deficiency," Journal of the Optical Society of America A 29(3), 313-320, for the 8 percent and 0.4 percent European Caucasian figures, the 4 to 6.5 percent range in men of Chinese and Japanese ethnicity, and the differing male to female prevalence ratio. Gao, J.G. and Tian, M. (2023), "Prevalence of color vision deficiency among Chinese college students and their quality of life," International Journal of Ophthalmology 16(9), 1542-1548, for the 17,303 student sample, the 4.11 percent male and 0.16 percent female rates, the 2.20 percent overall figure, and the university major restrictions. Yu, H., Liu, R., Kumar, R., Singh, S. and Kumar, R. (2026), "Urban color in public design: a review of spatial aesthetics and behavioral impact in Chinese and South Korean cities using structural equation modelling approaches," Frontiers in Psychology 17, DOI 10.3389/fpsyg.2026.1692740, PMC12886025, for the 73 studies, the median β of about 0.43 for mid-range colour complexity, the β of about 0.30 for dominant hue and chroma, the 2 to 4 percent energy reduction, and the Xi'an inverted-U peaks of about 0.86 for complexity and 0.84 for coordination. That review synthesises studies which quantified colour from street-view imagery and drone footage, and its findings are drawn from urban environments rather than screen interfaces. Zhang, X., Abdul Rani, M.F. and Zhang, Y. (2026), "Exploring color symbolism in ancient Chinese architecture and furniture design", Journal of Chinese Architecture and Urbanism 8(1), for the finding that colour decisions in Ming and Qing dynasty architecture and Forbidden City furniture reflected philosophical doctrine, material availability and sociopolitical function together rather than a fixed colour-to-element mapping. Xie, J.Z., Tarczy-Hornoch, K., Lin, J., Cotter, S.A., Torres, M. and Varma, R., "Color Vision Deficiency in Preschool Children: The Multi-Ethnic Pediatric Eye Disease Study," Ophthalmology, 2014, vol. 121, pages 1469 to 1474, for the prevalence figures in boys of 5.6 percent non-Hispanic white, 3.1 percent Asian, 2.6 percent Hispanic and 1.4 percent African-American, and 0 to 0.5 percent in girls across all groups, from roughly 4,005 preschool children aged 3 to 6 screened under a single protocol in Southern California. Those figures describe that study population and are cited as evidence that prevalence varies between groups, not as national rates. Roberson, D., Pak, H. and Hanley, J.R. (2008), "Categorical perception of colour in the left and right visual field is verbally mediated: evidence from Korean," Cognition 107(2), 752-762, PubMed 17931614, for the obligatory yeondu and chorok boundary in Korean, the faster cross-category than within-category discrimination by Korean speakers where English speakers showed none, the right visual field lateralisation in fast responders, and the weakening under verbal interference. Roberson, D., Hanley, J.R. and Pak, H. (2009), "Thresholds for color discrimination in English and Korean speakers," Cognition 112(3), 482-487, PubMed 19619872, for the absence of any difference in just-noticeable-difference discrimination thresholds between the two language groups despite their differing supra-threshold category boundaries. Sun, V.C. and Chen, C.C. (2018), "Basic color categories and Mandarin Chinese color terms," PLOS One 13(11), PMC6261562, for the 59 native Mandarin-speaking observers in Taiwan, the 280 Natural Color System samples under D65 illumination, the 32 single-character terms, the principal component analysis, the finding that Mandarin basic colour categories resemble those in the World Color Survey while being represented by wide-spread and inconsistent terms among speakers, the scatter in Mo, Tie and Cai, and white, blue and purple not explaining enough variance to qualify as major factors. That study sampled Taiwanese speakers, so term usage may differ on the mainland. Cultural practices described here vary by community and generation, so treat them as starting points for research rather than rules.