Sign language interpretation

Criterion 1.2.6 asks for sign language interpretation of prerecorded audio, and it sits at level AAA, which is why most teams have never had to think about it. The reason it exists is the part worth understanding, because it is not obvious from the outside: a sign language is a language of its own, not a transcription of the spoken one. British Sign Language and American Sign Language have their own grammar, their own idioms, and no genetic relationship to English or to each other. Facial expression and the space in front of the signer carry grammatical meaning, which is why an interpretation is a translation and not a relabelling.

That is what captions cannot fix. Captions present the spoken words in written form, so for a reader whose first language is a signed one they are a second language, read at speed, while also watching the picture. Many Deaf readers are entirely comfortable with that and many are not, in the same way that subtitles in your third language are technically sufficient and quietly exhausting. Reading age in the written language of the surrounding country is, on average, lower among people who have signed from birth, which has nothing to do with ability and everything to do with the written language being learned second, largely through the eye.

Presentation is where designs usually go wrong. The common pattern is a small inset in a corner of the video player, and the common failure is making it small enough to be free. Hands move fast, fingerspelling is fine detail, and the face is grammar, so a thumbnail that keeps the picture uncluttered has thrown away most of the language. Better provisions give the signed track real size: a companion stream the same size as the video, a picture-in-picture the reader can scale and move, or a version of the whole video with the interpreter composed in at a usable size. Whatever the arrangement, the reader chooses it, and the choice belongs in the same place as the caption and audio description controls rather than on a separate page.

Two practical notes. Production cost is real, so it lands first where it pays best: safety information, legal and medical content, anything a public body publishes, and evergreen video that will be watched for years. And an interpreter is a person, so their track has its own needs: even lighting, a plain background, contrasting clothing, no cropping at the wrists, and a shot wide enough for signs that use the whole signing space. Getting the framing wrong is the quickest way to spend the entire budget on a track nobody can read.