--- name: thesis-defense-viva description: "Use when preparing for the viva voce / oral thesis defense — building the examiner's question set, rehearsing the presentation, and the pre-defense polish (slides, summary, the 'what if I'm stuck' responses) for a life-science or bioinformatics PhD." license: MIT --- # Thesis Defense & Viva Voce (Life Sciences / Bioinformatics) The viva is not a fact recitation; it is an examiner's structured interview about **whether the work is yours, whether it is sound, and whether you know its edges.** Preparation is question-driven: you rehearse the specific probes an examiner will actually ask, not a generic "good luck." This skill builds the question set from your own thesis, scripts the presentation, and defines how to answer when you are genuinely stuck. ## Hard rules 1. **No fabricated citations.** Anything you cite under questioning (a prior result, a protocol, a benchmark) must resolve to a record you can defend. Do not bluff a reference number; an examiner will catch a wrong citation faster than any other slip, and it reads as integrity, not memory. 2. **No claim without provenance.** If you assert a number or a conclusion in the defense, be able to point to the figure/table/page that supports it. "The result is in Table 4.2" is the habit to train; "I think it was about 30%" is the habit that sinks you. 3. **No silent failure.** If a probe touches a method or a control you have not fully run, do not gloss over it. Name the gap and its status ("I have not run X; here is what I would do and what I expect") — an admitted, bounded gap is scored far better than a confident dodge. 4. **Answer the question asked, not the one you wanted.** A 30-second direct answer, then elaboration only if invited. Rummaging through the thesis while you think is the single most common live mistake. ## When to use - Rehearsing the oral defense / viva voce. - Building the examiner question set a week or two before the defense. - A final "what will I be probed on?" walk-through of your own chapters. ## When NOT to use - Fixing the underlying reproducibility gaps the exam would expose — that is `thesis-methods-reproducibility`, run it first. - The written thesis itself — `thesis-structure`. - Rehearsing a *conference* talk or a *grant* talk (different stakes, different length) — use `comm-conference-talk` or `grant-proposal-craft`; a defense is 10–20 min of talk + 1–3 h of questioning. ## Prerequisites - The thesis draft is essentially frozen — you can answer "what's on page N?" quickly because you know your own document. - Your supervisor or a mock examiner to run at least one full dry run. ## Build the question set from your thesis Generate the probes your *own* document invites, in the examiner's order of attack. This is the highest-leverage prep: it is the exact list they will walk. 1. **The framing questions** (first, to place you): - What is the one contribution, in a sentence? - Why does this matter, and why was it hard before your work? - What did *you* do versus the group / the tool / the collaborators? 2. **The soundness questions** (the core): - For each major result: what was the control, and why does it support the claim? - What alternative explanation did you rule out, and how? - What is the weakest assumption in your method, and what would break it? 3. **The edge-of-work questions** (the stress test): - What if the sample size / the cutoff / the normalization were different? - Where does your result *not* hold? (Examiners reward a stated boundary.) - What would you do with 6 more months, and what would you do first? 4. **The provenance questions**: - Trace one number in your results to its source file / figure / parameter. - If a method "standard protocol" was used, name the version and the parameter. Rehearse each until the direct answer is under 30 seconds; only then add depth. ## The presentation (10–20 min) - It is the **thesis-scale** summary, not a chapter read: contribution → the one experiment / result that carries it → the boundary. Mirror `thesis-structure` in miniature. - Slides are the visual aid, not the script: figure-forward, minimal text, every panel labeled. The defense slides are a *separate* deck from the written thesis figure set. - Open with the contribution, not the background — the examiner knows the background; they need your result and your claim. ## When you are genuinely stuck - **Don't bluff.** "I'm not certain; here is how I would find out" is the correct move and is scored above a confident wrong answer. - **Buy time, don't freeze.** Repeat the question in your own words to confirm it; a 5-second rephrase is normal and reads as engagement, not delay. - **Scope the answer.** If the question is broader than you can answer, answer the part you can and name the part you can't: "On X, my data says …; the Y aspect I did not measure." ## Pitfalls - **Rummaging through the thesis** while composing an answer — pre-locate your key figures; know where every number lives. - **Answering the question you hoped they'd ask** instead of the one they asked. - **Bluffing a citation or a version** — caught fast, and it reads as an integrity problem, not a memory slip. - **No stated boundary** — "my method is robust" with no "except when …" is the tell of someone who has not stress-tested their own work. - **Treating the examiner as hostile** — they are checking your grasp; the tone you keep is part of the score. - **Forgetting the "what would you do differently" question** — the classic final, and one you can prepare a clean, specific answer for. ## Validation - A full mock defense run (at least one), with the question set answered cold for the framing and soundness categories. - Every major result has a rehearsed control-justification in under 30 seconds. - One live trace from a results number to its source figure/file, done aloud. - A written "stuck" script: for each known-weak area, the bounded-answer line you will actually say. ## References - `thesis-structure` — the miniature the defense slides must mirror. - `thesis-methods-reproducibility` — the provenance habits behind every answer. - Program-specific viva format (panel size, allowed items, time) lives in your department's PhD regulations — the authoritative local source. ## Cross-references - thesis-structure — the architecture you present and defend. - thesis-methods-reproducibility — the provenance discipline the exam probes. - manuscript-structure — for polishing the defense's one-page contribution statement into words an examiner will reuse. (Run the `meta/integrity-audit` skill on the final slides + question-set answers before the defense. Content skills in this repo do not cross-reference `meta/` slugs as bullets, so it is referenced in prose here.) ## Changelog - **1.0.0** (2026-06-18) — Initial release. A question-set builder keyed to your own chapters, a 10–20 min presentation script, the "when you're stuck" responses, and the rehearsal validation, tuned for life-science / bioinformatics defenses.