accessibility
Plain language
also called reading level (wcag), plain English (govuk), lower secondary reading level (wcag)
Writing interface copy at the reading level of the people using it, with short sentences, familiar words, and the point first, rather than at the level of the org chart.
Plain language is not simplified language, and it is not a tone. It is writing at the reading level of the people who have to act on the writing, which usually means short sentences, words people already own, the active voice, and the thing the reader must do placed at the front rather than after the justification. WCAG puts a number on the bar: success criterion 3.1.5 Reading Level asks that text not require reading ability more advanced than the lower secondary education level, and where it does, that a version or a summary at that level be available. It sits at Level AAA, which is why so many teams treat it as optional. The population it serves does not experience it as optional.
It is worth separating from readability, because the two words get used interchangeably and they name different failures. Readability is a property of the setting: the size, the contrast, the measure, the leading, whether the letterforms hold up at that size. Plain language is a property of the words: whether a person who can see the sentence perfectly well can also get the instruction out of it. A sentence can be beautifully set and impossible to parse, and a plain sentence can be set in grey nine pixel type. They fail independently, and fixing one has never once fixed the other.
The audience is much wider than the accessibility framing suggests, which is the usual shape of a curb cut. It includes people with cognitive and learning disabilities and people with low literacy, and it also includes everyone reading in a second language, everyone reading on a phone in a queue, and everyone reading while frightened, which is most people at the moment they meet an error message or a bill. It compounds with the reader-side controls: someone using dynamic type at its largest setting sees a handful of words per line, and a forty word sentence at that size is a paragraph they have to hold in their head. The place it pays best is wherever the interface blocks the way, so the copy on an interstitial, a permission prompt or a consent screen is worth more editing attention than anything on the happy path.
In practice the failures are boringly consistent. Copy inherits the vocabulary of the organisation rather than the reader: the names of internal systems, the legal wording that was safe to sign off, “submit” instead of the thing being sent. The remedy is equally boring. Say what the reader has to do, then say the deadline, then say what happens if they miss it. Name things the way the reader names them. Then measure the result rather than arguing about it, because sentence length and word familiarity are countable, and a first draft that reads at university level almost always says the same thing in half the words.
Which word?
| If you want | say |
|---|---|
| the copy is correct but nobody can act on it | plain language |
| a dense block has to be read by anyone tired | chunking |
| judging the setting, not the typeface | readability |
| a screen is correct but exhausting to use | cognitive load |
| checking what survives when the design is stripped | reader mode |
| the words and icons themselves are the barrier | symbol support |