You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Understanding what a top-band investigation actually looks like is one of the fastest ways to raise your own. This lesson works through annotated structural exemplars across three popular topic areas — language and gender, child language acquisition, and language change — and then teaches you the skill that separates good investigations from very good ones: evaluation. Because the investigation is assessed mainly on AO1, AO2 and AO3, the exemplars below are read against those objectives throughout, and the closing sections show you how to evaluate your own findings honestly and connect them to wider debates without over-claiming.
Across topics, moderators reward the same underlying qualities. The table contrasts top-band work with competent-but-limited mid-band work so you can locate your own draft.
| Quality | Top band | Mid band |
|---|---|---|
| Research question | Precise, focused, linguistically motivated | Broad, vague or merely "interesting" |
| Data | Selected with a clear rationale; well presented | Adequate but sometimes too much, too little, or poorly chosen |
| Analysis | Multi-level, interpretive, theory-informed | Single-level (often lexis only), descriptive |
| Theory (AO2) | Integrated and accurately attributed; evaluated | Bolted on at the end; used superficially or mis-attributed |
| Evaluation | Honest; weighs limitations and alternatives | Brief or absent |
| Writing (AO1) | Clear, academic, accurate terminology | Generally competent, with slips in register or terms |
Key Definition: Evaluation — the critical assessment of your own investigation: its strengths and limitations, the reliability and validity of its findings, alternative interpretations of the data, and how the conclusions relate to wider linguistic debates. Evaluation is woven through the analysis as well as gathered in the conclusion.
Title: "How do male and female participants use hedging when disagreeing in a small mixed-gender study group?"
Introduction (~250 words). States the research question; defines hedging with examples; sets up the relevant frame — Lakoff (1975) on women's language, Holmes (1995) on hedging as relational, Cameron (2007) on gender as performance; explains why a study group is a productive context (informal, mixed-gender, with low-stakes disagreement).
Methodology (~350 words). Audio-recorded one 45-minute group of four students (two male, two female); transcribed the 15 most interactionally rich minutes; defined hedge categories (epistemic modals, approximators, shields) and applied them consistently; secured consent, anonymised speakers, acknowledged the observer's paradox and the limits of a single group.
Analysis (~1,000 words). Lexis: female speakers hedged more overall (7.8 vs 4.2 per 100 words), but the types differed — women favoured epistemic modals ("might," "could"), men favoured approximators ("sort of"). Pragmatics: many female hedges prefaced disagreement and functioned relationally (inviting others in, softening face-threat) rather than marking uncertainty — supporting Holmes over a bare deficit reading. Discourse: hedges clustered at the onset of disagreements for both genders, suggesting a shared politeness function. Throughout, theory is integrated and each claim is tied to a specific extract.
Conclusion (~250 words). Hypothesis partially supported: a gender difference in frequency exists, but its interpretation is relational, not deficit-driven. Limitations (one small group, single context, observer's paradox) are weighed, and same-gender comparison is proposed as further research.
Examiner-style commentary (why this is Top-band): The strength is that the analysis never stops at counting. The frequency gap (AO1) is immediately pushed into a functional, contextual reading (AO3) and tested against accurately attributed competing theories (AO2: Lakoff vs Holmes vs Cameron). The hypothesis is reported as partially supported — the honesty that signals genuine evaluation rather than a forced result. A Mid-band treatment of the same data would likely stop at "women hedge more, which supports Lakoff," conflating frequency with deficit and missing the relational function entirely.
Title: "Overgeneralisation of past-tense morphology in one child aged 2;6 to 3;0"
Introduction (~250 words). Focuses on overgeneralised past tense ("goed," "runned"); outlines the U-shaped pattern (correct irregulars → overgeneralised forms → correct again); frames the analysis with Berko's "wug" test (productive morphology) and Brown's order of morpheme acquisition, and notes the innate-versus-usage-based debate (Chomsky's nativism set against usage-based, frequency-sensitive accounts).
Methodology (~300 words). Recorded the student's younger sibling across play, mealtimes and story time over six weeks; transcribed six five-minute extracts; obtained parental consent and used a pseudonym; coded every past-tense form as correct-regular, correct-irregular, overgeneralised or other, with fixed definitions.
Analysis (~1,100 words). Morphology: overgeneralisation rose from roughly 10% of irregular contexts in Week 1 to about 45% by Week 4, then dipped — the predicted U-shape. Lexis: lower-frequency irregulars were overgeneralised first ("dig"→"digged"), while very high-frequency forms ("went," "was") resisted — pointing to a frequency effect. Syntax/pragmatics: past-tense forms clustered in narrating and imaginative play. The discussion holds two ideas in tension: the productive, rule-like overgeneralisation supports an emergent internal rule, while the frequency effect supports usage-based learning — without pretending one "wins."
Conclusion (~200 words). Summarises the trajectory, judges the evidence genuinely mixed between accounts, and flags limitations (one child, six weeks, few contexts), proposing a longer longitudinal design.
Examiner-style commentary (why this is Top-band): The investigation earns AO2 by refusing a tidy verdict — it lets the data support both a rule-based and a usage-based reading and says so, which is exactly the conceptual maturity moderators reward. The fixed coding scheme protects reliability (AO1), and the frequency finding is interpreted, not just reported (AO3). A weaker version would announce that the child "proves Chomsky right," over-claiming from one short case and ignoring the frequency pattern that complicates the story.
Title: "How has the lexis of recipe introductions changed between 1960 and 2020?"
Introduction (~250 words). Frames the recipe as a functional genre with stable conventions but shifting culture; sets the research question about lexical change and what it reveals about attitudes to food and cooking.
Methodology (~350 words). Built a small corpus of 30 recipe introductions — ten each from the 1960s, 1990s and 2020s — controlling for dish type (main courses) and source (cookbooks for home cooks); used AntConc for frequency lists and keywords; states the sampling rationale.
Analysis (~1,050 words). Semantic fields: 1960s lexis clusters around instruction and thrift ("economical," "nourishing"); 2020s around pleasure, authenticity and wellness ("indulgent," "authentic," "vibrant"). Borrowing: a marked rise in loanwords (Italian, Japanese), reflecting globalisation. Adjectives: practical in the 1960s ("quick," "cheap"), evaluative and sensory by the 2020s ("smoky," "stunning"). Address: a shift from impersonal instruction ("one should") to direct second-person address ("you'll love this"), read through Fairclough's synthetic personalisation. Changes are connected cautiously to social trends rather than asserted as their proof.
Conclusion (~200 words). Summarises the lexical drift, evaluates the corpus's limits (small, cookbook-only), and proposes extending to online recipes.
Examiner-style commentary (why this is Top-band): Diachronic studies tempt students into sweeping cultural claims; this one stays disciplined, treating the corpus as suggestive and attributing the address-term shift to a named, correctly applied concept (Fairclough). The controlled sampling (dish type and source held constant) is what licenses the comparison at all — a point a moderator notices immediately. A Mid-band version would list "old vs new words" without the controlled design, the synthetic-personalisation lens, or the caution about a 30-text corpus.
Title: "How does a barista construct service and authority through language across ten customer interactions in one coffee shop?"
Introduction (~250 words). Frames the service encounter as a site where occupational identity and institutional power are performed through routine talk; states the research question; introduces the relevant frame — Drew and Heritage's notion of institutional talk (with its goal-orientation and asymmetries), Fairclough's synthetic personalisation (addressing a mass audience as if individually), and Brown and Levinson's politeness theory for the face-work of service.
Methodology (~350 words). Audio-recorded, with the consent of the employee and the franchise manager and with all customers anonymised, ten complete service interactions during one shift; transcribed all ten with a light convention set; coded each turn for directives, interrogatives, formulaic service phrases ("can I get you anything else?") and personalised address (first-name use, "for you"). Acknowledges the observer's paradox and the single-site limitation.
Analysis (~1,000 words). Discourse: the encounters share a tight, recurring structure (greeting → order → upsell prompt → payment → closing), evidencing the goal-oriented routine of institutional talk. Lexis/pragmatics: heavy use of formulaic, personalised phrasing ("your usual?", "enjoy") manufactures rapport efficiently — Fairclough's synthetic personalisation realised in micro-form. Grammar: the barista's directives are almost always mitigated as interrogatives ("would you like…?") rather than bald imperatives, preserving customer face while still steering the transaction; the customer's requests, by contrast, are often bald ("flat white"), a reversal that reveals where institutional service places the face-burden. Variation: personalised address rises with regular customers and falls with one-off visitors, suggesting rapport is calibrated, not automatic. Theory is integrated and each claim quoted from the data.
Conclusion (~200 words). Service authority here is exercised through routine and mitigation rather than overt command; the rapport is real but systematically produced. Limitations (one employee, one site, one shift, the observer's paradox) are weighed, and comparison across different service settings is proposed.
Examiner-style commentary (why this is Top-band): The investigation reads occupational language as power exercised softly — through structure and mitigation — rather than reaching for the cruder "boss tells worker what to do" model, and that nuance is what evidences conceptual understanding (AO2). The reversal it spots (mitigated staff directives, bald customer requests) is a genuine finding, not a description, and it is interpreted contextually (AO3). The controlled coding and quoted examples secure AO1, and the calibration of rapport to customer type shows the analyst noticing variation within the data rather than flattening it. A Mid-band version would simply list polite phrases and conclude "the barista is friendly," missing the systematic face-work and the customer-side asymmetry entirely.
Exemplars teach as much through their failure modes as their successes. The patterns below recur in lower-scoring investigations; learning to recognise them in your own draft is the fastest route to improvement. For each, the table gives the symptom, the underlying cause, and the remedy.
| Weakness | What it looks like | Underlying cause | Remedy |
|---|---|---|---|
| No genuine research question | A vague title ("Language and gender") with no question to answer | Topic chosen before it was narrowed | Reframe as a how/which/to-what-extent question with a defined context and variable |
| Feature-spotting | A list of devices ("there is alliteration, a simile, a question") with no functions | Stopping at rung 2 of the analysis ladder | Add "which functions to…" to every observation; reach rungs 3–4 |
| Single-level analysis | Everything is lexis; grammar, pragmatics and discourse are absent | Comfort with vocabulary, unfamiliarity with other levels | Deliberately work the same data through three levels |
| Theory bolted on | Theorists named only in a final paragraph, or mis-attributed | Reading treated as decoration, not as a lens | Integrate concepts as you analyse; verify every attribution |
| Over-claiming | "This proves teenagers can't spell" from twenty texts | Confusing a small sample with a population | Use tentative verbs; state the sample's limits |
| Data drowning the analysis | Pages of transcript, a paragraph of comment | Failure to separate "report" from "data" | Quote only discussed extracts; appendix the rest |
| Absent evaluation | The conclusion restates findings as certainties | Treating evaluation as optional | Weave limitations and alternatives through the analysis and conclusion |
There is a thread running through every row: the weak investigation describes, the strong one interprets and evaluates. A moderator can usually tell within a page which kind they are reading, because the strong investigation keeps answering the question "so what does this mean, and how confident can I be?" while the weak one keeps stopping at "here is a thing I noticed."
Coursework Tip: Before you submit, run a deliberate self-audit against this table, row by row. For each weakness, find the strongest counter-example in your own draft — a sentence that reaches rung 4, a moment of genuine multi-level analysis, an honestly stated limitation. If you cannot find one for a given row, you have just identified your highest-value revision.
Evaluation is not an optional flourish; it is integral, and the strongest folders evaluate throughout, not only at the end. Interrogate your own work with these questions.
Key Definition: Triangulation — strengthening validity by combining multiple methods or data sources, e.g. pairing discourse analysis of recorded talk with questionnaire data on the same participants' attitudes, so the two readings can corroborate or qualify each other.
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.