HookForty-six of forty-eight interruptions came from men — then the finding fell apart
In 1975 Don Zimmerman and Candace West recorded thirty-one segments of everyday conversation around a California university campus. In the mixed-sex pairs, men produced 46 of the 48 interruptions — 96%. It became one of the most quoted numbers in sociolinguistics: proof, apparently, that men use talk to dominate women.
Then Geoffrey Beattie recorded roughly ten hours of university tutorials — 557 interruptions, more than ten times the original evidence base — and found men and women interrupting at almost identical rates. Neither study settles the question: 48 interruptions is a tiny sample, tutorials are not casual chat, and whether an overlap counts as an 'interruption' or a supportive nudge is a judgement call baked into the coding. That tension is Paper 2 Section A in miniature. The exam does not want you to recite who found what. It wants you to hold a theory in one hand and its limitations in the other, and weigh them — every AO2 mark lives in that gap.
ModelLanguage and gender — four lenses on the same data
Robin Lakoff's Language and Woman's Place (1975) launched the deficit position: women's talk, she claimed, was marked by hedges ('sort of'), tag questions ('isn't it?'), 'empty' adjectives ('divine'), precise colour terms and rising intonation on declaratives — a style learned in childhood that made women sound unsure of their own claims. Crucially, Lakoff introspected: she counted nothing.
The dominance position read gendered talk as power in action: Zimmerman and West's interruption figures, and Pamela Fishman's analysis of 52 hours of couples' home recordings, in which the women asked around two and a half times as many questions as the men — doing the routine maintenance work that kept conversation alive at all. The difference position (Deborah Tannen, 1990) reframed the same behaviours as two subcultures: rapport-talk versus report-talk, learned in single-sex childhood peer groups — different, not deficient.
The current diversity (dynamic) position dismantles the question itself. William O'Barr and Bowman Atkins tracked Lakoff's features across 150 hours of courtroom speech and found they clustered by status, not sex: expert witnesses of both sexes dropped them, inexperienced witnesses of both sexes used them. They renamed the style powerless language. Janet Hyde's 2005 meta-analysis found around 78% of measured gender differences were small or near zero, and Deborah Cameron's The Myth of Mars and Venus (2007) argues that the differences we 'hear' are largely stereotype steering perception. Representation matters alongside use: lexical asymmetry (master/mistress, bachelor/spinster) and marked forms (actress, 'lady doctor') encode expectations before anyone opens their mouth.
ModelLanguage and power — who may say what to whom
Shân Wareing's split is the starting frame: instrumental power is backed by sanction — a judge, a head teacher, a contract; comply or face consequences — while influential power has to be earned through persuasion: advertising, politics, journalism. Norman Fairclough's Language and Power (1989) supplies the machinery. Power in discourse is the visible asymmetry of unequal encounters; power behind discourse is the prior question of who set the conventions and whose English counts as 'professional' in the first place. His term synthetic personalisation names the mass-media trick of addressing millions as 'you', one reader at a time.
Asymmetry is measurable, which makes it analysable. In unequal encounters the powerful participant allocates turns, opens and closes topics, interrupts without penalty, asks the questions rather than answering them, and can issue bald imperatives. Politeness theory explains the softening everyone else performs: Brown and Levinson model talk as the management of face — positive face (the wish to be approved of) and negative face (the wish not to be imposed on). A nurse's 'just pop your sleeve up for me' wraps an imperative in minimisers because even instrumental power usually dresses itself in mitigation. The sharp analytical question in any data is: who is doing the face-work, and for whose benefit?
MechanismOccupational English — precision, belonging and gatekeeping
Paul Drew and John Heritage (1992) identified what makes institutional talk different from conversation: it is goal-oriented, its turn-taking is constrained (in court, witnesses answer and never ask), and professionals share inferential frameworks — insider ways of hearing, so 'chest pain, radiating left' is a diagnosis-in-progress to a paramedic and merely a description to everyone else. Jargon therefore leads a double life. Inside the group it compresses meaning with surgical precision; at the boundary it excludes — sometimes accidentally, sometimes as deliberate gatekeeping.
Two concepts organise almost any occupational-language answer. John Swales's discourse community: a group with shared goals, mechanisms of intercommunication, its own genres and lexis, and a threshold of expert members — a hospital ward, a kitchen brigade and a subreddit's moderating team all qualify. Étienne Wenger's community of practice shifts the focus to learning-by-doing: mutual engagement, a joint enterprise and a shared repertoire that newcomers acquire on the job — the shorthand, the running jokes, the email sign-offs. Almut Koester's workplace research adds the humane detail examiners like: phatic, off-task talk is not wasted time, because the relationships it builds are what task-talk runs on.
DataClass, age and the sociolinguistic variable
William Labov invented the modern method in New York in 1966. He asked staff in three department stores a question whose answer had to be 'fourth floor', then noted whether they pronounced the /r/ — a prestige feature in New York English. At upmarket Saks, 62% used some rhotic /r/; at mid-market Macy's, 51%; at discount S. Klein, 20%. Pronunciation stratified precisely by the store's social position — and rates rose when he pretended not to hear and staff repeated themselves in careful speech.
Peter Trudgill ran the British equivalent in Norwich (1974) on walking versus walkin': the non-standard [n] was near-universal among working-class men in casual speech — and in self-report interviews women claimed more standard usage than they actually produced, while men claimed less. That over-claiming of non-standardness is covert prestige: vernacular forms carry status inside the group precisely because they signal toughness and solidarity rather than institutional approval, the mirror image of the overt prestige attached to Standard English.
Age works the same way once you stop treating it as a birthday. Penelope Eckert's Detroit-area high-school study found language tracked social practice — school-oriented 'jocks' versus locally oriented 'burnouts' — better than raw age or parental class. Jenny Cheshire's Reading playground research found teenagers' non-standard grammar rose with their adherence to peer 'vernacular culture'. Lesley Milroy's Belfast work supplies the mechanism: dense, multiplex social networks — where your neighbour is also your workmate and your cousin — police vernacular norms. Your idiolect is the fingerprint left by every network you have passed through.
CaseEthnicity and Multicultural London English
Between 2004 and 2010, ESRC-funded projects led by Jenny Cheshire and Paul Kerswill recorded young speakers in inner-London Hackney and outer-London Havering and documented a new variety: Multicultural London English (MLE). It emerged in high-diversity friendship groups where children acquired English largely from each other rather than from a single local adult model, blending Caribbean, African, South Asian and traditional London influences. Signature features: the pronoun man ('man's hungry'), the quotative this is + speaker ('this is me: no way'), the intensifier bare, the invariant tag innit, and th-stopping ('dis', 'dat').
The crucial exam point: MLE is a multiethnolect. It is spoken across ethnic boundaries — white, Black and Asian Londoners in mixed networks all use it — so the tabloid label 'Jafaican', with its implication of fake Jamaican, gets the research exactly wrong. Related concepts earn AO2 depth: code-switching (Mark Sebba's work on London Jamaican shows speakers moving between varieties within a conversation to do identity work) and Ben Rampton's crossing — using a variety you have no ethnic 'claim' to, which can read as affiliation or as appropriation depending on who is doing it, to whom, and where. Attitudes lag the linguistics: the Accent Bias in Britain project (2019–20) found MLE ranked near the bottom of accent-prestige ratings even when listeners heard identical scripted content — variety difference misread as competence difference.
MechanismOne transcript, three readings — building the AO2 paragraph
The Section A essay hands you a statement to evaluate, not a topic to describe. The working method: fix your line of argument early, then make each paragraph a claim tested against evidence — a named study with its finding, its limitation, and a rival reading. The transcript below (invented, simplified) shows why the same forty words of data can serve three different theories — which is exactly the flexibility the top band rewards.
A team meeting, original invented data. RAV (chair): "right — actions from Monday (.) Priya, the venue" / PRIYA: "yeah so I rang them and I think we're [nearly—" / RAV: "[did you get a price" / PRIYA: "er (.) not a final one? they said Friday (.) sorry" / RAV: "fine. Jordan — numbers." Reading one, dominance: Rav interrupts at the overlap, allocates every turn by name and issues verbless directives ('Jordan — numbers') — power in discourse, textbook Fairclough. Reading two, deficit: Priya hedges ('I think'), uses a rising declarative ('not a final one?') and apologises — Lakoff's features on cue. Reading three, dynamic: swap the names and nothing changes. O'Barr and Atkins would call Priya's features powerless rather than female — she is performing deference to the chair role, and Rav's bluntness belongs to the chair, not to men. A paragraph that runs all three readings and then adjudicates — on sample, on context, on what the features are actually doing — is a level-5 AO2 paragraph.
VocabularyKey terms the mark scheme pays for
TrapsMisconceptions that cost marks
ExamWhat examiners want
The Section A essay is worth 30 marks: 20 for AO2 (concepts and issues, evaluated) and 10 for AO1 (terminology and coherent expression). The stem is always an idea to be weighed — practise against original stems like 'Evaluate the idea that differences between men's and women's language have been exaggerated' — so state a position in your opening paragraph and test it, rather than touring theories in date order.
Make the unit of evidence name, date, finding, limitation: 'Fishman (1983) — women asked around two and a half times as many questions — but three couples, one culture, one decade' does more AO2 work than a paragraph of summary. Numbers lift bands: 46 of 48, 62% at Saks, 78% of effect sizes near zero. For AO1, precision pays — variable, covert prestige, multiethnolect, face-threatening act — and so does defining your key term in the first paragraph.
The recurring examiner complaints are a checklist of what not to do: theory-dumping without evaluation stalls in the middle bands; treating a 1975 introspective claim as settled fact signals no critical distance; and essays that never return to the wording of the question lose the argument thread AO1 rewards. Around 45 minutes is the right budget — plan for five testing paragraphs, not ten listing ones.