Rank 2 by frequency | 664 questions in corpus (14.7% of all questions)
Inference questions flip the normal LR script. Instead of giving you an argument to evaluate, the stimulus hands you a set of facts — observations, rules, data points — and asks what follows from them. You are not judging someone else's reasoning; you are doing the reasoning. The answer choices are candidate conclusions, and your job is to find the one that the stimulus actually proves or supports.
Unlike Flaw, Weaken, Strengthen, or Assumption questions, an Inference stimulus typically contains no argument and no conclusion. What you get is a fact set: a collection of statements, observations, data, or rules that you are asked to treat as true. Everything on the page is evidence; nothing is a claim being defended.
Because there is no conclusion in the stimulus, the correct answer is the conclusion. You must generate it rather than evaluate one that is handed to you. Support flows from the stimulus to the correct answer — the opposite of Strengthen/Weaken, where an answer choice supports or attacks a given conclusion. This makes Inference a fundamentally different cognitive task: you are not judging reasoning, you are producing it.
The underlying skill tested is your ability to draw valid conclusions from evidence — to recognize logical entailments, combine multiple premises, and distinguish between what must be true, what is likely true, and what merely could be true. At the hardest end, it tests fluency with formal logic: conditionals, contrapositives, and quantifiers.
The standard of proof varies by subtype. Under Must Be True, the answer has to be 100% logically guaranteed by the stimulus — no wiggle room, no exceptions. Under Most Strongly Supported, the answer does not have to be certain; it just has to be the best-supported statement among the five choices. Think of it as roughly 80% proven: imaginable exceptions do NOT make the answer wrong. Under Must Be False, the answer must be 100% incompatible with the stimulus information.
Regardless of subtype, valid inferences come from a narrow set of moves. An answer may directly restate or paraphrase information already in the stimulus. It may combine two or more facts to produce something new that is logically entailed — linking overlapping concepts, applying a rule to a case, chaining conditionals. It may state the contrapositive of a conditional ("if A then B" becomes "if not B then not A"). If the stimulus contains a chain like A→B and B→C, the transitive inference A→C is fair game.
What never counts: anything that requires new information the stimulus does not provide. Real-world knowledge, plausible assumptions, common sense — none of these justify an inference unless the stimulus itself gives you the bridge.
Inference questions appear in six recognizable subtypes. They all test the same underlying skill but use different standards of proof and different stimulus constructions.
Subtype 1 — Must Be True (MBT). The answer must be 100% logically guaranteed by the stimulus. Stems include "If the statements above are true, which one of the following must also be true?", "Which one of the following conclusions follows logically from the statements above?", and "Which one of the following can be properly inferred from the passage above?"
Subtype 2 — Most Strongly Supported (MSS). The answer need not be airtight; it just has to be the best-supported inference of the five. Stems include "Which one of the following is most strongly supported by the information above?" and "The statements above, if true, would most strongly support which one of the following?"
Subtype 3 — Must Be False / Cannot Be True. The answer must be logically impossible given the stimulus. "If the statements above are true, which one of the following CANNOT be true?"
Subtype 4 — Complete the Argument / Fill-in-the-Blank. The stimulus ends with a blank; the answer fills it in. Usually graded by the MSS standard. "Which one of the following most logically completes the passage?"
Subtype 5 — Conditional/Formal Logic Inference. The stimulus contains explicit if-then statements, and the correct answer is usually the contrapositive or a chain of linked conditionals (A→B, B→C, therefore A→C).
Subtype 6 — Quantifier-Based Inference. The stimulus uses all, some, most, many, few, none. Key trap: "some" can include "all." Chaining works for universals ("All A are B" + "All B are C" yields "All A are C"), but "most" statements only combine weakly ("Most A are B" + "Most A are C" only proves "Some B are C").
A consistent method works across every subtype. The order matters: understand the facts first, identify the standard of proof second, and generate predictions before touching the answer choices.
Step 1 — Treat every statement as true evidence. Read the stimulus carefully and resist the urge to look for a conclusion — there usually isn't one. Everything is a premise.
Step 2 — Identify the subtype from the stem. MBT requires 100% certainty; MSS allows the best-supported answer; MBF requires logical impossibility; Complete the Argument follows MSS. Your standard of proof depends on which you're dealing with.
Step 3 — Look for linkable concepts. Two statements that share a term can often be combined to produce a new inference. For conditional stimuli, lay out all the conditionals, write their contrapositives, and look for chains. For quantifier stimuli, track exact strength — do not let "most" slip into "all" or "some" slip into "most."
Step 4 — Make a prediction before reading the answers. You will not always predict the exact correct answer, but prephrasing several candidate inferences sharpens your eye and makes it much harder for traps to tempt you.
Step 5 — Test each answer against the standard. For MBT, ask "Does this HAVE to be true?" For MSS, ask "Is this the BEST-supported statement of the five?" Use process of elimination — knock out extreme, out-of-scope, and distorted answers first, then work through the remainder.
Correct Inference answers come from two basic patterns. The first is direct restatement: the answer paraphrases a stimulus fact using different words but identical meaning. When you see an accurate paraphrase of something the stimulus says, do not hesitate — restatement is a perfectly valid inference.
The second is logical combination: the answer combines two or more stimulus facts to produce a new statement that neither fact supports alone. This often involves linking overlapping concepts, applying a general rule to a specific case, or drawing a contrapositive. On harder questions, the combination involves three or more facts and requires chaining conditionals or recognizing quantifier overlaps.
Both patterns share one feature: every element of the correct answer is traceable back to something the stimulus says. There is no bridging assumption you have to import from outside the passage.
Wrong answers cluster into five recognizable shapes. Spotting them by shape lets you eliminate quickly instead of second-guessing each one.
Trap 1 — Out of scope / new information. The answer introduces an idea or entity the stimulus never mentions and that cannot be derived from it. ANY answer importing outside information should be eliminated immediately, no matter how reasonable it sounds.
Trap 2 — Too extreme / stronger than support. The answer uses always, never, all, none, only where the stimulus supports only a weaker claim. Strongly worded statements rarely must be true unless the stimulus explicitly supports them. When the stimulus says "some," the answer cannot say "most"; when it says "usually," the answer cannot say "always."
Trap 3 — Unsupported relationships. The answer proposes causal, comparative, or proportional relationships the stimulus does not establish. Correlation in the stimulus becomes causation in the answer; a mentioned group becomes a comparative claim. These are extremely common because they sound sensible.
Trap 4 — Distortion. The answer echoes stimulus wording but subtly changes meaning — shifting the direction of a conditional, reversing what the stimulus actually said, or swapping in a close-but-different concept. "Sounds right" is often the tell.
Trap 5 — Reversed conditional. The stimulus says "if A then B" and the answer says "if B then A." This is a logical fallacy (affirming the consequent), and it is one of LSAC's most reliable traps on conditional-logic stimuli.
LSAC scales difficulty on Inference questions through six levers. Hard questions usually combine several at once.
Long, dense stimuli that contain multiple facts hard to hold in working memory at once. Conditional logic requiring chaining across three or more conditionals, often combined with contrapositives. Quantifier interactions — combining "most" and "some" where the inference is non-obvious. Disguised conditionals hidden behind words like unless, only if, whenever, every, without. Multiple possible inferences, where several things could be inferred and you must find the specific one the answer choices target. And subtle word choice in answers, where the correct answer uses different wording from the stimulus and requires you to recognize logical equivalence.
The reasoning structures that show up most often in hard Inference stimuli are conditional chains, quantifier overlaps, comparative relationships, definitional links, temporal or causal sequences, disjunctive reasoning ("either A or B; not A; therefore B"), and contrapositive triggers where the correct answer is the contrapositive of a stated conditional. Recognizing which structure is at play lets you know which tools to reach for.
Inference overlaps superficially with several other question types. Keeping the distinctions clean prevents you from applying the wrong strategy.
vs. Necessary Assumption. Inference asks "What follows from these facts?" Assumption asks "What must be true for this argument to work?" The critical structural difference: Inference stimuli typically have NO argument, while Assumption stimuli ALWAYS have one. If there's no conclusion being defended, it's not an Assumption question.
vs. Identify the Conclusion. Main Conclusion asks you to identify a conclusion that is already WITHIN the stimulus. Inference asks you to derive a conclusion that is NOT in the stimulus — the stimulus is just evidence.
vs. Sufficient Assumption. Sufficient Assumption gives you an argument with a gap and asks what would fill it. Inference gives you facts and asks what follows from them. Different direction of inference, different starting material.
vs. RC Inference. A critical distinction: LR Inference (especially Must Be True) requires the answer to be provable from the stimulus — 100% logically guaranteed. RC Inference allows reasonable extrapolation — what the author would likely agree with, not just what is strictly entailed. The same word, two different standards.
Every Inference question uses one of the stem patterns below. Recognizing them instantly tells you which standard of proof applies and cues you into derivation-from-facts mode.