Critical Reasoning — Inference, Assumption and Tone

Most of a CAT reading-comprehension set is not asking what the passage said. It is asking what follows from it, what it must be assuming, what would weaken it, and what the author is doing in a particular paragraph. These are reasoning questions that happen to be attached to a passage, and they have a property worth exploiting: the wrong options are usually wrong for one of a small number of reasons. Learn those reasons and you stop choosing between four plausible sentences and start eliminating on stated grounds. The commonest of those grounds is not falsehood — it is strength. An option can be perfectly reasonable, entirely consistent with the passage, and still wrong because the passage does not go that far.

Inference — What Must Be True, Not What Is Likely

An inference question asks what the passage commits you to. The test is not "is this reasonable?" but "could this be false while everything in the passage stays true?" If it could, it is not an inference — however sensible it sounds.

  • The safest inference is usually the dullest one. Options that feel like they teach you something new are usually going beyond the passage. The correct answer often reads as an obvious restatement — that is a feature, because a restatement cannot fail the "could it be false?" test.
  • Watch the quantifier. "Some economists argue" does not license "economists argue"; "in most years" does not license "always"; "one study found" does not license "research shows". CAT builds a great many wrong options purely by widening a quantifier, and nothing else in the sentence changes — which is why the option looks right.
  • A cause is not established by a correlation. If the passage says two things rose together, an option saying one caused the other is an addition. Unless the passage argues for the mechanism, the causal option is wrong however natural the causal story is.
  • Worked example. "Between 1990 and 2010, the share of household income spent on food fell in every state for which data exist. Over the same period, the share spent on health and education rose. Whether households were choosing to spend differently, or simply had less choice as these services moved from public provision to private purchase, the data cannot say." Which follows? Not "households became wealthier" — the passage gives shares, not amounts, and a falling food share is compatible with falling income. Not "privatisation raised health spending" — the passage explicitly refuses to choose between explanations. What does follow is narrow and dull: the composition of household spending changed over those two decades. That is the inference, and its dullness is the evidence that it is safe.

Assumption — the Gap the Argument Needs

An assumption is something the argument needs to be true but never says. The reliable way to find one is the negation test: negate the candidate option and ask whether the argument collapses. If it does, that was the assumption. If the argument survives, the option was merely relevant.

  • Worked example. A city plans to cut traffic congestion by widening its main road from four lanes to six. Officials expect journey times to fall by a fifth. One assumption: the number of vehicles using the road will not rise substantially once it is widened. Negate it — suppose the wider road attracts many more vehicles — and the expected saving disappears. So it is an assumption. Contrast with "the road surface is in good repair": negate it and the plan still makes sense, so that is relevant but not assumed.
  • An assumption is never stated in the passage. If you can point to the sentence, it is a premise, not an assumption — and an option that quotes the passage back at you is a common wrong answer in this question type.
  • Weaken and strengthen are the same machinery, run in either direction. To weaken an argument, attack its assumption; to strengthen it, confirm the assumption. So finding the assumption first answers all three question types, which is why it is worth the extra fifteen seconds.
  • A weakener need not disprove anything. It only has to make the conclusion less likely. Options that try to demolish the argument outright are often too strong to be the intended answer, while a modest, specific counter-consideration is exactly right.

Purpose and Tone — What the Author Is Doing

  • "The author mentions X in order to…" is a question about function, not content. The answer is a verb: to illustrate, to concede, to qualify, to rebut, to introduce a contrast. Read the sentence before and after X — the function of a sentence lives in its neighbours, not in itself.
  • A concession is not agreement. "Admittedly, the early results were poor" inside a supportive argument does not make the author a critic. Authors routinely state the opposing case in its strongest form before answering it, and a candidate who reads that paragraph alone will get the tone question exactly backwards.
  • Tone is usually moderate. "Sceptical", "measured", "cautiously optimistic" beat "outraged", "dismissive" and "euphoric" almost every time. Published academic and journalistic prose is rarely extreme, and CAT selects from exactly that register — so an extreme tone option is a distractor unless the passage is unmistakably heated.
  • "Which of the following would the author most likely agree with?" is an inference question in disguise. Apply the same test: could the author hold the passage's position and reject this statement? If yes, it is not the answer.
🎯 Why elimination beats selection here
All four options are written to be attractive; only one is defensible. Choosing the "best" one invites you to compare their appeal, which is the comparison the paper-setter designed. Rejecting three on stated grounds — too strong, added a cause, widened a quantifier, contradicted by paragraph two — leaves the answer standing without your having to like it.

Key takeaways

  • An inference must be true if the passage is true. Test each option by asking whether it could be false while the passage stays true.
  • The safest inference is usually the dullest. An option that teaches you something new has probably gone beyond the passage.
  • "Some", "often" and "one study" do not license "all", "always" and "research shows". A widened quantifier is CAT’s commonest wrong option.
  • A correlation stated in a passage never licenses a causal option. If the mechanism is not argued, the cause was added by the option.
  • Find an assumption by negating the option: if the argument collapses, that was the assumption; if it survives, the option was merely relevant.
  • A weakener only has to make the conclusion less likely. An option that tries to demolish the argument is usually too strong to be the answer.
  • A concession is not agreement. Authors state the opposing case at its strongest before answering it, so never judge tone from one paragraph.
  • Eliminate on stated grounds rather than choosing what feels closest. All four options are written to be attractive; only one is defensible.

Practice questions (8)

Attempt each one before opening the answer. Every explanation names the tempting wrong option as well as the right one, because that is where marks are lost.

  1. Passage: "Between 1990 and 2010, the share of household income spent on food fell in every state for which data exist, while the share spent on health and education rose. Whether households chose to spend differently, or had less choice as these services moved from public provision to private purchase, the data cannot say." Which can be inferred?

    1. Households became wealthier over the period
    2. The composition of household spending changed over the period
    3. Privatisation of health and education raised household spending on them
    4. Food became cheaper relative to health and education
    Show answer

    Answer: B — The composition of household spending changed over the period

    Option B restates what the two reported shares establish, and it cannot be false while the passage is true — which is the test. Option A does not follow: these are shares, not amounts, and a falling food share is compatible with falling income. Option C is the explanation the passage explicitly declines to choose. Option D adds a price story the passage never mentions — shares can move because of quantities, prices or income, and nothing here distinguishes them.

  2. Argument: "A city will widen its main road from four lanes to six, and expects journey times to fall by a fifth." Which is an assumption of this plan?

    1. The road surface is currently in good repair
    2. The number of vehicles using the road will not rise substantially after widening
    3. Public transport in the city is inadequate
    4. Widening the road will cost less than building a new one
    Show answer

    Answer: B — The number of vehicles using the road will not rise substantially after widening

    Apply the negation test. Negate B — many more vehicles use the wider road — and the expected saving in journey times vanishes, so the argument depends on it. Negate A, C or D and the plan still stands: a road can be widened whatever its surface, whatever the buses are like, and whatever the alternative would have cost. Those options are relevant to the decision but not assumed by the argument, which is the distinction this question type is built on.

  3. A passage states: "One survey of 400 firms found that those offering remote work reported lower attrition." Which option is wrong specifically because it widens a quantifier?

    1. A survey found an association between remote work and lower reported attrition
    2. Research shows that remote work reduces attrition
    3. Some firms offering remote work reported lower attrition
    4. The survey covered 400 firms
    Show answer

    Answer: B — Research shows that remote work reduces attrition

    Option B commits two upgrades in six words: "one survey" becomes "research shows", and an association becomes a reduction — a quantifier widened and a cause added. Options A and C stay at the passage's strength, and D is a stated fact. This is the single most productive elimination in CAT verbal: read every option for the words that quietly grow.

  4. An author writes: "Admittedly, the scheme’s first two years produced little. But the design flaw it revealed has since been corrected, and the evidence from the third year is encouraging." The concession in the first sentence functions to:

    1. Signal that the author opposes the scheme
    2. Acknowledge the strongest objection before answering it
    3. Introduce a new topic unrelated to the scheme
    4. Show that the evidence is inconclusive either way
    Show answer

    Answer: B — Acknowledge the strongest objection before answering it

    "Admittedly … But …" is the standard shape of a concession followed by a rebuttal: the author grants the objection at its strongest and then removes its force. The author's position is on the far side of the "but", so this is a supportive passage, not a critical one. Option A is the mistake this construction is designed to produce, and it is why tone should never be judged from a single sentence.

  5. An argument concludes that a new bus lane reduced average commute times, citing a fall in times after it opened. Which most weakens the argument?

    1. The bus lane cost more than the city had budgeted
    2. A large factory on the route closed in the same month the lane opened
    3. Some commuters say the lane has made cycling less safe
    4. Commute times also fell in a nearby city that built no bus lane
    Show answer

    Answer: B — A large factory on the route closed in the same month the lane opened

    The argument needs the lane to be the cause of the fall. Option B supplies a rival cause operating at exactly the same time and on exactly the same route, which is the most direct attack available. Option D is a genuine weakener too — a general trend would explain the fall without the lane — but it is weaker here because a nearby city is a looser comparison than the same route in the same month. Options A and C are about whether the lane was a good idea, which is a different question from whether it caused the fall.

  6. A passage carefully reviews evidence on both sides of a claim and ends without endorsing either. The author’s tone is best described as:

    1. Dismissive
    2. Measured
    3. Indifferent
    4. Polemical
    Show answer

    Answer: B — Measured

    "Measured" describes an author who weighs evidence and declines to overstate a conclusion — which is what the passage does. "Indifferent" is the tempting near-miss and is wrong: careful review is the opposite of not caring, and an indifferent author would not have reviewed the evidence at all. "Dismissive" and "polemical" are extremes, and extreme tone options are rarely right because published prose of the kind CAT selects is rarely extreme.

  7. A passage reports that two variables rose together over a decade and offers no mechanism. Which option must be rejected, and on what ground?

    1. "The two variables moved together over the decade" — too narrow
    2. "The first variable drove the increase in the second" — a cause was added
    3. "The relationship between the variables deserves study" — overstated
    4. "Data on both variables exist for the decade" — contradicted
    Show answer

    Answer: B — "The first variable drove the increase in the second" — a cause was added

    A passage that reports co-movement without a mechanism supports no causal claim, so option B adds something the passage does not license — the ground is addition, not falsehood, and the causal story may well be true. The other three pair a defensible statement with a mislabelled ground: the first is an accurate restatement rather than too narrow, the third is a mild claim rather than an overstatement, and the fourth is implied by the passage rather than contradicted.

  8. You have narrowed an inference question to two options and sixty seconds remain in the section. The better method is:

    1. Choose the option that sounds more insightful about the topic
    2. Find the word where the two differ and check which one the passage licenses
    3. Choose the longer option, since inferences need qualification
    4. Re-read the whole passage from the beginning
    Show answer

    Answer: B — Find the word where the two differ and check which one the passage licenses

    Two surviving options almost always differ in one load-bearing word — a quantifier, a connective, or a causal verb — and that word is what the question is testing. Locating it takes seconds and gives a defensible decision. Option A inverts the correct instinct: the more insightful-sounding option is usually the one that went beyond the passage. Option D spends sixty seconds re-reading text you have already understood, when the decision lives in the options.