Why Family Emoji Fall Apart: Joined Sequences Explained

Send 👨‍👩‍👧‍👦 to the wrong place and you do not get a family. You get a man, a woman, a girl and a boy standing in a row, taking up four times the space and looking nothing like what you sent.

That is not a bug. The family emoji genuinely is four separate people with invisible glue between them, and the glue is the first thing to go when text passes through software that does not know about it.

What the glue is

The joining character is called the Zero Width Joiner, written U+200D. It has no width, no appearance and no meaning of its own. Its entire job is to tell a font that the characters on either side should be drawn as one thing rather than two.

It is much older than emoji. The joiner entered Unicode in 1993 for scripts such as Arabic and Devanagari, where letters change shape depending on what they connect to. Emoji borrowed a mechanism that had been quietly doing typographic work for two decades.

When a font has artwork for the full joined sequence, you see one glyph. When it does not, the fallback is to draw each component separately and ignore the joiners. There is no error and no warning, which is why the failure looks like a mistake you made.

How many there are

We counted every joined sequence in our dataset. There are 188 of them, and they cluster heavily:

• 174 are people and body emoji. Professions, gendered variants, hair colours, couples and families.
• 7 are faces, including 😺 style variants and a small number of expressive combinations.
• 2 are animals, the service dog and guide dog built from a dog joined to another symbol.
• 2 are flags, which use the joiner differently to build regional banners.
• The remainder are scattered across food and objects.

By complexity: 160 use a single joiner, 16 use two, and 12 use three. The three joiner group is where the trouble concentrates.

The longest ones in the set

The most elaborate sequences are the kiss and family emoji:

• 👩‍❤️‍💋‍👨 kiss between a woman and a man, which is eight characters long
• 👨‍❤️‍💋‍👨and 👩‍❤️‍💋‍👩, also eightCopy
• 👨‍👩‍👧‍👦family of four, seven charactersCopy

Read the kiss sequence aloud and it is almost comic: woman, joiner, heart, colour selector, joiner, lips, joiner, man. Eight characters producing one small picture. On a platform with a character limit, that single emoji costs as much as a short word.

It also explains a behaviour people notice and cannot account for. Press backspace once on a family emoji in some apps and a child disappears, leaving a family of three. You have not deleted the emoji. You have removed one person from the sequence, and the font redrew what was left.

Where they break, specifically

Four places, in order of how often they cause trouble.

Old emoji fonts. A device whose font predates a given sequence has no artwork for it, so it draws the parts. This is the most common cause by a wide margin and it affects the newest sequences first.

Text fields with length limits. A field that counts characters rather than visible glyphs will see seven where you see one. Some truncate mid sequence, which produces half a family.

Databases with narrow columns. Older systems configured for a limited character range drop what they cannot store. The joiners are usually the casualty, leaving the components intact and separate.

Anything that normalises text. Search indexes, slug generators and some content filters strip characters they consider meaningless, and an invisible zero width character looks exactly like something meaningless.

If you see a stray ♂ or ♀ sign after a person emoji, you are looking at exactly this. A gendered variant pairs a person with a gender sign, so a failed join leaves that sign stranded by itself. The flag rendering piece documents a close cousin of this failure.

Counting characters, and why it matters

You can see the problem for yourself without any tools. Put a family emoji into any box that shows a character count and watch the number jump by seven.

This has real consequences in a few places. Posts with hard limits lose space disproportionately to composed emoji. Database fields sized for a certain number of characters hold fewer messages than expected. Systems that charge per message segment can split a message in an unexpected place, and splitting mid sequence is what produces the half rendered families people occasionally receive.

It also affects anything that sorts or deduplicates text. Two visually identical families can be encoded differently if one includes an extra colour selector, which means a strict comparison treats them as different strings. That is a genuine source of bugs in anything that stores user supplied text, and it is invisible during testing unless someone thinks to try it.

How to tell a composed emoji by eye

There is no visual marker, which is the awkward part. A few reliable heuristics:

• More than one person is visible. Families, couples and people holding hands are composed, with the exception of a few older single character versions.
• A person with a job or an object. Every profession emoji is a person joined to a thing.
• A gendered variant. If the same emoji exists in man, woman and neutral forms, the gendered ones are composed.
• Hair colour. All four hair variants are joined.
• It was added recently. Newer additions skew heavily composed.

The inverse is the useful version of this rule. A single yellow face, a single animal, a single object or a single symbol is almost always one character and will render anywhere. That is the set to reach for when reliability matters more than specificity, and most of them are in the face emoji collection and the objects category.

Why Unicode built it this way

The alternative was worse. Encoding every family combination as its own character would have meant thousands of new code points, since you would need one for every arrangement of parents, children and skin tones.

The joiner approach keeps the character count manageable by composing from parts already in the standard. One man, one woman, one girl, one boy and some glue covers a combination that would otherwise need its own permanent entry.

The cost is exactly what this article is about. Composed sequences are fragile in a way single characters are not, and the fragility is invisible until it is not. It is a reasonable trade, and it is why the emoji set is roughly 1,750 entries rather than tens of thousands.

The profession emoji are the same trick

👩‍🚒 woman firefighter is a woman joined to a fire engine. 👨‍🍳 man cook is a man joined to a frying pan. 👩‍💻 woman technologist is a woman joined to a laptop.

Once you know the pattern it is hard to stop seeing it, and it explains why these emoji look like two things awkwardly adjacent on platforms with weaker artwork. Some vendors genuinely do draw a person next to an object rather than integrating them.

It also explains the gaps. There is no profession emoji for anything that lacks a suitable object to join to, which is why the set covers cooks, farmers and pilots but not most office jobs. The full range is on our people and body page, and the individual job glyphs are collected under work emoji.

The couples, and the reason this group is so crowded

Couple and kiss emoji are the most heavily composed group in the set, and the history explains the clutter.

Originally there was one couple emoji and one kiss emoji, each a single fixed picture. Adding same sex versions meant either encoding every combination as a new character or building them from parts. Unicode chose parts, which is why 👩‍❤️‍💋‍👨 contains two people, a heart, a colour selector and a pair of lips held together by three joiners.

The payoff is that the same mechanism later allowed skin tones to be set per person, something a fixed picture could never support. The cost is that these are the longest and most fragile sequences on the keyboard, and they fail more visibly than anything else. A broken family looks like several people. A broken kiss looks like a small scene has exploded.

There is also a quieter consequence. Because the components are ordinary emoji, a platform that does not support the sequence still shows something comprehensible rather than nothing. That graceful degradation was part of the design argument, and it is why the fallback is a row of people rather than a row of empty boxes.

Hair colour works differently again

👨‍🦰 red hair, 👨‍🦱 curly hair, 👨‍🦳 white hair and 👨‍🦲 bald are built by joining a person to a hair component that exists only for this purpose.

Each hair component has its own entry in our meanings index. Those hair characters are real and you can type them alone, at which point you get a disembodied lock of hair. They were added in 2018 specifically so that hair colour would not require a separate character for every person emoji, which is the same economy argument as the families.

These break in a particularly ugly way. A failed join leaves a person followed by a floating patch of hair, which is more obviously wrong than a split family and considerably harder to explain to whoever received it.

Testing before you commit

If a sequence has to work somewhere important, check it rather than assuming.

• Paste it into the destination and look at the result, not at your own keyboard preview.
• Count the characters if the field has a limit. A family emoji will eat seven of them.
• Avoid them entirely in usernames, file names, slugs and anything that gets searched. Single character emoji are far safer there.
• Prefer a single character alternative where one exists. 👪 family is one character and renders everywhere, unlike the composed families.

The last of those matters most, because a simpler option almost always exists. If the specific composition does not matter, the single character version will survive places the sequence will not.

What this means for the future of the set

Unicode has leaned harder on composition over time, not less. Recent additions increasingly build on existing characters rather than introducing new ones, which keeps the standard compact and pushes more complexity onto fonts and rendering engines.

The practical consequence for anyone writing is that newer emoji are more likely to be composed, and therefore more likely to break on older devices. An emoji added last year is not just missing artwork on an old phone, it may fall apart into three visible pieces.

That is the real argument for checking the age of an emoji before relying on it, which you can do on our Unicode versions pages. The ones that never break are the oldest and simplest, several of which we dated in the 25 oldest emoji, and the related invisible character problem shows up again in how number emoji are assembled.

If something on your screen has already fallen apart and you want to work out why, start with the tofu box guide and the per platform comparisons on our Android and iPhone pages.