The logical fallacies the GMAT loves to test.

Here is the thing nobody tells you about GMAT Critical Reasoning: the exam is not testing whether you know the Latin name for a fallacy. It will never ask you to identify “post hoc ergo propter hoc” or pick the word “equivocation” from a list. What it tests is whether you can feelwhere an argument is weak. The stems are practical — weaken this, find the flaw, choose what would help you evaluate it — and every one of them rewards the same skill: spotting the exact joint where the reasoning is held together with tape.

The good news is that the GMAT recycles a small set of these broken joints. Arguments on the test are short, and they break in predictable ways. If you learn to recognize the handful of shapes below, you stop reading the five answer choices as five fresh puzzles and start reading them as five attempts to either name the flaw, exploit it, or distract you from it. The right answer is almost always the one aimed at the joint you already found.

You are not memorizing a glossary. You are building a pattern library. The test never asks for the name — it asks you to act on the shape. Recognize the shape and the correct answer narrows itself to one or two before you finish reading the choices.

The big one: correlation treated as causation

If you learn only one shape, learn this one. It is the single most-tested reasoning flaw on the GMAT, and it shows up in Critical Reasoning and in the data-driven arguments inside Data Insights.

Correlation implies causation

The flaw: two things happen together, so the argument concludes one caused the other.

Example:“Employees who attended the optional wellness seminar took 20% fewer sick days last quarter. Therefore the seminar improved employee health, and we should make it mandatory.” The two facts are linked, but linkage is not a mechanism. Maybe the kind of person who volunteers for an optional seminar is already health-conscious and was always going to take fewer sick days.

How it appears:on a Weaken question, the correct answer breaks the causal link — it shows the effect would have happened anyway, or that the two variables are connected by something else entirely. On a Strengthen question, the correct answer rules out the rival explanations. The trap answer typically restates the correlation in stronger words and dresses it up as proof.

Reversed causation

The flaw:the argument has the arrow pointing the wrong way — B actually caused A, not A caused B.

Example:“Profitable firms spend the most on advertising, so heavy advertising is the path to profitability.” Plausibly the reverse: firms advertise heavily because they are already profitable and can afford it. A correct weakener will quietly flip the arrow.

The common third cause

The flaw: A and B are correlated only because some third factor C drives both.

Example:“Cities with more bookstores have higher household incomes, so opening bookstores raises local incomes.” Far more likely, an affluent, educated population both supports bookstores and earns more — education is the hidden C driving both. When you spot a causal claim, train yourself to ask three questions in order: could the effect cause the cause? Could something else cause both? Could it just be coincidence? The right answer usually lives in one of those three doors.

Necessary versus sufficient conditions

This one is quieter than causation but just as common, and it trips up strong students because the logic feels airtight at a glance.

The flaw: the argument treats a condition that is required as if it were enough, or vice versa.

Example:“You need a strong analytics background to get promoted to manager here. Priya has a strong analytics background, so she'll be promoted.” The background is necessary — you can't get promoted without it — but the argument slid into treating it as sufficient. Having it does not guarantee the promotion; plenty of other things could block her.

How it appears:watch for the words. “Required,” “must,” and “only if” signal a necessary condition. “Enough,” “guarantees,” and “ensures” signal a sufficient one. The flaw is almost always the argument quietly upgrading one into the other. The correct answer on a flaw question describes exactly that confusion in abstract terms.

Sampling problems

Unrepresentative sample and hasty generalization

The flaw: the argument draws a broad conclusion from a sample that is too small, too narrow, or self-selected to stand for the whole group.

Example:“We surveyed visitors to our premium airport lounge, and 90% said they fly more than twenty times a year. The average traveler clearly flies a great deal.” The sample is drawn from exactly the high-frequency flyers most likely to be in a premium lounge. It tells you nothing about the average traveler.

How it appears:the correct weakener shows the sample differs in a relevant way from the population the conclusion is about. The trap answer attacks the sample size when size was never the problem — a survey of fifty thousand lounge visitors is still useless for the claim. The issue is who was sampled, not how many.

Survivorship bias

The flaw:the argument studies only the survivors — the cases that made it through some filter — and forgets the ones that dropped out and left no trace.

Example:“Every founder profiled in this bestselling business book dropped out of school, so dropping out is a smart bet for ambitious entrepreneurs.” The book profiles winners. The thousands who dropped out and failed never got a chapter, so they are invisible to the analysis. The denominator is missing.

How it appears:any time a conclusion is built from a group that already passed a test — funded startups, returning customers, completed projects — ask what happened to the ones that did not pass. The correct answer points at the missing failures.

The percent-versus-number trap

This is a GMAT signature, and Data Insights leans on it heavily because the section is built around reading tables and charts where percentages and raw counts sit side by side. The trap is simple: a percentage and an absolute number can move in opposite directions, and an argument that swaps one for the other looks valid until you separate them.

The flaw: the argument reasons about a percentage as though it were a count, or about a count as though it were a rate, ignoring that the base can change.

Example one — rising rate, shrinking base: “The share of our revenue coming from new customers rose from 10% to 25% this year, so we're acquiring far more new customers than before.” Not necessarily. If total revenue collapsed because existing customers left, the new-customer slice could be a larger share of a much smaller pie — fewer new customers in absolute terms, higher percentage.

Example two — rising count, irrelevant to rate: “The number of defective units shipped rose 30% last year, so our quality control is deteriorating.” If the factory shipped twice as many units, a 30% rise in defects actually means the defect rate fell. The raw count went up while quality improved.

How it appears:whenever an argument cites a percentage to support a claim about a quantity, or a quantity to support a claim about a rate, your antenna should go up. The correct answer almost always reintroduces the variable that was hidden — the base for a percentage, or the total for a count.

Part-whole errors

The flaw: the argument assumes what is true of the parts must be true of the whole (composition), or that what is true of the whole must be true of each part (division).

Composition example:“Each department in the company cut its own costs this year, so the company's total costs must have fallen.” Not if a new shared expense — a company-wide system, say — was added on top, or if one department's cut shifted work and cost onto another.

Division example:“This fund returned 12% last year, so the bonds inside it returned 12%.” The average says nothing about any single holding; some positions may have lost money while others soared. What holds for the aggregate need not hold for the components.

How it appears:the correct answer shows a case where the part and the whole diverge — an interaction, an addition, or an average that hides spread.

The softer flaws

These show up less often as the central crack of an argument, but they appear constantly as wrong answer choices and as the flaw in flaw-description questions. Knowing their shape lets you eliminate fast.

False dichotomy

The flaw: the argument presents only two options when more exist, then knocks one down to force the other.

Example:“Either we slash the marketing budget or we go bankrupt. We can't go bankrupt, so the budget has to be cut.” There may be a dozen other levers — new revenue, renegotiated leases, a different cut. The two-door frame is the trick.

Circular reasoning

The flaw:the conclusion is smuggled into the premises — the argument assumes the very thing it claims to prove.

Example:“This is the most trusted consulting firm in the region because no other firm has earned this level of trust.” The premise and the conclusion are the same sentence wearing different clothes. Nothing independent supports the claim.

Equivocation

The flaw: a key word changes meaning partway through the argument, so the two halves do not actually connect.

Example:“Our investment is sound, because the building it funded was constructed with sound engineering.” “Sound” means financially safe in the conclusion and structurally solid in the premise. The shared word hides a broken bridge.

Appeal to authority or popularity

The flaw: a claim is treated as true because an authority endorsed it or because many people believe it, with no actual evidence behind either.

Example:“Three out of four executives we asked prefer this software, so it must be the most efficient option.” Preference is not efficiency, and the executives may not have measured anything.

Ad hominem and straw man

Ad hominemattacks the person making the argument rather than the argument itself: “The analyst who recommended this merger once worked for the target, so the recommendation is worthless.” The bias is worth noting, but it does not by itself make the analysis wrong. Straw mandistorts an opponent's position into a weaker one and then refutes that: answering a call for “tighter expense review” by arguing against “eliminating all spending.” On flaw questions, both of these are common wrong answers that sound damning but miss the real crack.

The master move: no alternative explanation considered

Sitting underneath almost every causal and sampling flaw is one umbrella weakness, and it is worth naming on its own because so many GMAT arguments share it: the argument considers exactly one explanation and never rules out the others.

A study finds a result and the author jumps to their favorite cause. A program runs and the author credits the program for the outcome. In nearly every case, the correct weakener is the sentence that says “but here is another thing that could explain the same result.” And the correct strengthener is the sentence that rules one of those alternatives out. When you cannot immediately find the joint, ask the catch-all question: what else could account for this? That question alone solves a large share of Weaken and Strengthen items.

Most GMAT arguments fail the same way: they fall in love with one explanation and never check the alternatives. “What else could cause this?” is the most productive question you can ask in the entire Verbal section.

How to actually use this on test day

You will not have time to run a checklist of twelve fallacies on each question. You do not need to. The workflow is faster than that, and it is the same whether the question lives in Critical Reasoning or in a Data Insights argument:

1. Find the conclusion first. Underline it mentally. Everything else in the stimulus is either evidence for it or noise.
2. Name the gap, not the fallacy.Ask what the author had to assume to get from the evidence to that conclusion. The gap usually matches one of the shapes above — a leap from correlation to cause, from sample to population, from percent to number, from part to whole.
3. Predict the answer before you read the choices. If the flaw is causal, you already know a weakener will attack the link and a strengthener will defend it. Walk into the answers with a prediction, not an open mind.
4. Attack the exact joint. The correct answer hits the specific gap you found. Choices that are true but irrelevant, or that attack a different joint, are the traps. An answer can be a real-world fact and still be wrong because it does not touch thisargument's weak point.

That is the whole method. Conclusion, gap, prediction, attack. The fallacy names are just labels for the gaps you will learn to feel. If you want to go deeper into the underlying logic, there are excellent free and openly licensed introductory logic textbooks — reading one will sharpen your instinct for what makes an inference valid. But do not confuse studying logic with practicing GMAT questions. The exam rewards speed at spotting these specific shapes under time pressure, and that comes only from doing real questions and reviewing why each wrong answer was built to tempt you.

The short version

Correlation is not causation, and its cousins — reversed arrows and hidden third causes — are the most-tested shapes on the exam. Necessary is not sufficient. A sample has to look like the population it is about. Percentages and raw numbers can move in opposite directions. Parts and wholes can diverge. False dichotomies, circular reasoning, equivocation, and the appeals are softer flaws that show up most often as wrong answers. And underneath nearly all of it is one question: what else could explain this? You will never name these on test day. You will just see the shape, predict the flaw, and pick the answer that attacks the joint.

The platform

Zakarian GMAT's Critical Reasoning chapter teaches each of these shapes with original worked arguments, then drills them in timed problem sets where every wrong answer is tagged by the trap it represents. The error log's six-tag taxonomy surfaces whether your CR misses are conceptual (you didn't see the flaw) or strategic (you saw it but picked the wrong attack). The sample chapter is free if you want to see the teaching first.