LNAT inference questions: why 'must be true' and 'most likely' demand different reading moves

An LNAT inference question is, in plain language, a multiple-choice item whose correct answer is a statement the passage guarantees, hints at, or leaves room for, depending on the stem. The phrase 'can be inferred' is doing far more work in that stem than candidates usually realise. It is not a synonym for 'is suggested somewhere' or 'could be argued'; it is a contract between the passage and the candidate, and the right answer is the option that respects that contract exactly. Most candidates reading this have probably been told, in one form or another, that inference questions reward 'reading between the lines'. The advice is true but useless without the operative definitions: what counts as a line, what counts as between, and what counts as a fair inference. This article is a working tutor's map of those definitions, built around the three sub-types of inference stem that recur in Section A of the LNAT and around the linguistic signals that separate the answer from its most plausible sibling, the distractor.

Section A of the LNAT is a 42-question multiple-choice paper built around 12 argumentative passages, each with 3 or 4 items. Roughly a third of those items, in practice, are inference questions. The other item families in Section A include explicit comprehension, application, tone, and assumption questions, all of which intersect with inference work at the edges. Candidates preparing through the LNAT (Hukuk) Hazırlık Kursu & Özel Ders programme on the brand site will recognise that Section A is scored on a 0–42 scale and that every correct item carries equal weight, which is one reason why inference accuracy has an outsized effect on a candidate's percentile. A reader who loses two inference items per passage and gains them back elsewhere does not recover the score; the sub-types behave differently and reward different reading moves.

Stems first: why the verb in the question does all the work

The single most important habit to build for LNAT inference questions is to read the stem before reading the passage. Candidates who read the passage first and only then look at the stem are reading twice, in the worst possible order: they have already decided what the passage is about, and they are now searching the option list for whatever vaguely matches. The stem is the contract; the passage is the evidence; the option list is the test. Read the contract, then read the evidence, then test the options against the contract. That three-step loop is the entire method, and it is the method that the diagnostic assessment in the brand's programme is built to surface.

Three sub-stems dominate LNAT inference items, and they are not interchangeable. The 'must be true' stem, sometimes phrased as 'which of the following is most strongly supported by the passage' or 'which statement is best supported', demands an answer that is logically forced. If the passage is silent on the option's content, the option is wrong, even if it is true in the real world. The 'most likely to be true' or 'which can most plausibly be inferred' stem allows a softer contract: the answer must be consistent with the passage and go slightly beyond what is explicitly said, but it must not contradict the passage. The 'author would most likely agree' stem is a disguised inference question; it asks what the author's belief system would licence, which is in turn inferred from the author's argumentative moves. A common mistake is to treat these three stems as if they were one. They are not, and the candidate who collapses them pays for it on the distractor lines.

Word choice inside the stem is also a signal. 'Could', 'might', and 'most likely' weaken the contract; 'must', 'necessarily', and 'is best supported' tighten it. The same passage can host two different inference items, one with each contract, and the right answer for one is the wrong answer for the other. In my experience as a tutor, this is the single largest source of avoidable error in inference work. Candidates treat the passage as a fixed object and the stem as decoration; in fact, the stem determines the threshold of evidence, and the passage is read through that threshold.

For a worked example, take a passage arguing that mandatory sentencing laws reduce judicial discretion and that this trade-off is acceptable in cases involving violence. A 'must be true' item might ask which of the following the author would accept. The right answer is usually the narrowest one: 'Discretion is incompatible with consistent sentencing'. A 'most likely' item on the same passage could legitimately go further and ask what the author would think of a non-violent mandatory regime. The right answer there is more interpretive, but it must still respect the author's logic. Treating the two as the same question produces a wrong answer with high confidence, which is the worst possible outcome on a multiple-choice paper.

Scope: how to keep an inference from outgrowing the passage

Scope is the second pillar of inference work, and it is where most candidates lose marks without realising it. An inference's scope is the set of topics, agents, and timeframes the passage is actually about. The most common scope error is to import a claim from outside the passage into the answer. Candidates who have read widely on the topic of the passage will reach for what they already know, and the right answer is the one that survives without that knowledge. The other scope error is the mirror image: under-scoping an inference by limiting it to a single sentence in the passage when the passage as a whole is doing the work. The right scope is almost always the passage, sometimes a paragraph, occasionally a sentence, and never smaller than a phrase.

A practical scope method I teach in the brand's preparation programme works as follows. Read the stem. Read the relevant section of the passage. Before looking at the options, write a one-sentence version of the answer in your own words. That sentence is the scope, and the right option is the one that matches it without expanding it. If your one-sentence version is longer than 25 words, the stem is probably a 'must be true' item and you are over-scoping. If your version is shorter than 12 words and feels like a fragment, the stem is probably a 'most likely' item and the option will be allowed to add a connecting word or two. The candidate who can write a one-sentence answer before reading the options has, in my experience, the best chance of staying inside scope on the distractor lines.

Scope also has a quantitative edge. On the LNAT, candidates have roughly 95 minutes for Section A, which works out to about 32 minutes per group of 12 passages or around 2 minutes 15 seconds per item. A scope error is expensive in two ways: it forces a re-read of the passage section, which is usually impossible inside the 2-minute budget, and it tips the candidate toward a confident wrong answer. Build the discipline of writing the one-sentence version; the time spent is recovered threefold when the options fail to seduce you into a scope-expanded distractor. The diagnostic assessment in TestPrep İstanbul's LNAT (Hukuk) Hazırlık Kursu & Özel Ders surfaces this habit by timing each item and flagging scope-inflated answers.

Signal words and the three linguistic fingerprints of a correct inference

Once the stem and the scope are in place, the candidate is reading the options with a sharper eye, and that is where signal words do most of their work. Three linguistic fingerprints separate the answer from its distractor siblings. The first is modal restraint. Correct inferences in 'most likely' items use modal verbs that mirror the stem: 'might', 'could', 'tends to'. Distractors in those items use flat declaratives: 'is', 'will', 'always'. A candidate who is calibrating to 'most likely' and sees an option that drops the modal is looking at a distractor. The second fingerprint is causal economy. Correct inferences describe a single causal move, sometimes two, never three. Distractors often chain three or four causal steps that the passage only half-supports. The third fingerprint is agent alignment. Correct inferences name the agent the passage names; distractors swap the agent for a related one.

Let us work a small example. The passage argues that increased screen time in adolescents is associated with reduced sleep quality and that this association is not yet established as causal. A 'most likely to be inferred' stem gives us four options. Option A: 'Screen time causes insomnia in teenagers.' Option B: 'Reduced sleep quality in adolescents is associated with screen time use.' Option C: 'Teenagers who reduce screen time will sleep better.' Option D: 'Schools should ban phones after 9pm.' Option A drops the modal, asserts causation the passage denies, and is a distractor. Option B mirrors the passage's hedged language and is the answer. Option C asserts a forward causal claim the passage does not licence. Option D is a recommendation the passage never makes; it is the classic outside-scope distractor. Three of the four options fail for distinct, identifiable reasons, and a candidate who knows the three fingerprints can name the failure mode of each distractor without guessing.

The fingerprints also have a tempo function. Once a candidate internalises them, the option list reads as a sorting exercise rather than a deliberation. Option A fails on modal restraint; option B is the answer; option C fails on causal economy; option D fails on scope. The candidate has not weighed the options against each other; the candidate has tested each option against a rule. In a 2-minute-per-item budget, that is a 30-second saving per item, and over 14 inference items in Section A that is a 7-minute reserve, which is enough to spend on the genuinely hard items where the four options are all plausible. Build the rules; spend the saved time on the items that need it.

Distractor anatomy: why three of four options feel right

The LNAT's inference distractors are not random. They are engineered to share surface vocabulary with the passage while committing one of the three failures above. The first distractor type is the hedge-dropper, which removes a modal or qualifier from the passage's language and turns a hedged claim into a flat one. The second is the scope-bloater, which adds a noun or agent that the passage does not name. The third is the causal-chain extender, which adds a step the passage only implies. The fourth, rarer but present, is the tone-shifter, which moves a claim from descriptive to evaluative or vice versa.

For the candidate, the practical question is how to keep these from feeling right. The simplest counter-move is the negation test. Read the option. Negate it. Does the negated version still respect the passage? If yes, the option is either too weak to be the answer (in a 'must be true' item) or a sibling distractor (in a 'most likely' item). If the negated version contradicts the passage, the option is consistent with the passage and worth a closer look. The negation test is a 5-second operation, and it eliminates a large share of scope-inflated distractors in one pass. It is not a substitute for reading the stem and the scope carefully, but it is a reliable second filter.

For 'must be true' items, the secondary test is the necessity check. If the option were false, would the passage still be true? If yes, the option is not forced by the passage, and it is a distractor. If no, the option is necessary to the passage and is the answer. The necessity check is the most discriminating test for the 'must be true' stem, and it is the test most candidates skip because it requires holding the passage and the option in mind at the same time. Practise it. In the brand's preparation programme, the inference drill blocks are designed so that candidates do the necessity check on every 'must be true' item for the first 20 minutes and then fall back on it whenever a 'must be true' item gives them trouble on the timed paper.

Three sub-types of inference stem, side by side

A compact view of the three sub-types helps, especially for candidates who are mid-preparation and trying to triage their practice sets.

The table is not a key to be memorised. It is a triage tool. When a candidate sits with an inference item, the table tells them which of the three tests (necessity, negation, or agent-alignment) is the most discriminating for that stem. That triage is, in my experience, what separates the candidates who finish Section A with 5 minutes to spare from the ones who run out of time on passage 11.

Pacing, practice design, and the score ceiling problem

Section A of the LNAT is not a comprehension test in the conventional sense. It is an inference test with a comprehension wrapper. The candidates who hit the score ceiling are not the ones who understand the passage most thoroughly; they are the ones who triage the items by stem and apply the right test to each. A useful practice design is one where, for the first 30 minutes of a practice block, the candidate pauses after each item and names the stem sub-type out loud. This is awkward. It is also the fastest way to build the triage habit, because it forces the stem into working memory before the option list is read.

The next design element is a 4–3–3 practice ratio. For every 10 inference items, four should be 'must be true', three should be 'most likely', and three should be 'author would agree'. That ratio roughly tracks the distribution candidates see in the live paper. Skewing practice toward a single sub-type is a common mistake that produces a candidate who is sharp on that sub-type and fragile on the others. The diagnostic assessment in the brand's programme records sub-type accuracy separately, which makes this skew visible to both tutor and candidate.

Finally, the practice should include a pacing review every five passages. Count the time spent on each item in the first five passages. The median time per item should be inside the 2-minute 15-second budget; the upper quartile should be inside 3 minutes. If the upper quartile is above 3 minutes, the candidate is over-reading the hard items, and the right intervention is to triage them as 'return to' and move on, picking them up in the last 10 minutes of the paper. The LNAT does not penalise unanswered items any harder than it penalises wrong ones, but it does penalise a candidate who spends 4 minutes on a single inference item and leaves three easier items with 30 seconds each. Pacing is a score lever, and the lever is mechanical.

Common pitfalls and how to avoid them

The pitfalls below are the ones I see most often in candidates who are otherwise strong readers. Each is paired with a concrete counter-move, and the counter-moves are designed to be cheap in time and reliable in effect.

• Reading the passage before the stem. The candidate reads 350 words, then looks at a stem that asks about a 40-word claim at the end of the passage. The counter-move is a hard habit: stem first, then passage, then options. The reading order takes 20 seconds longer on the first item and saves time on every subsequent item.
• Importing outside knowledge. The candidate knows the topic (climate, AI, criminal justice) and reaches for a fact they remember from elsewhere. The counter-move is the outside-knowledge test: could a reader who has never encountered this topic still choose this option? If no, the option is outside the scope and is a distractor.
• Modal-dropping. The candidate picks a flat declarative when the stem asked for a hedged claim. The counter-move is to read the stem, name the modal it uses, and reject any option whose modal is stronger than the stem's.
• Over-scoping the answer. The candidate writes a one-sentence version that needs 30 words to express. The counter-move is to compress that sentence to 15 words and discard the modifier; the modifier is usually the distractor.
• Treating 'author would agree' as a content question. The candidate looks for the option whose claim is true in the world. The counter-move is to ask instead what the author would say; sometimes the author would say something the world disagrees with, and the right answer is exactly that.
• Skipping the necessity check on 'must be true' items. The candidate reads the option, feels it is consistent with the passage, and selects it. Consistency is not enough; the option must be forced. The counter-move is the 5-second necessity check described above.
• Running out of time on the last passage. The candidate leaves passage 11 or 12 with 8 minutes and tries to read two passages cold. The counter-move is triage at minute 80: flag the two hardest items on each remaining passage, read the stems first, and answer the ones whose stems are 'most likely' before tackling the 'must be true' items. The 'most likely' items are usually faster and carry the same 1-mark weight.

Building a six-week inference plan that actually moves the score

A six-week plan that targets inference accuracy has four phases. In week 1, the candidate takes an untimed inference-only diagnostic of 30 items, mixed across the three sub-types, and records sub-type accuracy. The purpose is to identify the weakest sub-type; in most cases it is the 'author would agree' sub-type, because it requires holding the author's argumentative moves in mind. In week 2, the candidate drills 20 items of the weakest sub-type, untimed, with the necessity or negation test applied to every item. In week 3, the same 20 items are re-done timed, at 2 minutes 15 seconds per item, with the aim of holding accuracy above 80%. In week 4, the candidate mixes the three sub-types at the live distribution (4–3–3) and times every block. In week 5, the candidate sits full Section A papers and applies the triage habit. In week 6, the candidate takes a final inference-only diagnostic and compares it with the week 1 baseline. The candidates who see a 15–20 point lift on the 30-item diagnostic are the ones who have done weeks 2 and 3 in full; the candidates who skip those weeks plateau by week 4 and have nothing left to gain in week 5.

The plan is mechanical, but the discipline required is real. Most candidates reading this will identify the weakest sub-type, will not drill it twice, and will not see the lift. The plan rewards repetition more than brilliance. In a 6-week window, a candidate who drills the weak sub-type twice a week for 90 minutes each session will outscore a candidate who reads widely and practices irregularly. The diagnostic assessment built into the LNAT (Hukuk) Hazırlık Kursu & Özel Ders programme is calibrated to flag the candidates who are about to make that mistake, and the tutor's role is to redirect them into the drill blocks before week 4.

What to log after every practice block

Candidates who improve fastest log three numbers per block: items attempted, items correct, and median time per item. They also log, by sub-type, the items they got wrong and the failure mode of each (modal-drop, scope-bloat, causal-chain extend, tone-shift). The log is the source of truth; the candidate's memory of how the block went is unreliable. After two weeks of logging, the pattern is visible: the candidate who is dropping points on tone-shift distractors in 'author would agree' items needs a different intervention from the candidate who is dropping points on causal-chain extenders in 'most likely' items. A single practice block tells the candidate nothing; ten practice blocks tell the candidate everything.

Where inference questions sit inside the wider LNAT score

Section A of the LNAT is scored on a 0–42 scale, with one mark per correct item and no negative marking. The essay in Section B is read separately by admissions tutors and does not contribute to the Section A score. Within Section A, the inference sub-types are not weighted differently, but the candidate's pattern of accuracy across them is the most reliable predictor of the Section A percentile. A candidate who is 90% accurate on 'must be true' items, 80% on 'most likely', and 70% on 'author would agree' will outscore a candidate who is 80% on all three, because the harder sub-types carry the same mark weight and reward the same accuracy investment more. The skill of inference is therefore not a single skill; it is three skills, and the candidate who improves the weakest one moves the percentile the most.

There is also a soft transfer effect into Section B. The essay is graded on argument architecture, evidence handling, and counterargument, all of which are the same cognitive moves that the 'author would agree' inference sub-type tests. The candidate who has spent six weeks doing necessity checks on inference items is the candidate who, in the essay, will not over-claim the strength of their own evidence. That is not a coincidence; the inference sub-type and the essay are testing the same intellectual virtue, which is the discipline of not saying more than the warrant allows. The diagnostic in the brand's programme measures this transfer directly, and the candidates who show it are the ones whose Section A and Section B scores move together over the six-week plan.

Conclusion and next steps

LNAT inference questions are a contract exercise. The stem names the contract, the passage provides the evidence, and the option list tests whether the candidate has respected both. Candidates who build the habit of reading the stem first, who apply the necessity check to 'must be true' items and the negation test to 'most likely' items, and who drill the weakest sub-type twice a week for six weeks will move their Section A percentile more reliably than candidates who read widely and practice irregularly. The three fingerprints (modal restraint, causal economy, agent alignment) and the three sub-types (must be true, most likely, author would agree) are the spine of the work; everything else is practice design and pacing discipline. TestPrep İstanbul's diagnostic assessment is a natural starting point for candidates who want a baseline sub-type accuracy reading before they commit to the six-week plan.

FAQ

How are LNAT inference questions different from comprehension questions?

Comprehension questions ask what the passage explicitly says. Inference questions ask what the passage guarantees, hints at, or licences, depending on the stem. The difference is in the contract: a comprehension question is a closed contract, and an inference question is an open contract with a defined threshold.

Should I read the passage before the question on LNAT inference items?

No. Read the stem first, then the relevant section of the passage, then the options. The stem tells you the threshold of evidence; reading the passage first forces you to re-read it when the stem's threshold does not match your first reading.

How long should I spend on each inference item in Section A?

Budget around 2 minutes 15 seconds per item, with an upper quartile of 3 minutes. If an item is going past 3 minutes, triage it as a return-to and move on; the mark weight is the same as for an easier item, and pacing is a score lever.

Which inference sub-type should I practise first if I have three weeks?

The 'author would agree' sub-type is usually the weakest and the most discriminating. In my experience, candidates who drill that sub-type for 90 minutes twice a week for three weeks see the largest lift in sub-type accuracy, and the lift transfers into the 'most likely' sub-type as a side effect.

Does practising inference questions help the LNAT essay too?

Yes, indirectly. The essay rewards candidates who do not over-claim their evidence, and the necessity check built into inference work is the same intellectual discipline. Candidates who drill inference items for six weeks typically produce tighter Section B arguments as a side effect, though the transfer is not automatic and benefits from explicit feedback on the essay itself.

Frequently asked questions