GRE Math Subject Test common mistakes
GRE Math Subject Test · Understand the traps and how to avoid them.
Spending more than about 2.5 minutes on a single multiple-choice question before moving on, then running out of time on the last 15 questions.
The exam gives roughly 2.5 minutes per question on average, and early questions often look like routine calculus problems that should be quick, so candidates over-invest in them and do not build a time budget. Because there is no penalty for wrong answers, leaving questions blank at the end is pure lost expected score.
What to do instead
Practice with a hard per-question cap: mark a question, guess a letter, and move on after 3 minutes. Use a two-pass strategy: first pass takes all questions you can do in under 90 seconds, second pass returns to the marked ones. In the last 2 minutes, fill every blank with the same letter so no question is left unanswered.
Treating the five answer choices as independent options and solving each question fully from scratch instead of eliminating choices by estimation, plugging in values, or checking limiting cases.
Candidates carry over habits from proof-based coursework where the answer must be derived, and they distrust shortcuts as mathematically impure. But the test is scored only on the chosen letter, and many choices differ by sign, order of magnitude, or a single factor that a quick check exposes.
What to do instead
For every question, before computing, ask: can I plug in x = 0, 1, or a large value? Can I check the sign or the units? Can I eliminate choices that fail a special case? Drill this by redoing practice questions where you solve only by elimination and record how often it suffices.
Applying a convergence test to a series without first checking that the terms tend to zero, or confusing absolute convergence with conditional convergence when rearranging terms.
The nth-term test is presented early and feels trivial, so it is skipped under time pressure. Rearrangement is a subtle point: many candidates know the alternating series test but not that a conditionally convergent series can be reordered to any sum, so they assume any convergent series can be regrouped freely.
What to do instead
Make a one-line checklist for every series: (1) do terms go to 0? (2) are terms positive, so comparison and ratio tests apply? (3) if signs alternate, is it absolutely convergent? Only then apply the test. Memorize the standard counterexamples: harmonic series diverges, alternating harmonic converges conditionally, and p-series converge exactly when p > 1.
Computing a definite integral or derivative of a piecewise or absolute-value function by applying the formula across the whole interval, ignoring the point where the definition changes.
The integrand looks like a single expression, and the absolute value or piecewise break is easy to overlook. Candidates also forget that the fundamental theorem of calculus requires the integrand to be continuous on the interval, or they differentiate an integral with a variable in both limits without the chain rule.
What to do instead
Before integrating, locate zeros and breakpoints of the integrand and split the interval. For d/dx of an integral with limits a(x) and b(x), write f(b(x))b'(x) - f(a(x))a'(x) explicitly. Practice a set of problems with |x - c|, floor functions, and variable limits until splitting becomes automatic.
Assuming a matrix is diagonalizable, or that eigenvectors for distinct eigenvalues are automatically orthogonal, without checking hypotheses such as symmetry or distinct eigenvalues.
Linear algebra courses emphasize the spectral theorem for real symmetric matrices, and candidates overgeneralize it to all matrices. They also confuse algebraic and geometric multiplicity, and forget that a repeated eigenvalue may have too few independent eigenvectors.
What to do instead
Keep a hypothesis table: symmetric real matrices are orthogonally diagonalizable; distinct eigenvalues always give independent eigenvectors, but orthogonality needs symmetry; a matrix is diagonalizable iff each eigenvalue's geometric multiplicity equals its algebraic multiplicity. Test each claim in practice against a defective example such as [[1,1],[0,1]].
Computing a determinant or rank by row reduction while forgetting that row swaps and row scaling change the determinant, or concluding rank from the number of nonzero rows without reducing to echelon form.
Row reduction is learned as a single procedure for solving systems, and the bookkeeping rules for how each operation affects the determinant are treated as an afterthought. Partial reduction then produces a wrong determinant or an inflated rank.
What to do instead
Track a running factor: each row swap multiplies the determinant by -1 and each row scaling by the scalar, so divide at the end. For rank, reduce fully to row echelon form and count pivots, not nonzero rows of an intermediate matrix. Verify with a small 3x3 example where the answer is known.
Solving a counting or probability problem by multiplying probabilities for events that are not independent, or by counting ordered arrangements when the problem asks for unordered selections.
Independence is assumed because the events sound unrelated, and the ordered versus unordered distinction is buried in wording such as 'a committee' versus 'a president and vice president'. Candidates also double-count when a problem involves identical objects.
What to do instead
For each problem, state explicitly whether order matters and whether sampling is with or without replacement before choosing a formula. Use the complement for 'at least one' problems, and check small cases by hand. For identical objects, divide by the factorial of each repeated group.
Using the wrong variance formula for a sum or difference of random variables, adding variances when variables are subtracted, or adding standard deviations instead of variances.
The rule Var(X ± Y) = Var(X) + Var(Y) for independent variables is counterintuitive because the minus sign disappears, and candidates reason by analogy with expectations, where E(X - Y) = E(X) - E(Y). Mixing standard deviations and variances compounds the error.
What to do instead
Convert everything to variances first, apply Var(X ± Y) = Var(X) + Var(Y) only under independence (and add 2Cov(X,Y) otherwise), then take the square root at the end if a standard deviation is requested. Write the covariance term explicitly even when it is zero, so the independence assumption is visible.
Choosing a convergence test or a limit technique in the abstract algebra and topology questions by pattern-matching on keywords, without verifying that the structure satisfies the theorem's hypotheses (for example, that a group is abelian or that a map is a homomorphism).
Abstract algebra and topology questions are short and formulaic-looking, so candidates recall a theorem by name and apply it immediately. The distractors are usually answers that follow from the theorem when its hypotheses are violated, which makes the error feel correct.
What to do instead
For each abstract question, write the hypotheses of the candidate theorem in a single line and check each one against the given structure. Memorize the standard counterexamples that separate hypotheses: S3 is non-abelian, Z is not finite, the rationals are not complete, and a continuous bijection need not have continuous inverse. Use them to test whether a claimed implication is true.
Misreading the question stem on 'which of the following must be true' versus 'could be true' versus 'is false', and picking the first statement that is true rather than checking all five choices.
The quantitative questions train a compute-and-match reflex, but the logical-quantifier questions require evaluating every choice. Under time pressure, candidates stop at the first true statement even when the question asks for the one that must be true in all cases or the one that is false.
What to do instead
Circle the quantifier word in the stem before reading the choices. For 'must be true', try to construct a counterexample to each choice and reject it; for 'could be true', find one example that satisfies it. Only the choice surviving all five checks is correct, so budget time to evaluate every option.