We submitted 497 US and 373 UK digital twins to an identical battery of 15 mildly provocative everyday scenarios and analysed the in-voice first-person text they produced (≈754,000 words), holding stimulus, wording and response format constant across markets. Beyond surface dialect (spelling, slang), the two corpora diverge systematically along the dimensions cross-cultural pragmatics would predict. American twins favour an overt-heat emotional lexicon (mad, pissed, irritated), colourful boosters (freaking, as heck) and more direct complaint acts (snap, say something). British twins favour understatement (a bit annoyed), measured affect terms (annoyed, frustrated), de-escalation formulae (get on with it), conventionalised complaint (complain, have a word, kick off) and a markedly higher rate of reflexive apology (sorry 5.5% vs 2.6%). Crucially, both varieties hedge at near-identical density — but through different devices, consistent with Leech's (1983) tact strategies. The lexical divergence tracks the independently-measured behavioural divergence (UK twins "make a scene" at 5.1% vs the US 7.6%), suggesting the twins reproduce not only what their human counterparts say but the culturally-patterned relationship between feeling and its expression.
Cross-cultural pragmatics studies how the "same" communicative act is realised differently across speech communities. Three strands of that literature frame this study.
Face and politeness. Brown & Levinson (1987) model politeness as the management of face: positive face (the desire for approval and solidarity) and negative face (the desire to be unimpeded). Although Brown & Levinson claim universality and do not themselves rank nations, a substantial secondary literature associates Anglo-American interaction with positive-politeness / solidarity strategies — informality, expressiveness, camaraderie — and British interaction with negative-politeness strategies of non-imposition, indirectness and reserve (Wierzbicka, 2003; Sifianou, 1992; cf. Scollon & Scollon, 2001 on involvement vs. independence).1
Tact and understatement. Leech's (1983, 2014) Politeness Principle adds to Grice a Tact maxim ("minimize cost to other"). Indirectness, hedging and understatement are the strategies that serve it. Wierzbicka (2003, 2006) reads the Anglo aversion to imposing on another's autonomy as a cultural script realised grammatically — interrogative "whimperatives," downtoners, and a dispreference for bald assertions of feeling.
Realisation and modification. The Cross-Cultural Speech Act Realization Project (Blum-Kulka, House & Kasper, 1989) supplies the empirical apparatus: speech acts arrayed on a directness scale, and modulated internally by downgraders (mitigators, hedges — "perhaps," "a bit") versus upgraders / boosters (intensifiers). Holmes (1990, 1995) and Aijmer (2002) treat hedges and boosters as simultaneously epistemic and affective politeness devices. Finally, Fox (2004) gives the ethnographic prediction most directly testable here: the English reflex apology — saying "sorry" when not at fault — and a pervasive norm of understatement.
Each of these makes a falsifiable prediction about how US and UK speakers should narrate a small injustice. We test them on a corpus that, by construction, holds the provocation constant.
Corpus. Each digital twin is modelled on a real individual in its market. We fielded 15 self-contained scenarios (a slow waiter, a card declined in a queue, a third wrong order, a cracked phone atop varying "carried load," etc.) on two response channels — a felt-intensity rating and a free-text account of what the person would feel and do. We analyse only the in-voice qualitative text — the participant speaking in the first person — and explicitly exclude the model's analytic "reasoning" field. The result is 26,098 responses / ≈754k words (US mean 29.5 words/response; UK 28.2).
Design control. Because stimulus text, answer options and the 0–5 action scale are identical across markets, any systematic lexical or pragmatic divergence is attributable to the speaker population, not the prompt — a matched-guise property rarely available in naturalistic cross-cultural corpora.
Feature coding. We operationalised the constructs above as regex-matched feature families: downgraders/hedges, upgraders/boosters, an overt-anger ("heat") lexicon, a restraint/de-escalation lexicon, apology & politeness formulae, complaint-directness markers, and internalisation markers. For each feature we report the % of responses containing it and the rate per 1,000 words (length-controlled), with a two-proportion z-test on the response-level rate.
A note on significance. At n ≈ 13,000 per cell, even trivial differences reach p < .001. We therefore foreground effect size (ratios between markets) and treat the z-tests only as a floor. "ns" below means a contrast failed even that floor.
The sharpest divergence is in which words name the feeling. American twins reach for the hot, colloquial register; British twins for measured, almost clinical terms — or for vivid words used in a contained frame.
| Affect term | US % | UK % | US /1kw | UK /1kw | lean |
|---|---|---|---|---|---|
| mad (= angry) | 9.83 | 0.73 | 3.3 | 0.3 | US ×13 |
| irritated | 14.10 | 2.07 | 4.8 | 0.7 | US ×7 |
| pissed | 5.18 | 3.28 | 1.8 | 1.2 | US |
| annoyed | 33.54 | 43.10 | 11.4 | 15.3 | UK |
| angry | 2.88 | 7.23 | 1.0 | 2.6 | UK ×2.5 |
| frustrated | 1.24 | 2.94 | 0.4 | 1.0 | UK |
| fuming / livid | 0.07 | 2.06 | 0.0 | 0.7 | UK ×29 |
| Overt-heat cluster (any) | 20.10 | 10.23 | 7.2 | 3.8 | US ×2 |
All contrasts p < .01 except where noted. "Heat cluster" = mad / pissed / furious / raging / livid / fuming / seething / ticked off.
Americans are twice as likely to deploy an overt-heat term outright (20.1% vs 10.2%). The British high scorers are instructive: annoyed (the maximally understated option), angry and frustrated (measured, almost reportorial), and the idiomatic fuming/livid — vivid, but, as §3.2 shows, almost always wrapped in a downtoner ("a bit fuming"). The American mad/irritated pairing has no British equivalent.
The most theoretically interesting result is a null at the aggregate level. Hedge density is statistically indistinguishable — both populations mitigate almost everything they say (composite ≈ 3.2 hedges/response; ≥1 hedge in >98% of responses). What differs is the device.
| Downgrader | US % | UK % | lean & reading |
|---|---|---|---|
| a bit / a little / a touch | 23.82 | 41.97 | UK ×1.8 understatement (Leech tact; Fox) |
| kind of / sort of / kinda | 20.79 | 4.84 | US ×4.3 approximation |
| probably | 84.62 | 65.05 | US epistemic distancing |
| maybe | 54.93 | 64.53 | UK epistemic distancing |
| Any hedge | 99.91 | 98.89 | parity — both mitigate near-universally |
British mitigation runs through understatement of degree — the canonical "a bit annoyed" — exactly the device Leech files under the Tact maxim and Fox calls a defining English reflex. American mitigation runs through approximation and epistemic distance — "kind of pissed," "I'd probably snap." Both soften; only the British strategy systematically shrinks the reported magnitude of the emotion itself.
Overall booster frequency is again near-parity (≈48–50% of responses), but the expressive, slang intensifiers are distinctively American, consistent with a positive-politeness / high-involvement style (Tannen, 1984; Holmes, 1990).
| Booster | US % | UK % | lean |
|---|---|---|---|
| honestly / literally / seriously | 3.15 | 0.97 | US ×3.2 |
| super / freaking / frickin | 0.62 | 0.13 | US ×4.8 |
| as heck / as hell | 0.43 | 0.01 | US ×43 |
Fox's (2004) ethnographic claim is borne out quantitatively. UK twins apologise 2.1× as often as US twins — in scenarios where, as the aggrieved party, they have nothing to apologise for — and use bare politeness formulae more.
| Marker | US % | UK % | lean |
|---|---|---|---|
| sorry | 2.55 | 5.45 | UK ×2.1 |
| please | 0.15 | 0.43 | UK ×2.9 |
| polite(ly) | 1.83 | 2.48 | UK |
Placed on the CCSARP directness scale, the two populations realise the same grievance differently. Americans favour direct, on-record confrontation; Britons favour conventionalised or idiomatic realisations that route the complaint through a recognised social form.
| Realisation | US % | UK % | directness / lean |
|---|---|---|---|
| snap / sharp word / short with | 6.17 | 2.51 | US ×2.5 direct, on-record |
| say something | 1.83 | 0.97 | US direct |
| complain | 0.42 | 1.20 | UK ×2.9 conventionalised |
| kick off (make a fuss) | 0.00 | 3.59 | UK only idiomatic |
The mirror-image idioms are telling: where an American "says something" or "snaps," a Briton "has a word," "complains," or — the line they most often refuse to cross — "kicks off." The British de-escalation lexicon (get on with it 4.7% vs 0.1%; bite my tongue 1.5% vs 0.6%; calm 18.5% vs 13.9%) is correspondingly richer.
Combining §3.1 and §3.5, we compute the ratio of restraint-cluster to heat-cluster mentions in each market — a single index of how much each variety frames a provocation in terms of containing a feeling versus naming it hot.
The lexical gap mirrors the behavioural gap measured independently in the same study: UK twins "make a scene" in 5.1% of provocations versus the US 7.6%. The variety that talks with more restraint also acts with more restraint. The twins reproduce not just national vocabulary but the culturally-patterned relationship between feeling and its expression — the object cross-cultural pragmatics actually studies.
The results align, point for point, with the predictions of the politeness literature. American twins instantiate a positive-politeness / high-involvement profile (Tannen, 1984; Scollon & Scollon, 2001): overt affect, expressive boosters, direct on-record complaint. British twins instantiate a negative-politeness / non-imposition profile (Wierzbicka, 2003; Sifianou, 1992; Fox, 2004): degree-understatement, measured affect terms, conventionalised complaint, and the reflex apology. The §3.2 finding — equal hedge density, opposite device — is the most precise confirmation of Leech's (1983) framing: tact is a universal pressure, but each variety discharges it through a different, culturally-preferred strategy.
For population-scale simulation the implication is methodological. A model that merely swapped spellings would be a costume; one that reproduces the device-level machinery of politeness — and ties it to congruent behaviour — is reproducing the pragmatic competence of the modelled population. That is the fidelity claim worth making, and it is independently checkable, as here.
(i) Synthetic corpus. These are digital twins, not transcribed humans; the study evidences fidelity-to-population, not a fresh sociolinguistic survey of Britons and Americans. (ii) Polysemy. Regex matching cannot fully disambiguate (e.g. British mad = "crazy"); spot-checking indicates the angry sense dominates, and the heat-cluster result is robust to dropping mad entirely. (iii) Instrument-supplied tokens. The phrase "make a scene" and the 0–5 action labels were given in the prompt, so their near-parity is expected and excluded from the directness argument. (iv) Effect size over p. With n ≈ 13k, significance is uninformative; we have reported ratios throughout. (v) Sample. The UK panel is smaller and younger; the restraint pattern nonetheless holds within every generation.
Aijmer, K. (2002). English Discourse Particles: Evidence from a Corpus. Amsterdam: John Benjamins.
Blum-Kulka, S., House, J., & Kasper, G. (Eds.). (1989). Cross-Cultural Pragmatics: Requests and Apologies. Norwood, NJ: Ablex.
Blum-Kulka, S., & Olshtain, E. (1984). Requests and apologies: A cross-cultural study of speech act realization patterns (CCSARP). Applied Linguistics, 5(3), 196–213.
Brown, P., & Levinson, S. C. (1987). Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press.
Fox, K. (2004). Watching the English: The Hidden Rules of English Behaviour. London: Hodder & Stoughton.
Holmes, J. (1990). Hedges and boosters in women's and men's speech. Language & Communication, 10(3), 185–205.
Holmes, J. (1995). Women, Men and Politeness. London: Longman.
Larina, T. (2015). Culture-specific communicative styles as a framework for interpreting linguistic and cultural idiosyncrasies. International Review of Pragmatics, 7(2), 195–215.
Leech, G. N. (1983). Principles of Pragmatics. London: Longman.
Leech, G. N. (2014). The Pragmatics of Politeness. Oxford: Oxford University Press.
Scollon, R., & Scollon, S. W. (2001). Intercultural Communication: A Discourse Approach (2nd ed.). Oxford: Blackwell.
Sifianou, M. (1992). Politeness Phenomena in England and Greece: A Cross-Cultural Perspective. Oxford: Clarendon Press.
Tannen, D. (1984). Conversational Style: Analyzing Talk Among Friends. Norwood, NJ: Ablex.
Wierzbicka, A. (2003). Cross-Cultural Pragmatics: The Semantics of Human Interaction (2nd ed.). Berlin: Mouton de Gruyter.
Wierzbicka, A. (2006). English: Meaning and Culture. Oxford: Oxford University Press.
1 The positive/negative-politeness ranking of national varieties is not claimed by Brown & Levinson, who argue for universality; it is the standard secondary reading associated with the authors cited. We follow that reading while flagging its status.