Start early, pay half.50% off with EARLY-GHOSTY — until 31 August.
GhostPrep TARA
Eerily accurate TARA practice questions

TARA Critical Thinking: the complete guide

What Critical Thinking actually tests, why Oxford and UCL bother testing for it, every one of the 7 question types with real questions from our practice bank, how the module is timed and scored, and where it all comes from.

Start practicing freeJump to the 7 types
By the GhostPrep teamLast updated

Written by the specialists behind our practice bank — over a decade writing, teaching and tracking the UK admissions tests.

What is critical thinking, actually?

Every academic argument is built from three moving parts: reasons, a conclusion, and often an assumption that's never actually written down. A reason is a statement offered as support.

A conclusion is the claim that support is meant to establish. An assumption is the missing link between the two — something that has to be true for the reasons to actually get you to the conclusion, even though nobody says it out loud.

Critical thinking, in the sense TARA tests it, is the skill of seeing that structure clearly — regardless of whether you agree with the conclusion, know anything about the topic, or find the writing persuasive.

You're not being asked whether an argument is right. You're being asked whether it works: do the reasons, if true, actually justify the conclusion, or is there a gap somewhere?

That's a different skill from what most school subjects reward. A history or English essay usually asks you to build an argument.

TARA asks you to take someone else's argument apart and diagnose it — which is closer to what you'll actually be doing at degree level: reading a source, an academic paper, or a policy document, and deciding how much weight its conclusion really deserves.

Why universities test for it

Every applicant to a course like PPE, Human Sciences, or History and Politics arrives with strong grades — that's the baseline, not the differentiator.

What varies far more is how well someone can independently evaluate an argument, spot where the evidence runs out, and avoid being persuaded by writing that sounds confident rather than writing that's actually sound.

That's a much harder thing to see on a UCAS form than predicted grades.

That's the specific gap TARA's Critical Thinking module is built to measure.

It doesn't test subject knowledge, so it can't be crammed the way A-level content can, and it doesn't track neatly with school grades — which is exactly why Oxford and UCL use it alongside the rest of an application, not instead of it.

Which courses require it, and how each department weighs it, differs by university and can change between admissions cycles, so it's worth checking your specific course's page directly rather than assuming.

Why it matters beyond the exam

The reason universities bother testing this at all is that it doesn't stop mattering once you have an offer.

Degree-level study in the humanities and social sciences runs on source evaluation: reading a historian's interpretation, a psychology paper's methodology, or a policy report's headline statistic, and working out how much of it actually holds up.

The exact skills TARA tests — separating a conclusion from its supporting reasons, spotting an unstated assumption, noticing when a correlation is being sold as a cause — are the same ones a first-year seminar will expect you to already have.

They're not exam-specific either. The same structure shows up in a news article's opening claim, a company's marketing statistic, or a colleague's case for a decision at work.

Getting fluent at spotting it now keeps paying off well past results day.

How the module works, and how it's timed

Critical Thinking is 22 multiple-choice questions in 40 minutes, each with five options and one correct answer. There's no penalty for a wrong guess, so every question is worth attempting.

For the purposes of the test, treat every reason you're given as true — your job isn't to fact-check the passage, it's to judge the logic connecting what's stated to what's claimed.

Forty minutes for twenty-two questions works out to just under one minute fifty seconds per question on average — but treating that as a strict per-question clock is a mistake.

A short passage testing Identifying an Assumption might take well under a minute once you've had some practice; a longer Matching Arguments question, where you're abstracting several arguments into logical patterns, can easily take three or four.

Budget flexibly across the module rather than watching the clock on every single question, and with a couple of minutes left and questions still unanswered on screen, guessing beats leaving them blank — there's nothing to lose.

Where it comes from: the TSA connection

TARA's Critical Thinking module didn't appear from nothing.

Oxford and UCL previously used the Thinking Skills Assessment (TSA) for many of the same courses, and TSA's Critical Thinking component used this same reason/conclusion/assumption framework, the same five-option multiple-choice format, and several of the same named question types.

When TSA was discontinued and UAT-UK introduced TARA for 2026/27 entry, the Critical Thinking module carried that structure across largely intact.

Practically, that means if you, or your school, already have TSA Critical Thinking practice materials, they're still genuinely useful preparation — the skill being tested hasn't changed, even though the test's name has.

We've written a full breakdown of exactly what's the same and what's different in our TSA vs TARA guide.

What's a good Critical Thinking score?

TARA scores Critical Thinking on its own scale from 1.0 to 9.0, reported separately from Problem Solving.

UAT-UK has stated that the scale is designed so a typical candidate scores around 4.5, and that roughly the top 10% of candidates score above 7.0 — scores are capped at 1.0 on the low end and 9.0 on the high end.

That means there's no fixed pass mark: a 5.5 puts you comfortably above the typical candidate, while anything above 7.0 puts you in genuinely rare territory.

Because Critical Thinking and Problem Solving are reported separately, it's entirely possible to score strongly on one and only averagely on the other, and universities that use TARA can see both individually.

For the full breakdown of how the October 2025 and January 2026 sittings actually scored, see our TARA scoring guide.

Here's where the key numbers sit:

The TARA score scale
Critical Thinking and Problem Solving are each reported separately on this scale.
1.0
9.0
4.5
Typical candidate
7.0
Top 10% score above this

The 7 types, with real practice questions

Each one below is a genuine question from the GhostPrep TARA practice bank — not a simplified version. Read the passage, have a go, then reveal the answer to see the reasoning laid out.

Which question type am I looking at?

A quick way to place any Critical Thinking question before you start reasoning through it.

Question type 1 of 7Practice bank · ID 164

Identifying the Main Conclusion

What it tests: Pick out the one statement every other sentence in the passage is working to support.

Approach: Read the whole passage before you look at the options — the conclusion is a judgement or recommendation, not a plain fact, and it can sit anywhere in the passage, not just the last line.

Step by step
  1. Read the whole passage once before looking at the question.
  2. Ask yourself: what is this passage ultimately trying to convince me of?
  3. Test each candidate answer — does it get supported by the rest of the passage, or does it do the supporting?
  4. Pick the statement sitting at the top of that support structure — the one nothing else outranks.

Common trap: The most “final-sounding” sentence isn't always the conclusion. Passages often close with one more piece of supporting evidence, not the point itself — treat the last sentence with the same suspicion as any other.

Passage

Ministers have proposed requiring encrypted messaging services to build in access points so that law enforcement can read suspects' messages. This plan should not go ahead. Any deliberate weakness built into an encryption system can, in principle, be found and exploited by hackers as well as by police, since code cannot distinguish a warrant from a break-in. Criminals aware of the backdoor would simply migrate to unregulated foreign apps beyond UK jurisdiction, so the measure would do little to hinder serious offenders while leaving ordinary users more exposed. Several major technology firms have already warned they would withdraw services from the UK market rather than comply, risking thousands of jobs.

Which is the best expression of the main conclusion of the argument?

AAny weakness deliberately built into an encryption system can be exploited by hackers as well as by police.
BBecause criminals would move to unregulated foreign apps, the backdoor requirement would fail to hinder serious offenders.
CThe government should abandon its plan to require encrypted messaging apps to include access points for law enforcement.
DAll government regulation of encryption technology is harmful and should be opposed.
EMajor technology firms would withdraw their services from the UK if forced to build in the proposed access points.
Reveal the correct answer & explanation

Correct answer: C

The passage opens by describing the proposal, then states straight away that it “should not go ahead” — everything that follows exists to defend that one position. The exploitable-backdoor point, the criminal-migration point, and the firms' withdrawal point are all reasons offered in support of C, not separate conclusions in their own right. E is the classic trap here: it's the final sentence, so it feels conclusive, but it's actually just one more piece of supporting evidence. D overreaches — the passage objects to this specific backdoor plan, not to encryption regulation in general.

Question type 2 of 7Practice bank · ID 86

Drawing a Conclusion

What it tests: Work out what must follow from the passage, even though that exact sentence never appears in it.

Approach: Treat every sentence in the passage as true, then ask what necessarily follows from combining them.

Step by step
  1. List out each stated fact or reason as if it were true.
  2. Combine only what's explicitly given — resist pulling in outside knowledge, even if it feels obviously true.
  3. Favour the most modest inference that's still fully backed by the passage over a bolder-sounding one.
  4. Reject any option that needs one more assumption to work — that's a job for a different question type.

Common trap: Options that generalise beyond the passage's stated scope — a different population, a different category, a stronger absolute claim — are the most common wrong answers here.

Passage

A 2023 government review found that most major messaging apps available in the UK now provide end-to-end encryption by default, compared with under half five years earlier. Officers surveyed by the review reported that, in cases involving encrypted messages, requests to providers for the content of those messages were refused in nearly all instances. However, requests for associated metadata — such as who contacted whom, and when — were granted in the majority of cases. The review did not examine encrypted voice calls or cloud storage services, and it made no attempt to measure whether the growth in encryption adoption has affected conviction rates for any type of offence.

Which of the following can be drawn as a conclusion from the passage above?

AThe rise in encryption adoption is the main cause of any recent change in conviction rates for the offences officers investigated.
BRequests for content from providers of encrypted voice-call services are refused in nearly all cases, just as with messaging apps.
CIn some investigations involving encrypted messages, officers obtained metadata about the communication even though the message content itself was withheld.
DProviders can never be compelled to hand over the content of encrypted messages under any circumstances.
EBecause content requests were mostly refused while metadata requests were mostly granted, providers were inconsistent in how they responded to law-enforcement requests.
Reveal the correct answer & explanation

Correct answer: C

Two statistics are given about the same set of cases: content requests were refused “nearly all” the time, while metadata requests were granted in “the majority” of cases. For both to be true at once, some cases must exist where officers got the metadata without getting the content — a small, careful inference straight from the numbers. D overclaims by turning “nearly all” into “never, under any circumstances.” B reaches into voice calls, a category the review explicitly didn't cover. A introduces a causal claim about conviction rates the passage says was never measured.

Question type 3 of 7Practice bank · ID 1245

Identifying an Assumption

What it tests: Spot the unstated step the argument needs to be true in order for its conclusion to follow.

Approach: Use the negation test: flip an option to its opposite and see whether the argument collapses.

Step by step
  1. Identify the conclusion, then the stated reasons supporting it.
  2. Ask what unstated link has to be true for those reasons to actually reach that conclusion.
  3. Apply the negation test to your best candidate: flip it to false and see if the argument falls apart.
  4. Discard options that are true-sounding background detail but don't actually hold the argument up.

Common trap: Plausible general knowledge about the topic is the classic distractor — it may well be true, but if the conclusion still stands without it, it isn't the assumption.

Passage

A study followed two groups: adults who read literary fiction for at least an hour daily, and adults who read no fiction at all. The fiction readers scored notably higher on a standard test of empathy, averaging fifteen per cent above the non-reading group, a gap researchers called substantial. Science journalists argued that governments wanting to raise national empathy levels should redirect funding from television drama subsidies to public library expansion, since the study showed that immersion in fictional narrative raises empathy, and library membership, unlike television viewing, can be tracked and incentivised through local schemes, making libraries the more practical policy lever.

Which is an underlying assumption of the argument?

AEngaging with fictional narrative through television does not raise empathy to the same extent as reading fiction does.
BExpanding public libraries will succeed in increasing the amount of fiction adults actually read.
CAdults who choose to read fiction do not already have higher empathy levels before they begin reading.
DEvery adult who reads fiction for at least an hour a day will show a rise in empathy test scores.
EThe empathy test used in the study accurately captures genuine empathetic behaviour rather than test-taking skill.
Reveal the correct answer & explanation

Correct answer: A

The conclusion shifts funding away from television drama specifically, on the grounds that “immersion in fictional narrative raises empathy.” But TV drama is fictional narrative too. Negate A — suppose TV narrative worked just as well — and there's suddenly no reason to single out television for the cut; the argument collapses. D fails the same test: negating it (not literally every reader improves) doesn't damage the case at all, since the argument only needs the average finding to hold.

Question type 4 of 7Practice bank · ID 1101

Assessing the Impact of Additional Evidence

What it tests: Decide whether a new piece of information, if true, would strengthen or weaken the argument.

Approach: Name the actual gap in the reasoning before you look at the options.

Step by step
  1. Name the actual gap in the reasoning before reading the options — a missing control group, an unrepresentative sample, a leap from correlation to causation.
  2. Re-check whether the question asks you to strengthen or weaken; it's easy to solve the opposite question under time pressure.
  3. Test each option against that specific gap, not just against the general topic.
  4. Rule out anything that's true but irrelevant to the gap you identified, or that applies to the wrong group entirely.

Common trap: Answers that are simply on-topic and sound positive (or negative) are the usual trap — they need to close the actual logical gap, not just relate to the subject.

Passage

A popular-science magazine reports on a twelve-week trial of a daily omega-3 supplement given to adults aged 60–75 who had mild memory complaints. Among those who completed the full trial and took the supplement every day, memory-test scores at the end were higher than scores at the start. There was no group in the trial that went without the supplement for comparison. On this basis, the article's researchers conclude that the supplement improves memory in older adults with mild memory complaints.

Which of the following, if true, most strengthens the argument above?

AThe supplement is derived from fish oil, which other research has linked to improved heart health in adults of all ages.
BParticipants who completed the trial were on average two years older than those who dropped out before it finished.
CAdults aged 40–59 who took the same supplement for twelve weeks reported feeling more able to concentrate at work.
DA separate group of similarly aged adults with mild memory complaints, given an inactive placebo instead of the supplement, showed no significant change in memory scores over the same twelve weeks.
EDuring the trial, participants were also encouraged to sleep more and reduce alcohol intake alongside taking the supplement.
Reveal the correct answer & explanation

Correct answer: D

The trial has no untreated comparison group, so the rise in scores could just be a practice effect or natural fluctuation rather than the supplement working. D supplies exactly that missing comparison: a matched group given a placebo showed no change, which rules out the alternative explanations and isolates the supplement as the likely cause. E looks like a positive trial detail but is actually a confound — it gives the memory rise another possible cause, which weakens rather than strengthens the claim.

Question type 5 of 7Practice bank · ID 314

Detecting Reasoning Errors

What it tests: Explain why the conclusion doesn't actually follow from the reasons given.

Approach: Name the specific kind of gap between the reasons and the conclusion — often correlation treated as proof of causation, as it is here.

Step by step
  1. State the conclusion and the reasons supporting it in your own words.
  2. Name the specific logical move being made — correlation treated as causation, a necessary condition treated as sufficient, a small sample generalised too far.
  3. Match that exact move to an option, rather than any option that criticises the passage validly but differently.
  4. Discard options that invent a claim the passage never actually made.

Common trap: More than one option often describes a real weakness somewhere in the passage — the correct one is whichever weakness the conclusion actually depends on.

Passage

A survey of English towns found that those with more charity shops per head on their high street reported higher average resident wellbeing scores than towns with fewer such shops. Council researchers concluded that opening more charity shops causes wellbeing to rise, and therefore recommended that town centres with empty units convert them into charity shops specifically to boost local wellbeing. The researchers noted that the surveyed towns varied widely in size, population and existing high-street composition, but argued this did not weaken their conclusion, since the pattern of more charity shops and higher wellbeing held across every town in the sample.

Which of the following best identifies the flaw in the argument above?

AIt overlooks the fact that charity shops depend on donated stock, a limitation the argument already implicitly accepts by discussing shop numbers rather than turnover.
BIt infers that because opening more charity shops raised wellbeing in the sample, councils are therefore morally obliged to prioritise charity shops over other regeneration schemes.
CIt treats the correlation between charity-shop numbers and wellbeing scores as proof that charity shops cause higher wellbeing, without considering that some other factor, such as a town's underlying prosperity or community engagement, might produce both.
DIt assumes without justification that every town surveyed had exactly the same number of charity shops, making the comparison invalid.
EIt relies on the emotional appeal that charity shops support good causes, rather than on the wellbeing data actually collected.
Reveal the correct answer & explanation

Correct answer: C

A correlation is observed across a varied sample and turned directly into a causal claim, with no consideration of a third factor — a town's underlying prosperity or community spirit, say — that could independently produce both more charity shops and higher wellbeing. Noting that the pattern held despite the towns' variation doesn't rule this out; it just means the confound would need to apply broadly too. D invents a claim the passage never makes — it explicitly compares towns with differing shop numbers.

Question type 6 of 7Practice bank · ID 147

Matching Arguments

What it tests: Find the option that shares the passage's underlying logical structure, regardless of topic.

Approach: Strip the argument down to an X/Y pattern before you touch the options.

Step by step
  1. Strip the passage down to an abstract X/Y pattern, ignoring the topic entirely.
  2. Write out the exact logical connective — “if X then Y, X is true, therefore Y” is a different pattern from “if X then Y, Y is true, therefore X.”
  3. Scan the options for the same shape, not the same subject matter.
  4. Remember the original argument may itself be invalid — that's fine, your job is to match structure, not judge soundness.

Common trap: The option that feels most topically similar to the passage is rarely the right one — TARA deliberately pairs a matching structure with an unrelated topic, and a mismatched structure with a similar-feeling one.

Passage

A wind farm can only receive the government's green subsidy if its turbines' capacity factor exceeds 40%. This turbine array's capacity factor exceeds 40%, so it must have received the green subsidy.

Which of the following most closely parallels the reasoning used in the argument above?

AA film can only win the top festival prize if it was shot on 35mm film. This film did not win the top festival prize, so it must not have been shot on 35mm film.
BA patient can only receive the experimental drug if they pass all three screening tests. This patient passed only two of the three screening tests, so she did not receive the drug.
CA restaurant can only earn a Michelin star if its head chef trained in France. This restaurant's head chef trained in France, so it must have earned a Michelin star.
DIf a beehive has a healthy queen, honey production increases. This beehive has a healthy queen, so honey production has increased.
EA cyclist can only make the national squad if they finish the time trial in under an hour. Tom did not finish the time trial in under an hour, so he did not make the national squad.
Reveal the correct answer & explanation

Correct answer: C

The passage's logic is: the subsidy requires a capacity factor above 40% (a necessary condition); that condition is met; therefore the subsidy was received. That's invalid — meeting one requirement doesn't guarantee every other requirement was also met. C makes the identical move: a Michelin star requires a French-trained chef; the chef is French-trained; therefore the star was won. D is a trap for skim-readers — there the stated condition genuinely is sufficient, so affirming it is valid reasoning, unlike the passage.

Question type 7 of 7Practice bank · ID 131

Applying Principles

What it tests: Identify the general rule an argument relies on, then find the option that follows that same rule in a different situation.

Approach: State the underlying principle in one plain sentence, stripped of the specific topic.

Step by step
  1. Identify the conclusion and the reasoning behind it, exactly as with any other type.
  2. Generalise that reasoning into one sentence stripped of the specific topic — a rule that could apply anywhere.
  3. Test every option against that general sentence, not against the original topic.
  4. Watch specifically for an option that shares the topic area but reaches the opposite conclusion — the standard trap.

Common trap: An option can sit in the same real-world domain as the passage — same topic, same tone — and still be wrong because it argues the opposite way. Match the rule, not the subject.

You'll notice this passage returns to the same encryption-backdoor debate as question 1. That's actually useful here: it shows that TARA doesn't care what a passage is about — it cares whether you can see how the argument is put together. Two questions can share a topic completely and still test entirely different skills.

Passage

Some politicians argue that companies should be forced to build ‘backdoors’ into encrypted messaging apps, so police can read the messages of suspected criminals. But criminals who want secrecy will simply switch to other encrypted tools, foreign apps, or coded language beyond any backdoor's reach. Meanwhile, ordinary users — including journalists, abuse victims, and small businesses — would lose the protection encryption provides, since any backdoor built for police could eventually be exploited by hackers too. A policy that fails to stop the people it targets while exposing millions of law-abiding users to new risks should not be introduced.

Which of the following best illustrates the principle underlying the argument above?

AAirport security scanners inconvenience every passenger to catch a handful of smugglers, and smugglers could find alternate routes anyway, yet the inconvenience is minor, so scanners should continue.
BSince it is not illegal for a customer to loudly criticise a shop's products while browsing, staff should be free to let them continue without stepping in.
CA school proposes banning all students from bringing mobile phones onto site to stop the few who cheat by texting answers during exams. But those students could simply hide notes or memorise answers instead, while removing phones from everyone would stop students contacting parents in emergencies or using revision apps. Since the ban would not stop the cheats but would deprive every other student of a useful tool, it should not be adopted.
DBecause most drivers already exceed the speed limit on motorways without causing accidents, the speed limit should be raised to match how people actually drive.
ESince determined drug users can always find a dealer regardless of enforcement, drug possession should be decriminalised so police resources can focus on treatment instead.
Reveal the correct answer & explanation

Correct answer: C

The principle: a measure that restricts everyone to stop a small group, that the targeted group can easily evade, and that strips a genuine benefit from the law-abiding majority, should not be adopted. C matches every part of that in a school setting. A is the classic trap — same topic area (security, futility, minor harm) but the opposite conclusion, since it recommends keeping the measure. E captures only the futility half and drops the harmed-majority element entirely.

22
Questions
40
Minutes
~1m50s
Average per question
0
Marks lost for guessing

Seven types down. Hundreds of questions to go.

You've now seen one example of each Critical Thinking question type. Register free to work through the rest of the bank at your own pace, with the reasoning explained the same way every time.

Start practicing free

Free to start · 10 practice questions a day · no card required

Next up
TARA Problem Solving: the 3 question types & the maths you actually need
Start here
TARA Test 2026/27: the complete guide for Oxford & UCL applicants