In almost every doctoral thesis defense the committee asks the same things: why you chose that design, how you justified your sample size, whether your instruments were validated in your population, whether you checked the assumptions of your analyses, whether you controlled for multiple comparisons, whether you report effect sizes, how far you can generalize, what your work adds to what is already published, and what its real limitations are. Those are twelve questions that come up defense after defense, whether you work in psychology, biology, economics, education or nursing. Below you have all twelve, why the committee asks them and how to approach each answer.
A defense feels unpredictable, but it rarely is. A competent committee keeps returning to the same fronts: the decisions you made and did not justify, the limits of what you can actually conclude, and your real command of the methods. If you prepare those questions in advance, you arrive with thought-out answers instead of improvising under pressure. And if you want the ones for your specific thesis, with your objectives and your analyses, there is a free defense simulator that generates them from your summary.
Why a defense is more predictable than it looks
The committee is not trying to catch you out on a trick: it wants to confirm that you understand what you did and that you recognize its limits. Almost every question comes from one of three places: a methodological decision that was never justified in the text, a conclusion that goes beyond what the data allow, or an analysis whose assumption or alternative you did not consider. Knowing those patterns is half the work.
On design and justification
1. Why did you choose this design and not another? They want to see that the choice answers your question, not that it was convenient. Tie design to objective: what your design can answer and what it rules out, and why the alternatives fit worse.
2. How did you justify your sample size? The flagship question. If you ran an a priori calculation (G*Power, simulation), walk them through it; if not, reason about the minimum effect you could detect. What does not work is "it was what I could get." Use the free sample size calculator if you need the logic.
3. Is your sample representative? Who can you generalize to? Be honest about the sampling limits (convenience, a single institution) and bound who your conclusions apply to. A committee punishes overgeneralizing more than admitting a limit.
On measurement and instruments
4. Why these instruments, and are they validated in your population? Cite validation in your language and population; if you used an adapted version, explain the process. If reliability in your sample was modest, raise it yourself first.
5. How did you ensure measurement quality? This is where reliability (alpha, omega) lives, and in psychometric work the factor structure and measurement invariance if you compare groups. If you compared groups without testing invariance, be ready to justify it.
On the statistical analyses
6. Did you check the assumptions of your analyses? Be clear on what each test required (normality of residuals, homoscedasticity, sphericity, multicollinearity) and how you verified it. If one failed, explain what you did (transform, robust test, non-parametric alternative). If you want to go through them one by one before the defense, there is a checklist in the guide on how to verify statistical assumptions.
7. Why this analysis and not an alternative? A member will often propose the model they would have used (a mixed model instead of repeated-measures ANOVA, an ANCOVA instead of comparing groups). Do not get defensive: acknowledge the alternative, say what it adds and why your choice was reasonable for your question and data.
8. Did you control for multiple comparisons? If you ran many tests, explain the correction (Bonferroni, FDR) or why you did not apply one. No control without justification is an easy criticism.
9. Do you report effect sizes, not just significance? A significant p with a trivial effect is not a finding. Keep your effect sizes and their practical interpretation in your field ready. If you are unsure how to read them out loud, here is the interpretation guide for Cohen's d, eta² and r.
On interpretation and contribution
10. You use causal language with a design that does not support it. If your design is cross-sectional or correlational, watch the verbs: "predicts", "produces", "causes". The answer is not to retract but to show you distinguish association from causation and to reframe your conclusions accordingly.
11. What does your thesis add that was not already in the literature? Prepare a two-sentence answer: the specific gap you fill and why it matters. It is surprising how many defenses stumble here for lack of a clear one.
12. What are the limitations and what would you do differently? Make your limitations the real, relevant ones, not the textbook ones. Showing you know exactly where your work is weak signals research maturity, not weakness. To walk in with that list already written, review the most common statistical errors in a thesis first: a good share of the limitations you end up defending start there.
The curveball
There is almost always an unexpected question that is not on any list: a twist on your mechanism, an alternative explanation of your results, an implication you had not considered. You do not prepare it by memorizing; you prepare it by knowing your thesis so well that you can reason live. If you have the twelve above covered, the curveball stops being scary.
Prepare the questions for YOUR thesis
Before the defense: this list is the general pattern, but what really changes the outcome is seeing the questions your specific work raises. That is why I built the Thesis Defense Simulator: paste your thesis summary (objectives, method and results) and it returns the questions the committee will most likely ask, why they ask them, and how to approach each answer. It is free and it works whatever your discipline. And if what you are defending is an undergraduate or master's dissertation rather than a PhD, use the simulator's dissertation version instead, tuned to that committee's level of demand.
And if what you are facing is a journal submission rather than the defense, its peer-review counterpart is the AI Paper Reviewer. If you would rather prepare the statistical side with someone who has sat on committees, you can book a free diagnostic.