Label and situation together complete the sentence

What it is

The few words on a button rarely make a sentence by themselves. They assemble with the title, the selection, and the current page. A dialog asks “Save changes to Untitled?” and the buttons answer Save / Don't Save / Cancel—question plus answer is one sentence. A button that says Yes forces the question to be read; if the question is truncated or scrolled away, the answer hangs. Label and situation together complete the meaning: the assembly has to hold in a glance, not by packing the whole sentence onto every button.

Why it happens

When the eye hits a button, working memory already holds situation: the filename, how many items are checked, the dialog’s first line. The button is the predicate; the situation is subject and object. If the assembly works, gaze need not bounce between title and button. It fails in two ways. The button talks past the question (the question is about saving, the key says Continue). Or the button assumes the question is still in view while a custom title bar or a short viewport has cropped it. The second is common in responsive layouts: complete on desktop, only Yes / No on a phone. People do not automatically restore a missing subject; they hit the primary by visual habit.

Studying it

Run a mutilated-context test: the same button pair with a full question, with only the second half, or with the question scrolled out of view. Add a mismatch condition where the question and the button verb point at different events.

Independent variables: visibility of the question, whether question and button co-refer, viewport height. Dependent variables: choice accuracy, gaze round-trips between title and buttons, later retelling of “what I agreed to.”

Eye tracking shows whether assembly actually happens. If people click almost without looking at the title, they are betting on the primary and the situation is not part of the sentence.

Where it stops holding

Toolbar Bold has no question to complete; the situation is the current selection and the label is already a full command. Speech that reads a button name has no surrounding title, so the name must be self-contained. A wizard primary that always says Next parks situation on a step indicator; if that indicator is invisible, the sentence is incomplete—use Next: shipping. Timed Yes / No in a game sometimes wants friction; a complete sentence is not the goal.

Applying it

• Write the dialog as a spoken question first, then let the buttons be its predicates; do not drop Yes / No and back-fill a title.
• Preview the narrow viewport: if the title is clipped, can the buttons alone still name the event? If not, put the verb back on the button.
• Do not answer with Yes / No a question that a custom header has already replaced.
• Verify by showing the button strip alone and the title alone. If the two halves do not name the same event, or either half is unintelligible while the other is off-screen, the sentence has not formed.

Related

• Within the group: E1.06.1 A button label should name the action that will happen, not a generic word · E1.06.3 Length is constrained by layout but must not give up clarity
• Adjacent: T2 Situated copy · E4.10 Modal dialogs · E6.05 Confirmation dialogs
• Search terms: sentence-style dialog · button in context · Yes/No labels