Inference questions ask what must be true given the passage. The one-step rule, why plausible options fail, and how to find the sentence behind each one.
11 min readMonthly & Premium
Quick answer
An inference question asks what must be true given what the passage states, not what is merely plausible. The stem usually reads "It can be inferred from the passage that". Every correct inference is built on one or two specific sentences, and the answer never travels more than one logical step beyond them.
The definition students get wrong
Ask a student what an inference question wants and you will usually hear something like "what the passage suggests" or "reading between the lines". Both descriptions are loose enough to be dangerous, because they license exactly the behaviour that loses points: constructing a reasonable story around the passage and picking the option that fits the story.
Here is the tighter definition. An inference is a statement that is not written in the passage but that must be true if the passage is true. Not likely. Not sensible. Must.
The test of a correct inference option is that you can point at one or two sentences and say: given these, this follows. If you find yourself needing an extra assumption, however obvious the assumption seems, the option is wrong. The extra assumption is where every distractor on this question type lives.
This is also why inference questions are the hardest for strong general readers and the most improvable with technique. A strong reader has a lot of world knowledge available and will happily supply the missing assumption without noticing. A disciplined reader asks where in the text that assumption came from, cannot find it, and moves to another option.
The one-step rule
Correct inference options are one logical step from the text. Exactly one. Almost never two.
A passage says people deprived of REM sleep "struggle with creative problem-solving and have difficulty regulating their moods". The question asks what a person deprived of REM sleep would most likely have difficulty with, and the answer is "finding creative solutions to problems". That is one step: the passage said creative problem-solving, the option says finding creative solutions. Essentially a paraphrase with a thin layer of application over it.
Students often reject that answer for being too easy. It seems like it is just restating the passage, and inference is supposed to be harder than that. So they look for something more sophisticated and select an option about, say, difficulty falling asleep the following night, which involves a real chain of reasoning about sleep pressure and which the passage never establishes.
Resist that instinct. On this exam, the correct inference is usually the least adventurous option that is still not stated verbatim. If an option requires you to be clever, it is probably wrong. If an option makes you think "well, obviously, that is basically what it said", it is probably right.
The practical form of the one-step rule: before you select an option, say out loud in your head "the passage says X, therefore Y". If you cannot fill in X with a specific sentence, or if getting from X to Y needs a "well, presumably" in the middle, keep looking.
Four ways an inference option goes wrong
Failure
What it looks like
How to catch it
Too far
True direction, one step too many. The passage supports a smaller claim and the option extends it.
Ask what extra fact you would need. If you need one, the option is too far.
Plausible but unsupported
A perfectly reasonable statement about the world that the passage never touches.
Try to point at the sentence. If you cannot, it does not matter how true it sounds.
Contradicts a qualifier
The passage says "most" or "does not fail" and the option says "all" or "fails".
Reread the sentence for hedging words. They are usually the load-bearing part.
Stated outright
The option is written in the passage almost verbatim.
An inference question asks for something not stated. A verbatim match is a signal, not a confirmation.
Nearly every distractor on this question type fits one of these. Naming the failure is faster than re-reading the passage.
The procedure, about 50 seconds
1
Identify the anchor
Every inference question is anchored to a specific part of the passage. If the stem names a paragraph, that is your anchor. If it names a subject ("elderly and low-income residents", "goldsmiths in seventeenth-century England"), find where that subject is discussed. Do not begin with the options.
2
Read the anchor sentences and the sentence after
Two or three sentences, read properly. The inference is built out of these. If you skip this step and go to the options, you will be judging plausibility rather than entailment, and plausibility is what the distractors are optimised for.
3
Predict the inference before reading the options
Ask yourself: what follows from this that the passage did not bother to say? Often you will produce something close to the correct option. Even when you do not, having a prediction changes how you read the four choices.
4
Test each option with "the passage says X, therefore Y"
For each surviving option, name the sentence. An option you cannot attach to a sentence is out, regardless of how sensible it is. This eliminates the plausible-but-unsupported family entirely, and that family is roughly a quarter of all distractors here.
5
Prefer the smaller claim
If two options both seem supported, take the one that claims less. Inference answers on this exam are conservative. The bolder option is nearly always the one that travelled a step too far.
Author-attitude inferences
A subtype worth separating out. Some inference questions ask what the author thinks rather than what is factually implied: "The author would most likely agree that", "The passage suggests that the author regards X as".
These are answered from evaluative language, not from content. Look for the places where the author stops reporting and starts assessing: "increasingly argue that", "should be regarded not as a luxury but as", "the single most underestimated constraint", "attracts so little political attention". Those phrases are the author speaking. Everything between them is reporting.
The common error is to infer the author position from the topic. A passage describing the damage done by soil degradation does not automatically imply the author supports mandatory regulation, subsidies, or any particular policy. Passages on this exam are overwhelmingly expository. The author has a position on what is true and important, and almost never a position on what should be legislated. Options that hand the author a policy view are wrong far more often than they are right.
The second common error is to mistake a reported view for the author view. "For centuries, sleep was viewed as a passive state" is a view the passage is about to correct. Students who take the first stated position as the author position get the attitude question exactly backwards, and passages open this way often precisely because it sets that trap.
Inference checklist
Find the anchor before you read the options.
Read the anchor sentence plus the one after it.
Predict the inference yourself first.
For every option, name the sentence it comes from. No sentence, no answer.
Prefer the option that claims less.
Treat hedges as load-bearing: most, some, believe, appears to, does not.
For author-attitude questions, use evaluative language, not topic.
Do not confuse a view the passage reports with the view the passage holds.
Practice: five inference questions
Two of these have a tempting option that is one step too far, and one has an option that is simply stated in the passage. Read all the reasoning, not just the reasoning for your answer.
Street trees deliver measurable benefits only once their canopies are established, which typically takes between eight and fifteen years. During the first three summers after planting, however, a young tree has a root system far smaller than its water demand, and mortality in that window is the largest single loss in most municipal planting programmes. Cities that fund planting but not the three years of irrigation that follow routinely replant the same locations twice within a decade, at a cost that exceeds what the watering would have been.
It can be inferred from the passage that a city that plants street trees but does not water them during establishment will most likely:
The passage states that such cities replant twice within a decade "at a cost that exceeds what the watering would have been". Higher replanting cost than watering cost means higher total spend. One step, from one sentence.This inverts the stated cost comparison. The passage says the replanting cost exceeds the watering cost, so total cost goes up, not down. This option is attractive because skipping a spending line intuitively saves money.Legal compensation appears nowhere. This is the plausible-but-unsupported family: it is the kind of thing that could be true of some city somewhere, and the passage gives you no sentence for it.The passage says the trees die, not that they survive at reduced size. An option that changes the outcome from mortality to stunting contradicts the anchor sentence.
Street trees deliver measurable benefits only once their canopies are established, which typically takes between eight and fifteen years. Municipal budgeting cycles rarely extend beyond four. This mismatch, more than any disagreement about the value of urban canopy, explains why planting programmes are announced far more often than they are completed.
It can be inferred from the passage that the author regards the eight to fifteen year figure as:
The author places the eight to fifteen year figure directly against four year budget cycles and then names the "mismatch" as the explanation for programme failure. The evaluative sentence is the third one, and it tells you exactly what the author thinks the figure means.Nothing suggests the figure is inflated. The author uses it as a premise, not as something to be corrected, and calling it an overestimate would undermine the argument the paragraph is making.The author explicitly sets aside "disagreement about the value of urban canopy" as the explanation. Attributing a negative view of trees to an author who is diagnosing why tree programmes fail reverses the position.Species variation is never raised. This option sounds like scientific caution, which is why it is offered, and the passage gives you no sentence to attach it to.
Decipherment usually depends on a bilingual text, a document that records the same content in a known script and an unknown one. The Rosetta Stone is the famous case, but the principle is general: without a bridge into a known language, an undeciphered script can be catalogued, counted and classified without ever being read. Scholars can determine how many distinct signs a system uses, and therefore whether it is likely alphabetic, syllabic or logographic, purely from the statistics of the surviving inscriptions.
It can be inferred from the passage that a written language whose script was never deciphered:
The final sentence says scholars can determine the number of distinct signs and therefore whether a system is alphabetic, syllabic or logographic, from statistics alone. That is classification without decipherment, which is precisely what the option claims.The passage says the opposite: an undeciphered script "can be catalogued, counted and classified". This option contradicts the anchor sentence directly, and it is tempting only if you stopped reading at "without ever being read".Trade contact is a reasonable line of speculation about bilingual texts and it is nowhere in the passage. This is the plausible-but-unsupported pattern operating on a reader who knows something about ancient history.Survival of materials is not discussed. The passage in fact refers to "the surviving inscriptions", which implies material did survive, so this option runs against the little evidence there is.
Replication is not a courtesy check on published work, it is the mechanism by which a finding becomes knowledge. Researchers examining a decade of results in one subfield reported that studies with small samples and flexible analysis choices were both the most likely to produce striking effects and the least likely to replicate. They noted that the two properties are connected: the smaller the sample, the more the result depends on which analytical path the author happened to take, and the more paths available, the greater the chance that one of them produces a publishable effect.
It can be inferred that the researchers described in the passage would be most concerned about a study that:
The passage names two properties, small samples and flexible analysis choices, and says they combine to produce striking but unreplicable effects. An option containing both properties plus the striking result is the direct application of that stated combination.The passage associates the problem with small samples, so a very large sample points the other way, and a modest effect is the opposite of the striking effects that concerned the researchers.A failed replication is described as part of the mechanism by which knowledge is tested, not as a cause for concern about the failing study. The passage treats replication attempts as the solution, not the problem.Commercial funding is a genuine concern in research generally and it is not mentioned anywhere here. It is the option supplied by a reader own knowledge of research controversies rather than by this passage.
In some languages, word stress falls on a fixed syllable and never moves. Finnish stresses the first syllable of every word, so a listener never needs to learn where the stress goes. In English, stress is neither fixed nor free: it is determined by a set of interacting rules with many exceptions, and it can change the grammatical class of a word, as in the noun REcord and the verb reCORD. Learners of English therefore have to store stress information word by word, an additional burden that speakers of fixed-stress languages do not face in their own.
It can be inferred from the passage that in a language with fully predictable stress:
If stress falls on a fixed syllable and never moves, it cannot be doing contrastive work, because contrast requires it to be somewhere else in the minimal pair. The passage supplies the contrastive case for English (REcord and reCORD) and states that in Finnish the listener "never needs to learn where the stress goes". One step from those two sentences.This is the English situation, stated explicitly in the last sentence, and the passage says speakers of fixed-stress languages do not face it. The option attaches the wrong language type to the burden.Word length is never mentioned. It is a plausible-sounding linguistic claim with no anchor in the text at all.The passage makes a narrow claim about one feature, stress, being easier. Extending that to overall difficulty is the too-far failure: a language can have simple stress and hard everything else, and the passage gives you nothing about the rest.
Practising inference without a test
The drill that moves this question type is not doing more inference questions. It is doing them slowly with a written justification.
Take a passage. For each inference question, write two lines before you look at the options: "The passage says ___" and "Therefore ___". Then answer. Then, for each of the three wrong options, write one line saying which failure it commits: too far, unsupported, contradicts a qualifier, or stated outright.
This takes about four minutes per question, which sounds unaffordable until you notice that you are not training speed, you are training a habit of demanding a sentence before accepting a conclusion. After fifteen or twenty questions the demand becomes automatic and the time collapses back to fifty seconds.
The measure of progress is not your score. It is whether you can name the failure of each wrong option without looking at the explanation. When you can do that reliably, you have stopped guessing between two plausible answers, which is where nearly all inference losses come from.
One closing thought. Inference questions feel like the intellectually serious part of the Reading section, and students give them a corresponding amount of time and anxiety. They are worth one raw point, like everything else. If you have used 50 seconds and you are down to two options, choose the more conservative one and go. The section does not reward the question you agonised over. It rewards the ten you finished.
Monthly & Premium
8 more sections in this lesson
You have read the opening. The rest covers the method in full, with worked examples and practice questions that explain why each wrong answer is wrong.
All 45 lessons: Reading, Listening, Writing and Speaking
Real exam audio and record-yourself speaking drills
A question asking for something that must be true given what the passage states, without being stated itself. Stems usually read "It can be inferred from the passage that" or "The author suggests that".
One logical step, essentially never two. If getting from a passage sentence to the option requires an extra assumption, however reasonable, the option is wrong.
Because plausibility is exactly what distractors are built for. The test is entailment, not reasonableness, so an option you cannot attach to a specific sentence is out no matter how sensible it sounds.
It signals that the answer is a consequence rather than a stated fact. It does not license speculation, and the correct option is still the one that follows directly from the text.
Use the evaluative language in the passage, phrases like "increasingly argue" or "the most underestimated constraint", rather than the topic. And check that the view you are attributing is the author position and not a view the passage reports in order to correct.
About 50 seconds. If you are down to two options at that point, choose the one that claims less and move on. It is worth one raw point like every other item.