Thesis defense questions: what committees actually ask, and how to practice

You know your thesis better than anyone in the room. That was never the problem. The problem is retrieving it, out loud, under questioning, while your committee watches. Here's what they'll ask — and how to train for it in the last two to six weeks.

A defense isn't a trivia contest. The committee is checking three things: that you did the work, that you understand where it sits in the field, and that you can reason about it live — including about its flaws. Which means nearly every question you'll face collapses into five archetypes. Prepare those, and you've prepared most of the exam.

The five question archetypes

1. The contribution question

"What is the contribution of this thesis?" "What do we know now that we didn't before you started?" You need one claim, delivered in a few sentences — not a tour of your chapters. If your answer starts with "Well, in chapter two…", you don't have an answer yet; you have a table of contents.

2. The method-defense question

"Why this method and not X?" Why interviews and not a survey, this corpus and not that one, a frequentist analysis and not a Bayesian one. Examiners aren't saying you chose wrong — they're checking whether you chose at all, or just inherited a method from your lab. A strong answer names the alternative, states the tradeoff honestly, and concedes what the alternative would have bought you.

3. The limitations question

"How far does this generalize?" "What result would break your conclusions?" This is where candidates hurt themselves most, because the instinct under stress is to defend everything. Examiners read defensiveness as fragility. A candidate who names their limitations before being cornered sounds like an expert; one who bluffs sounds like someone who got lucky. Know your three biggest limitations cold, and what you'd do differently.

4. The adjacent-field question

Usually from the external examiner or the committee member outside your subfield: "How does this connect to [a neighboring literature]?" "What should practitioners take from it?" It feels unfair — you didn't write about that. That's the point. They're checking you can locate your work on the wider map, not just inside your own hedge maze.

5. The friendly softball that's secretly hard

"Summarize your thesis in two minutes." It sounds like a warm-up. It's the hardest question on this list: you have two hundred pages of resolution and two minutes of bandwidth, and without a rehearsed compression you will narrate your literature review while the clock burns. This is the one answer worth learning close to verbatim — and it doubles as your opening, when nerves peak.

Prepare answers you can retrieve under stress

Here's the trap: re-reading your thesis feels like preparation and does almost nothing. Memory research is blunt about this — what builds recall is retrieval, pulling answers out of your head, not pouring the text back in. In the classic studies, learners who kept testing themselves held about 80% a week later; those who stopped once an item felt known dropped to roughly 35%. "I read it again last night" is the 35% plan.

So prepare the way actors do, on your own material:

• Write short answer beats, not essays. For each archetype, draft a 60–90 second answer. Attach an intention to each — what are you doing to the committee here: reassuring, conceding, drawing a boundary? In studies, rehearsing with a communicative intention beat deliberate memorization by more than half again (31% vs 19% recall). And when exact words slip, the intention tells you what the sentence was for, so you improvise back onto the rails.
• Practice out loud. Speaking adds a measurable 10–20% to retention over silent review, and it exposes the sentences you can write but can't say.
• Use fading cues on the few verbatim bits. For the two-minute summary and the first sentence of each answer, don't jump from reading to reciting — that cliff is where practice fails on fresh text. Descend a ladder: full text → text with blanks → first letters only → nothing. On newly written material, this is the one technique shown to work when plain from-memory recitation produces no benefit at all.
• Rehearse your concessions. People bluff under pressure because the bluff is fluent and the concession isn't. Drill the honest sentences — "that's a real limitation; here's what it does and doesn't threaten" — until they're the easy ones to reach.

Mock defenses work — the hard part is getting a good one

Every guide to the viva ranks a mock defense as the single most effective preparation, and the memory research explains why: it's retrieval practice, feedback, and pressure inoculation in one package. Studies on choking show that practicing under even mild evaluative stress — someone watching, a camera on — protects performance when the real stakes arrive, because the adrenaline becomes a familiar cue instead of a novel emergency.

The catch is that committee-quality questioning is scarce. Labmates are too kind and too close to your topic — they ask the questions you can already answer. Your supervisor has read the thesis too many times to impersonate a skeptical external. And what makes a real examiner hard isn't the first question; it's the follow-up. They don't accept your answer and move on. They press: "But that doesn't address X." Most mock setups never simulate that second beat, which is exactly where defenses are won or lost.

This is the gap LobsterClass's Defense & exam mode was built for: an AI committee — full panel, a methods hardliner, or a friendly examiner — that questions you on your thesis material, pushes back with follow-ups instead of accepting your first answer, and grades you on directness and on whether you were honest about your limitations rather than bluffing past them.

The final two weeks

Spacing research puts the ideal gap between sessions at roughly 10–30% of the time remaining — so at two weeks out that's a session every two or three days, tightening toward daily in the last week. Sleep is doing real work too: one study found a night's sleep between sessions roughly halved the relearning effort. Never stack two heavy sessions in one evening when you could split them across a night.

Then walk in and let them ask. You've already answered harder versions of every question in the room — you checked.