How to study for proof-based math exams

Studying for a proof-based math exam means shifting your target from procedures to precise statements. Memorize every definition word-for-word, learn each major theorem with a proof sketch you can reproduce from memory, build a bank of counterexamples that mark the boundaries of each theorem, and write complete proofs under timed conditions before exam day. Re-reading lecture notes — the habit that carried you through calculus — is the least effective thing you can do in real analysis or abstract linear algebra.

Why your calculus study habits stop working

Computational courses reward procedure fluency. You identify the problem type, retrieve the matching technique, and execute it; drilling twenty similar problems genuinely prepares you, because the exam will contain a twenty-first. Proof-based courses break this model. The exam asks you to construct an argument you have never seen, assembled from definitions and theorems you are expected to know exactly. There is no procedure to pattern-match against.

This transition is well documented — Lara Alcock's How to Study for a Mathematics Degree (2012) treats it as the central adjustment of a math education. The grading changes too: a proof earns full marks only if it is logically complete, and "knowing roughly how it goes" earns close to nothing. Re-reading notes produces fluency with the shape of the material, which is precisely the thing proof exams do not test.

What they do test breaks down into four trainable skills: stating definitions exactly, reproducing core proofs, recognizing the limits of theorems, and writing arguments under time pressure.

Memorize definitions word-for-word — yes, really

In computational math, a half-remembered formula is partial credit. In proof-based math, a half-remembered definition is nothing, because a definition is a tool you can only use whole. Consider continuity at a point:

$f$ is continuous at $a$ if $\forall \epsilon > 0\, \exists \delta > 0$ such that $|x - a| < \delta \implies |f(x) - f(a)| < \epsilon$.

Every part of this is load-bearing. The order of quantifiers is the entire content: $\delta$ is chosen after $\epsilon$, and may depend on it. Move the quantifiers and you get a different concept — in uniform continuity, the only change is that one $\delta$ must work for all points at once, and that single shift in dependency separates two major ideas of analysis. The inequalities, the direction of the implication, which variable is free: none of it is decoration.

There is a practical reason precision matters beyond exam questions that say "state the definition." Most proofs at this level begin with a definition — "Let $\epsilon > 0$ be given" — so a definition you cannot state verbatim is a proof you cannot start.

The standard is verbatim plus one plain sentence of your own ("continuity means you can force outputs close by keeping inputs close"). Verbatim recall without the plain sentence is recitation; the plain sentence without verbatim recall is a proof you cannot write down.

Make theorem–proof flashcards

For each major theorem, make a card with the statement on the front and a proof sketch on the back — the two to four moves that make the argument work, not a transcript. For the rank–nullity theorem,

$\dim(V) = \text{rank}(T) + \text{nullity}(T),$

the sketch is three lines: take a basis of $\ker T$; extend it to a basis of $V$; show the images of the added vectors form a basis of the range. If you understand the course, the routine details reconstruct themselves from the sketch. The sketch is the part that does not reconstruct itself — it has to be remembered.

Add a second card type for hypotheses: "What does the Mean Value Theorem require, and where is each hypothesis used?" ($f$ continuous on $[a,b]$, differentiable on $(a,b)$ — and you should know which step of the proof fails without each.)

These cards are dense with quantifiers, Greek letters, and operators, so the tool you build them in has to render real mathematical notation rather than flattening it to plain text. ExamTeX renders full LaTeX on every card; you can generate a deck directly from your lecture notes or start with a linear algebra deck and edit it toward your course. Because proof courses are cumulative — week ten quietly assumes week two — review the deck on a spaced repetition schedule rather than in a pre-exam binge.

Reproduce proofs from memory, not just read them

Reading a proof and following every step feels like understanding, but following is a much weaker state than producing. The general result is the testing effect: Roediger and Karpicke (2006) found that practicing retrieval beats re-studying for long-term retention. For proofs specifically, Hodds, Alcock, and Inglis (2014) showed that training students to self-explain each line — asking why the step holds and how it connects to what came before — measurably improved proof comprehension compared to ordinary reading.

Combine the two into a single routine:

1. Read the proof once, actively — for each line, ask what justifies it and where each hypothesis gets used.
2. Close the notes and reproduce the proof on paper, in full sentences.
3. Compare. The place where you stalled is almost always the key idea of the proof — mark it.
4. Reproduce it again two or three days later, from nothing.

You do not need this treatment for every proof in the course. Aim at the named theorems and the proofs your professor flagged or spent a full lecture on — typically one or two per week.

Build a counterexample bank

Theorems are fenced in by their hypotheses, and counterexamples are the fence posts. Collect them deliberately, because "prove or disprove" questions are a fixture of proof-based exams, and a stocked counterexample bank converts them from invention problems into recall problems.

Tempting claim (false)	Counterexample
Continuous implies differentiable	$f(x) = \lvert x \rvert$ at $x = 0$
Terms tend to zero, so the series converges	$\sum_{n=1}^{\infty} \frac{1}{n}$ diverges
Bounded sequences converge	$x_n = (-1)^n$
A pointwise limit of continuous functions is continuous	$f_n(x) = x^n$ on $[0,1]$
Matrix multiplication commutes	almost any $A, B$: $AB \neq BA$

Each counterexample is also a lesson in why a hypothesis exists: $\lvert x \rvert$ is exactly what the differentiability hypothesis of so many theorems is excluding. Put these on flashcards too — claim on the front, counterexample on the back — and you will find your grasp of the definitions sharpens as a side effect.

Practice writing proofs under time pressure

A proof exam is typically three to five proofs in two or three hours, and untimed problem sets do not rehearse that constraint. In the last two weeks before the exam, simulate it: closed book, a fixed clock, complete written arguments in full sentences.

Past exams are the best source if your department releases them. If it does not, you can generate a practice exam from your own course materials — ExamTeX produces proof-style questions with an answer key, typeset like a real exam paper, which makes the simulation feel like the event. When you grade yourself, check for the things a grader checks: is every claim justified, are the quantifiers in the right order, did you actually use each hypothesis you invoked? Our practice testing guide covers why this kind of rehearsal outperforms passive review so consistently.

Keep a definitions–theorems–techniques inventory

Maintain one living document per course with three sections:

• Definitions — every definition, stated verbatim.
• Theorems — each statement, its hypotheses, and a one-line proof idea.
• Techniques — the recurring moves of the course: prove set equality by double inclusion, the epsilon-over-two trick, induction on dimension, contradiction via minimality.

The techniques list is the one students skip and the one that matters most on exam day, because techniques are what you reach for when a problem is genuinely new. Update the inventory weekly; it takes minutes if it never falls behind. In the final week, it doubles as your blank-sheet test: reproduce the entire inventory from memory, and whatever fails to appear is your remaining study list.

The bottom line

Proof-based exams test whether you can state definitions exactly, deploy theorems with their hypotheses intact, and construct complete arguments against a clock — none of which re-reading trains. Memorize definitions verbatim with a plain-sentence gloss, learn proof sketches and reproduce them from memory on a spaced schedule, stock a counterexample bank, and write full timed proofs before the real one. The course stops feeling like a wall of abstraction and starts feeling like an inventory you control.