Emoji · Long read
Emoji were not invented in 1999, and Unicode did not adopt them in 2010. Both dates are repeated everywhere and both are wrong. Here is the actual sequence — from a badly-selling Japanese handset in 1997 to a standard of 3,953 characters — with every claim traced to a primary source.
1997 → today
Quick answer
The first emoji set shipped in November 1997 on the J-Phone SkyWalker DP-211SW, a SoftBank handset carrying 90 characters. The famous set — 176 emoji designed by Shigetaka Kurita for NTT DoCoMo — came in 1999 and is the one that made emoji ordinary in Japan, because i-mode was popular and the DP-211SW was not. Kurita publicly agreed he had not been first once the earlier set was documented in 2019.
Unicode agreed to encode emoji in August 2007. The first characters landed in Unicode 5.2 in 2009, and the bulk of the Japanese carrier sets followed in Unicode 6.0 in 2010 — which is the year most accounts give as the beginning, three years after the actual decision.
Timeline
Four phases, running from a carrier feature nobody outside Japan could see to a standard that grows mostly by recombining what it already has.
Emoji begin as a carrier feature on Japanese phones, incompatible between networks and invisible to everyone else.
On 19 September 1982, Scott Fahlman posted to a Carnegie Mellon bulletin board suggesting that :-) be used to mark posts meant as jokes. It is not emoji; it is punctuation read sideways. But it establishes the idea that a message needs a tone marker the words cannot carry.
The second half of the proposal is usually forgotten. Fahlman suggested :-( not for sadness but to flag posts that were *not* jokes, on the grounds that those were becoming the rarer case.
The J-Phone SkyWalker DP-211SW, released in November 1997, carried 90 emoji — including an early pile of poo. The handset sold poorly and the set was largely forgotten, so the credit went elsewhere for twenty years.
Emojipedia established this in 2019. Shigetaka Kurita, long named as emoji's inventor, publicly agreed, noting that the J-Phone DP-211SW came before his own set.
Shigetaka Kurita designed a 176-character set for DoCoMo's i-mode service, each drawn on a 12×12 pixel grid. It was not the first set, but i-mode was popular and the DP-211SW was not, so this is the one that made emoji ordinary in Japan.
MoMA acquired the original set in 2016, cataloguing it as 1998–1999. Its gallery label names Kurita's sources: manga, emoticons, and the Zapf Dingbats typeface — the same dingbat tradition that had been supplying printers with pictorial characters for decades.
DoCoMo, au (KDDI) and SoftBank each ran their own emoji, with their own code assignments. Sending a message between networks meant passing it through a conversion table, and anything without a match arrived as a blank or the wrong picture. This is the problem Unicode was eventually asked to solve.
Google and Apple push the Japanese sets into the Unicode Standard so a character survives the trip between networks.
Google began converting the Japanese carrier emoji to Unicode private-use code points — a workable stopgap that got Gmail handling them, though private-use characters mean nothing outside the system that assigns them. It made the case for encoding the set properly.
In August 2007 the Unicode Technical Committee agreed to support encoding emoji, working from principles developed by a dedicated subcommittee. The proposal came from Google — Mark Davis, Kat Momoi and Markus Scherer — with Apple's Yasuo Kida and Peter Edberg, and input from Microsoft.
This is the date the common retelling skips. Emoji were not adopted in 2010; that is simply when the bulk of the characters shipped, three years after the decision.
iPhone OS 2.2, released in November 2008, added an emoji keyboard for the Japanese market, where a carrier could not sell a phone without one. It predated Unicode's emoji encoding entirely. Users elsewhere unlocked it with third-party apps, which is the first sign of demand outside Japan.
Unicode 5.2 carried 114 characters intended as emoji, alongside symbols drawn from the Japanese broadcasting standard ARIB. The encoding work had started, but the bulk of the carrier sets was still outside the standard.
The remaining 608 Japanese carrier emoji were added in Unicode 6.0, completing the job. From this point a Japanese carrier emoji was an ordinary Unicode character: it could travel between networks, operating systems and applications as text.
Unicode labels this era Emoji 0.6 in retrospect. The numbering came later, so the earliest emoji carry a version that never shipped as a release.
The keyboard leaves Japan, emoji reach every phone, and the set starts growing by public proposal.
iOS 5 made the emoji keyboard available as a standard option in every region. This is the single largest step in emoji's spread: a feature built for one national market became a default on the most visible phone in the world.
Android 4.4 shipped native colour emoji, and by the middle of the decade every major platform rendered the same characters with its own artwork. Emoji stopped being a thing your phone either had or did not have.
Unicode formalised emoji as a versioned standard of its own with Emoji 1.0, introducing skin-tone modifiers based on the Fitzpatrick scale. In November, Oxford Dictionaries named 😂 its Word of the Year — the first time a pictograph took the title.
Oxford leaned on SwiftKey data showing 😂 accounted for around 20% of UK and 17% of US emoji use that year, up from 4% and 9% in 2014.
Growth shifts from new characters to new sequences, and the argument moves from what exists to what it should look like.
Apple replaced its revolver artwork for U+1F52B with a bright green water pistol in iOS 10. Twitter, Google, Microsoft, Facebook, Samsung and WhatsApp all followed by 2018, and the code point that once showed a firearm on every platform now shows a toy on all of them.
Nothing changed in Unicode. One code point, unchanged, went from weapon to toy purely through vendor artwork. Nothing demonstrates more plainly that the standard fixes the character and not the picture.
Unicode opened a public proposal process, and the additions of this period reflect it: more foods, more animals, more disability and accessibility representation, and gender and hair-colour variants built as sequences rather than new characters.
Emoji 15.1 added its new concepts almost entirely as sequences — directional variants of existing people, for instance — rather than as new code points. It is the clearest statement of the modern strategy: combine what exists instead of encoding something new.
The current release brings the total to 3,953 emoji, but only 1,388 of those are basic emoji. The rest are sequences: skin tones, gender and hair variants, families, flags and keycaps, all assembled from parts already in the standard.
Before emoji
The two are often blurred together, but they solve the same problem by different means. Emoticons are improvised: ordinary punctuation arranged to look like a face, invented because the character set of the day offered nothing better. Western emoticons read sideways, as Fahlman's :-) does. Japanese kaomoji read upright — (^_^) needs no tilt of the head — and draw on a much larger pool of characters, which is why they can manage arms, tables and weather.
Emoji went the other way. Rather than borrowing characters to approximate a picture, they are characters that are pictures, encoded as themselves. That distinction is why an emoji renders as artwork on every platform while an emoticon renders as whatever the font does with a colon and a bracket — and why emoji, not emoticons, ended up needing an international standards body.
How it works
Almost every argument about emoji — why yours looks different, why one arrived broken, why the total is so large — comes back to these four mechanics.
Every emoji is a Unicode character with a permanent identity — 🔥 is U+1F525 everywhere, forever. What you see is a font, drawn by Apple, Google, Microsoft, Samsung or whoever made your keyboard. The character is standardised; the picture never was.
Older symbols that existed as text before they were emoji need a hint to render in colour. That hint is U+FE0F, an invisible character appended to the base. It is why ☮️ is two code points and 🔥 is one, and why a stripped variation selector turns a colourful emoji into flat black type.
A ZWJ (U+200D) glues separate emoji into one glyph. A family, a professional role, a flag on a pole: these are not single characters but recipes. Where a platform does not know the recipe, it falls back on showing you the ingredients side by side.
The five tones added in Emoji 1.0 are modifier characters appended to a base emoji rather than separate emoji. One waving hand plus five modifiers produces six renderings from two characters, which is why the People & Body group is far larger than any other.
By the numbers
Counted from Unicode’s own data for Emoji 17.0, using fully-qualified sequences. The shape of this table is the story of the modern set: one group is larger than all the others combined, because its entries multiply out across skin tones and variants.
| Unicode group | Count | Why it is that size | Browse |
|---|---|---|---|
| People & Body | 2,418 | Enormous because almost every entry multiplies out across five skin tones, two genders and several hair colours. | Hands & Gestures |
| Flags | 270 | Mostly country flags built as pairs of regional indicator letters rather than as individual characters. | — |
| Objects | 266 | Tools, devices, clothing and household things — the group that dates fastest as the objects themselves change. | Objects & Tech |
| Symbols | 224 | Signage, arrows, zodiac glyphs, hearts and status marks. Many were ordinary typographic characters long before they were emoji. | Symbols & Reactions |
| Travel & Places | 219 | Vehicles, buildings and landscapes, plus the weather and sky symbols that double as mood setters in captions. | Weather & Nature |
| Smileys & Emotion | 171 | The smallest group doing the most work: faces carry tone, and tone is what plain text is worst at. | Smileys & Emotion |
| Animals & Nature | 160 | Creatures and plants, used literally about as often as they are used for personality and nicknames. | Animals & Nature Life |
| Food & Drink | 131 | Meals and ingredients, several of which have picked up slang meanings well away from the food itself. | Food & Drink |
| Activities | 85 | Sports, games, arts and celebrations — the group most often used as a label rather than a reaction. | Activities & Play |
Case study
U+1F52B was encoded as a pistol, and for years every platform drew a firearm. In August 2016 Apple announced it would ship a bright green water pistol instead, and by April 2018 Twitter, Google, Microsoft, Facebook, Samsung and WhatsApp had all followed. Today the character shows a toy everywhere, and Unicode itself now names it water pistol.
Not one thing changed in the standard. The same code point, sent from the same keyboard, simply arrived meaning something else — decided by half a dozen design teams rather than by a committee or by the people using it. It is the sharpest illustration of what Unicode does and does not control, and it is worth remembering whenever an emoji seems to say something different than you intended: you sent a character, and the recipient’s phone chose the picture.
Questions
Shigetaka Kurita is usually credited for the 176-emoji set he designed for NTT DoCoMo in 1999, and that set is the one that made emoji mainstream in Japan. He was not first, though: SoftBank, then trading as J-Phone, shipped 90 emoji on the SkyWalker DP-211SW in November 1997. Kurita himself acknowledged this in 2019 after Emojipedia documented the earlier set.
November 1997, on the J-Phone SkyWalker DP-211SW, which carried 90 emoji. The more famous set — Kurita's 176 emoji for DoCoMo's i-mode service — followed in 1999. The 1997 handset sold badly, which is why the 1999 set was credited as the first for roughly two decades.
The Unicode Technical Committee agreed to encode emoji in August 2007. The first characters explicitly intended as emoji arrived in Unicode 5.2 in 2009, and the remaining 608 Japanese carrier emoji were added in Unicode 6.0 in 2010. Unicode later labelled this early material Emoji 0.6 retrospectively, which is why the oldest emoji carry a version number that never shipped as a release.
It is Japanese: 絵 (e, picture) plus 文字 (moji, written character) — literally picture-character. The resemblance to the English words emotion and emoticon is a coincidence rather than an etymology, which is why emoji and emoticons are two separate things with unrelated histories.
An emoticon is built from ordinary punctuation, like :-) or the Japanese kaomoji (^_^). An emoji is a Unicode character in its own right, rendered as a picture by the platform's font. Emoticons came first, in 1982, and were a workaround; emoji were designed as characters from the start.
Unicode standardises the character, not the artwork. Every platform draws its own font, so the same code point can be a different colour, angle or expression on each. The pistol emoji is the clearest case: U+1F52B showed a revolver everywhere until Apple redrew it as a water pistol in 2016, and by 2018 every major vendor had followed, without a single change to the standard.
Emoji 17.0 counts 3,953 in total, but only 1,388 are basic emoji. The rest are sequences assembled from existing parts: skin tones, gender and hair variants, families, keycaps and flags. That is why the People & Body group alone accounts for over 2,400 entries.
Yes. Unicode accepts public proposals, which have to argue for expected usage, distinctiveness and completeness, and explain why the idea is not already covered by an existing character. Most are rejected. Recent releases lean towards building new concepts out of sequences rather than encoding new characters at all.
Research notes
Dates on this page come from Unicode’s own documentation wherever Unicode documents them, and from the primary record everywhere else.
Unicode's own account of how emoji were encoded: the August 2007 committee decision, the 114 characters in Unicode 5.2 and the 608 added in Unicode 6.0, and the Google, Apple and Microsoft contributors involved.
The 2019 research establishing that SoftBank's 1997 DP-211SW carried 90 emoji two years before DoCoMo's set, including Shigetaka Kurita's own acknowledgement.
The museum record for the original 176 DoCoMo emoji, acquired in 2016 and drawn on a 12×12 pixel grid.
Fahlman's own account of the 19 September 1982 bulletin-board post that proposed :-) and :-( as tone markers.
The official totals used on this page: 1,388 basic emoji and 3,953 including every sequence.
The version-by-version chart behind the release dates cited here and on every emoji page on this site.
Keep exploring
Each entry carries the Unicode version it was first defined in, taken from the same Emoji Versions chart cited above.
Browse emojiThe upright Japanese emoticons that ran alongside emoji rather than being replaced by them.
Browse kaomojiThe typographic characters that predate emoji entirely, and still work as plain text.
Browse symbols