Inference questions flip the usual task. There's no argument to critique, just a set of statements you accept as true, and five choices, one of which is proved (or most strongly supported) by those statements. You're not evaluating reasoning; you're doing the reasoning.
The core discipline is restraint. Real-world "inference" means a reasonable guess; LSAT inference means something the statements actually establish. The gap between those two standards is where wrong answers live. An answer can be sensible, likely, and consistent with everything stated, and still wrong, because consistent isn't proved.
Two practical consequences follow. First, soft answers win: claims with "some," "at least one," or "not all" are easier to prove than claims with "most," "all," or "never," so the correct answer skews modest. Second, the power move is combining statements, the right answer usually falls out of linking two or three sentences, often through a shared term.
Stem language that signals this type:
Scan for the term that appears in two different sentences, that's usually the joint the test wants you to connect. Chain conditionals, combine a general rule with a specific case, or intersect two quantified claims.
Between two answers pointing the same direction, the softer one is more provable. "Some managers attended" survives scrutiny that "most managers attended" fails.
For every answer ask: which sentences establish this? If you can't point to the support, the answer is speculation, even if it's obviously true in real life. Outside knowledge is a liability here.
"Must be true" demands airtight proof. "Most strongly supported" tolerates one small step beyond the text, but only one, and smaller than you think.
An original question written in the LSAT style, not from a real test.
All of the ceramics in the museum's new wing were donated by local collectors. Some of the ceramics in the new wing are more than four hundred years old. Every ceramic more than four hundred years old is fragile.
If the statements above are true, which one of the following must be true?
Pick an answer to check your reasoning.
"All," "none," "always," and superlatives demand near-total proof. Unless the stimulus itself speaks in absolutes, extreme answers are almost always wrong.
The classic wrong answer says what a reasonable person would think next, a prediction or recommendation the statements suggest but never establish. Plausibility is not proof.
Given "every A is B," the trap answer claims "every B is A." If the stimulus contains conditionals, expect at least one reversal or negation among the choices.
Take the free diagnostic to see exactly where your points are going, or talk it through with a 170+ tutor, free, for 15 minutes.
Must-be-true answers are fully proved by the statements. Most-strongly-supported answers may take one very small step past the text. In both cases the correct answer is dramatically better supported than the other four.
Because weak claims need less proof. "Some" requires one example; "most" requires a majority; "all" requires everything. The test exploits this by writing attractive-but-extreme wrong answers.
When the stimulus is conditional-heavy, a quick chain diagram pays off. When it's quantified everyday prose, translating each claim into plain terms and hunting for the shared-term link is faster.
Yes, a large share of RC questions are inference tasks scoped to a passage. The same standard applies: supported by the text, not merely consistent with it.
Identify the conclusion the whole argument exists to sell.
Bridge general rules and specific cases, in both directions.
Find the unstated premise the argument can't live without.
Inference's tightest case, with 11 drills.
Or head back to all question types · What is the LSAT?
Last updated July 9, 2026 · by the theLSATguru team