The two channels should not repeat each other verbatim

What it is

Voice and screen should not ship the same string twice. A car HUD already shows “turn right 200 m”; the speaker then reads those words, occupying the ear so road sound cannot enter, while the eye gains no extra field. A smart display that paints the full TTS as a chat bubble has put one string into two sinks. Verbatim duplication looks like careful dual coding. It is the slow channel replaying what the fast channel already has.

Why it happens

Identical codes cost twice the time and add no second representation. Speech is slow; TTS of on-screen paragraphs serializes what the eye already holds and blocks the ear from the road, a passenger, or the next real conclusion. Complementarity wants different encodings of the same situation: vision for precise fields, speech for gist — or for the one field that should not have to be searched. Verbatim duplication is often lazy implementation: one string piped to two outlets. Listeners learn to ignore one channel; when the two later diverge, the ignored one is actually gone. Overlap on the conclusion phrase is fine. Overlap on the paragraph is not.

Studying it

The same message three ways: A, both channels verbatim; B, gist on speech, fields on screen; C, screen only. Dependent measures: memory for fields that appeared on only one channel, subjective annoyance, eyes-off-road in a driving mock-up. Independent variable: string length. Do not make which channel leads by a few hundred milliseconds the independent variable.

People will say on a survey that they “also like hearing it.” Accuracy often does not go up. Score field recall and annoyance, not “likes having two channels.”

Where it stops holding

A critical alert may be dual-coded on purpose (the same warning written and spoken) for robustness, not density — mark that as an exception. Language learning that writes down what was heard is a pedagogical goal, not product laziness. Verbatim waste is worst when the eyes are on this display. Overlap of a two- or three-word label can pass; paragraph-level overlap cannot.

Applying it

• Do not pipe the same string to TTS and to the card title.
• Speech carries one proposition. The rest stays on screen.
• Diff the TTS script against visible strings. Allowed overlap is the conclusion phrase, not the paragraph.
• How to check: sample live spoken scripts against screenshots. Large red overlapping spans are verbatim duplication. Then ask listeners for fields that were only on screen; if they cannot answer, they shut the visual channel because the ear was already reciting it.

Related

• Same group: M3.05.1 Screen takes lists and detail; voice takes the conclusion · M3.05.3 Eyes-free needs its own design, not a degraded GUI
• Nearby: M3.10 Multimodal complementarity of voice and screen · M3.11 Trimming speech output · M3.03 Reading long lists aloud
• Search terms: do not duplicate voice and screen verbatim · redundant multimodal · complementarity not redundancy