Verbosity

A screen reader does not read out everything it knows. It reads out what its user has told it to read out. Verbosity is the collection of settings that makes that decision: whether roles are spoken, whether states are spoken, whether descriptions are spoken, whether punctuation is announced by name, whether the row and column of a table cell come with the cell. NVDA groups these under document formatting and speech settings, JAWS ships three named verbosity schemes, and VoiceOver hides them under a rotor of its own. The details differ. The consequence does not: two people can reach the same button and hear two different sentences.

This is the fact that should end most arguments about label length. An accessible name is spoken at every verbosity level, because a thing with no name cannot be referred to at all. An accessible description is not. Descriptions are commonly delayed, truncated, or switched off entirely, and experienced users switch them off precisely because most of them are noise. So a hint that only exists in a description is a hint most of your readers will never hear, and moving the load-bearing half of it into the name is usually the whole fix.

The other half of the lesson is about restraint. Verbosity exists because assistive technology users are reading at speed, often at 400 words a minute or more, and every extra word you add to a control is a word they pay for on every pass through the page. Turning it down is how a practised user makes an interface usable. That is why “we will just add it to the description” is not a plan, and why padding a name with instructions (“Submit, click here to submit your completed application”) makes the interface slower for exactly the people the padding was aimed at.

Never test with verbosity turned up. Turn it down to its lowest useful setting, walk the interface, and see what survives. What you hear there is what most of your readers hear too.