Critical Thinking Test: Practice, Examples and Preparation

A critical thinking test measures how well you evaluate information, arguments and conclusions. These tests are common in law, consulting, management, policy, graduate schemes and professional roles where decisions must be based on evidence rather than instinct. Unlike a standard verbal reasoning test, a critical thinking test does not simply ask whether you understood a passage. It asks whether a conclusion follows, whether an assumption is hidden, whether an argument is strong, or whether a statement is supported by the evidence.

The best-known example is the Watson Glaser Critical Thinking Appraisal, published by Pearson TalentLens. Pearson describes Watson Glaser as measuring the ability to analyse arguments, evaluate information and make objective decisions. Candidate-facing practice material for Watson Glaser commonly separates the test into skills such as inference, recognition of assumptions, deduction, interpretation and evaluation of arguments. Those five labels are useful even when you are preparing for a different critical reasoning test, because they describe the main mental operations employers want to see.

Most marks lost on a critical thinking test are not lost to weak reasoning. They go because the candidate answers the question they would ask in ordinary life instead of the one on the screen. The Inference subtest exposes this within a few items: you read a short passage, you are given a statement that goes beyond it, and you rate that statement on a five-point scale running true, probably true, insufficient data, probably false, false. The whole subtest lives in the gap between probably true and insufficient data, and candidates who know something about the subject fill that gap from memory. Every gap filled that way costs a mark.

The other four subtests tighten the same screw in different ways. Each has its own response scale, its own standard of proof, and its own rule about what the text licenses you to say. Working out which rule is live in which subtest is most of the preparation.

These ten questions are in Watson-Glaser format. The Watson-Glaser Style Practice Pack runs these question types at full length — 5 timed tests, 150 questions, a worked explanation on every one.

The five formats inside a critical thinking test

Inference: a five-point rating, not a true or false call

You are given a passage presented as fact, then a statement not in it, and you rate that statement true, probably true, insufficient data, probably false or false. This is the only subtest with an ordinal scale rather than a binary one, and the scale is not a confidence slider. True means the statement follows with nothing added. Probably true means the passage makes it more likely than not. Insufficient data means the passage neither supports nor undermines it, whatever you personally believe. The permitted inputs are the passage plus knowledge every adult shares, and that second clause is the trap: specialist knowledge is not shared knowledge, so subject experts over-infer more than novices do.

Recognition of Assumptions: does the statement depend on it

A statement is given, then a proposed assumption, and you answer made or not made. There is no body of facts to weigh, which is the mechanical difference: you are not asking whether the assumption is reasonable or true, only whether the statement collapses without it. Negate it and reread the statement; if the statement still stands, the assumption was not made. Wrong answers come from accepting plausible background beliefs as assumptions, and from accepting assumptions phrased more strongly than the statement needs, with words such as only or always.

Deduction: necessity, with no room for likelihood

Premises are given, then a conclusion, and you answer conclusion follows or does not follow. You must treat the premises as true even where they are plainly false about the real world, which separates this from Inference. Nothing probabilistic counts: a conclusion that is very likely given the premises still does not follow. The recurring wrong answers are quantifier slips, such as reading all A are B as all B are A, and scope errors such as reading not all as none.

Interpretation: beyond reasonable doubt

A short body of information is given, then a conclusion, and you answer follows or does not follow. This sits between Deduction and Inference: it permits a small inductive step that Deduction forbids, but demands far more than the balance of probability behind probably true. Wrong options generalise from a described group to a wider population, turn a proportion into a headcount without giving a base, or convert a comparison such as more than last year into an absolute such as high.

Evaluation of Arguments: strong or weak, not true or false

A question of public debate is given, then a single argument for or against it, and you answer strong or weak. This is the only subtest judging quality rather than entailment. An argument is strong only if it is both directly relevant to the question asked and important rather than trivial, so an argument can be entirely true and still weak. Weak arguments are typically true statements about a neighbouring question, appeals to how many people hold a view, or restatements of the question offered as their own reason.

Scoring and timing, and what they mean for strategy

The usual model for this family is raw scoring: one mark per correct response, nothing deducted for a wrong one. There is no partial credit on the five-point Inference scale, so rating a statement probably true when the key says true earns exactly what false would earn, which is nothing. Under raw scoring the instruction is simple: leave nothing blank, because an unanswered item and a wrong item are worth the same.

Some versions apply formula scoring instead, where a fraction of a mark is deducted for each wrong answer and nothing is deducted for an item left blank. Where that model is in force, guessing at the buzzer lowers your score. On the binary subtests this matters more than it first appears, because a coin flip between two options has a negative expected value once wrongs are penalised, so a blank is genuinely worth more than a guess. Confirm which model applies before you sit, because the two models call for opposite behaviour in the last minute.

Timing varies more than format does. A widely used short form runs 40 items against a single 30 minute clock, roughly 45 seconds an item across all five subtests. Longer forms of around 80 items give proportionally more time. Some versions impose per-section timers, so time saved on Deduction cannot be spent on Evaluation of Arguments. Some are an indefinite bank against a clock, where nobody is expected to finish and accuracy per item attempted carries the score. Adaptive versions adjust difficulty as you go and usually forbid returning to an earlier item, which kills the flag-and-come-back habit that works on fixed forms.

A worked example

Passage: a regional rail operator replaced the ticket machines at 14 of its 30 stations last year. Across those 14 stations, recorded ticketless travel in the following twelve months fell. Across the remaining 16 stations it was unchanged. The operator has not published maintenance costs for either group.

Statement: ticketless travel fell at every one of the 14 stations that received new machines.

Correct rating: insufficient data.

The reasoning runs in two steps. First, locate what the passage asserts: a fall across the 14 stations taken together, which is an aggregate figure. Second, ask what an aggregate permits you to say about its members. A group total can fall while individual members rise, so the passage supports no claim about any particular station. Nothing in the text makes the statement more likely than not, and nothing contradicts it. That is insufficient data.

Each wrong option is built on a specific misapplied rule. True treats a property of the whole as a property of every part, which is the composition error and the most common single mistake on this subtest. Probably true applies a real-world plausibility judgement, reasoning that most stations probably did improve, when the scale asks what the passage makes likely, not what the world makes likely. Probably false imports outside knowledge about uneven rollouts, which is the same error pointing the other way. False requires the passage to contradict the statement, and a silent passage contradicts nothing; candidates pick it when they confuse absence of evidence with evidence of absence.

How to prepare for this type specifically

Drill the scale before you drill the content. Take any short factual paragraph, write four statements about it, and force each into one of the five Inference categories, noting which words license that rating. If you cannot point at the words, the answer is insufficient data.

Build the negation habit for Recognition of Assumptions until it is automatic. Say the proposed assumption aloud in the negative, then reread the statement and ask whether it still makes sense. That converts a vague judgement into a mechanical check that survives time pressure.

Practise accepting absurd premises. On Deduction items, work deliberately with premises you know to be false, because the reflex to correct them produces does not follow on items that plainly do.

For Evaluation of Arguments, separate the two tests explicitly. Ask first whether the argument is true, then whether it answers the question asked. Candidates who merge those into one judgement mark true statements strong and lose a block of marks in one subtest.

Finally, settle the scoring question in advance. Read the instruction screen carefully for any mention of marks deducted for incorrect answers, and decide your behaviour for the closing minute before you need it rather than inventing it under pressure.

Frequently asked questions

TestSolve is independent and not affiliated with any test provider or employer named on this page. All product names and trademarks belong to their respective owners.