Style-Shifting: Audience Design in Speech

Speech Context

Setting formality 50

Addressee status 50

Social distance 50

Prestige-consciousness 60

Live Speech

—…

P(standard) = 50%P(vernacular) = 50%

Standard rate (last 20)

—

Utterances

How it works

Sociolinguists have long known that a single speaker does not use one fixed way of talking — they shift between a more standard variant and a more vernacular one depending on who they're talking to and where. Labov's "attention to speech" model tied this to formality; Allan Bell's audience design model reframed it as speakers primarily responding to their addressee — accommodating upward toward higher-status or less-familiar listeners.

This simulator scores each simulated utterance with a logistic model over three normalised context cues (each rescaled to −1…+1), scaled by a personal prestige-consciousness trait k (0…1) that sets how strongly this speaker reacts to context at all:

z = k · (1.4·formality + 1.0·status + 0.7·distance)
P(standard) = 1 / (1 + e^(−z))

Setting formality — a courtroom vs. a kitchen table; the strongest driver of style in Labov's original work.
Addressee status — speaking to a boss vs. a peer; speakers accommodate upward (Bell 1984, Giles' Communication Accommodation Theory).
Social distance — a stranger vs. a close friend; more distance favours careful, standard speech.
Prestige-consciousness — an individual trait; at k=0 the speaker ignores context entirely and stays at 50/50.

Every ~1.1s the model draws one simulated word pair (e.g. "going to / gonna") using P(standard) as a weighted coin flip — a real Bernoulli trial, not a scripted animation — and the choice travels from the speaker to the listener as a coloured token, then joins the rolling 20-utterance history strip.

Standard variant Vernacular variant

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.