Case Study 2 — The Mock Defense: Turning an Ambush into Ownership

A composite, realistic scenario built from common defense-prep patterns. Names are fictional. This is the same ownership lesson as §16.8, watched in the week before a defense — the contrast to Case Study 1's scope fight: there the danger was doing too much; here it is failing to own what you did.

The setup

Amara Bello had finished writing. The dissertation — humidity and drift corrections for low-cost PM2.5 sensors, the scope held to two contribution chapters after the narrowing in Case Study 1 — was printed, signed off by her advisor, and in her committee's hands. The defense was nine days out. She felt the specific dread of someone who has done the work but has not yet had to say it out loud: she could write about her method for pages, but the thought of a committee member firing a question across the table made her stomach drop.

Her advisor, Dr. Okonkwo, had a remedy he insisted on for every student: a mock defense. Two postdocs and a senior PhD student from the lab agreed to play the committee. They had read her chapters. They were told to be tough but fair — to ask the questions a real examiner asks, not gotchas.

The ambush that wasn't

Twenty minutes in, after Amara's practice talk, the senior student leaned forward and asked the question Amara had been quietly hoping no one would: "Your correction was validated on a single city. Why should anyone believe it transfers anywhere else?"

Amara's first answer was the one §16.8 warns against. She heard herself say: "Well, the climate there is pretty representative, so I think it would probably generalize fine."

The room went quiet in the way that tells you something just went wrong.

The diagnosis: Amara had done three things the defense punishes. She minimized a real limitation ("pretty representative") instead of naming it. She speculated without evidence ("probably generalize fine") — a claim her data could not support. And she sounded surprised by her own weakness, as if single-city validation were an accident rather than a deliberate, defensible scoping choice. The limitation was fine. Her relationship to it was the problem. As §16.8 puts it, the fatal failure is not "my method had a weakness" — it is being unaware of the weakness, or pretending it is not there.

Dr. Okonkwo stopped the mock. "You know this answer cold," he said. "You scoped to one city on purpose. Say that."

The three-question rebuild

That evening Amara did what she should have done before the mock: she ran every major choice in the thesis through the three questions from §16.8 — Why did I do it this way? What are the alternatives and why did I reject them? What are the limitations and what would I do differently? She wrote the answers out, longhand, for each chapter. The single-city question, rebuilt:

The limitation had not changed. What changed was that Amara could now hold it — name it, justify the scoping, and point to where the thesis already acknowledged it.

The second mock

Three days later they ran it again. Same question, from the same student: "Why should anyone believe this transfers?"

This time: "It's a real limitation, and I'm explicit about it in §7.4 — my claim is that the correction is validated in this deployment, not that it generalizes. I chose depth over breadth deliberately: three years against a reference monitor in one city, rather than thin coverage across several with no ground truth. I'd expect partial transfer to similar humid climates because the failure mode is physical, not site-specific — but that's a hypothesis my data can't confirm, so I name multi-city replication as the first future-work item, with the same held-out-months protocol. I'm not claiming more than I measured."

The student nodded and moved on. There was nothing left to push on — Amara had already said the thing the question was fishing for.

What the mock cost — and saved

The mock cost an afternoon and one bruised ego in front of three colleagues. It saved Amara from delivering that first answer to her actual committee, where "it would probably generalize fine" would have invited twenty minutes of exactly the questioning she feared — not because the work was weak, but because the answer invited the committee to test whether she understood her own limits. The mock surfaced the one weakness in her command of the work while it was still cheap to fix. At the real defense, the single-city question came up almost verbatim. She was ready, and the room felt like a conversation between colleagues, not an interrogation — which, as §16.8 notes, is what a defense is at its best.

Lessons

1. The defense tests ownership, not perfection. The committee already knew the work was single-city; they were testing whether Amara knew what that meant. Owning the limit is the whole game.
2. Minimizing a weakness is worse than the weakness. "Pretty representative, probably fine" signals you either don't understand the limitation or hope no one notices — the two things a defense is built to catch.
3. Don't speculate past your data. "I'd expect partial transfer because the failure mode is physical, but my data can't confirm it" is honest and strong; "it'll generalize fine" is a claim you can't defend and shouldn't make.
4. Prep is mechanical: three questions per choice. Why this way, what alternatives, what limitations — answered for every chapter, this covers almost any question a committee can ask, because almost every question is a version of one of the three.
5. Rehearse out loud, ideally to a mock committee. Amara could write the answer; she could not yet say it. Saying it — and getting it wrong once, safely — is what converted dread into recall.

One sentence to remember: A defense doesn't ask whether your work is flawless — it asks whether you know exactly where the flaws are and why you chose them; the mock defense is where you find out, while it's still free.