📑 On this page
⚡ The short answer
A statistical argument takes a number from a sample and stretches it over a whole group. Weaken it by showing the sample is small, self-selected or untypical, or that the number misleads: percent vs absolute, a missing base rate, a changed measure. Strengthen it with bigger, random, independent evidence that points the same way.
🎬 The 30-second story
The 9-out-of-10 forward
1
“Survey: 9 out of 10 CLAT toppers use RankRocket! Sending to all 14 groups.”
📲
Rohan
Rohan's morning ritual: forward first, read later.
2
“Small print: 'poll of our Instagram followers who cleared CLAT'.”
🔍
Zoya
Zoya opens the link Rohan didn't.
3
“Ten people replied, all already following the app. Of course nine use it.”
🧮
Meher
Meher does the maths.
4
“…I've sent it to the Nani group too.”
🙈
Rohan
Too late to unsend.
Tara's verdict: The number is real; the sample is the problem. Ten self-selected fans of an app describe its fans, not CLAT toppers. Ask who was counted, what was counted and compared with what.
How does CLAT word survey and statistics questions?
| Stem in the paper | What it really asks | First move |
|---|---|---|
| “Which, if true, would most strengthen the main argument?” on a passage built on 'new research' (2023) | Widen the evidence. | Look for more studies, other countries, independent teams. |
| “Which weakens the argument that India has tackled the pandemic better…?” (2021) | Do the numbers count what they claim? | Find the option that says the data under-counts. |
| “Which strengthens the argument that the decline precedes the pandemic?” (2021) | Which statistic has the right date? | Pick the figure from the right period, not the scariest one. |
| “Unemployment rose by almost three times / three per cent” (2021) | Read the units. | Times, per cent and percentage points are different claims. |
→ scroll to see more
What does every statistical argument quietly assume?
Every survey argument has one skeleton: a number from a sample, stretched to a population. The hidden bridge is always the same: the people counted are like the people we're talking about, and they were counted correctly.
🗺️ Argument map
The 9-out-of-10 forward, mapped
The data
An Instagram poll found 9 of 10 toppers who replied use RankRocket
↓
🌉 Hidden assumption (the bridge)
The 10 who replied are typical of all CLAT toppers
↓
🎯 Conclusion
Toppers use RankRocket, so you should too
What each new fact does
- −Only the app's own followers saw the pollWeakens · hits the gap — self-selected sample
- −Ten replies, out of thousands who cleared CLATWeakens · hits the gap — too small to generalise
- ＋An independent survey of 2,000 randomly chosen NLU first-years found the sameStrengthens · hits the gap — large, random, independent
- ∅RankRocket's office is in NoidaNo effect · hits the conclusion — no effect
Every statistics question attacks or repairs that dashed bridge.
📌 Never attack the number itself
The figures are premises: the poll happened and nine said yes. You attack what the number is made to mean: who it represents and what it measured.
How do you check a statistic in four questions?
- 1
Who was counted?Size, selection, typicality. Ten people or ten thousand? Did they choose to answer? Do they look like the group in the conclusion?
- 2
What was counted?Reported or actual? Self-reported habits? A definition that changed midway? A percent hiding a tiny number?
- 3
Compared with what?Last year, the base rate, or a group that didn't get the treatment? A number alone means nothing.
- 4
Point the stemWeaken: the option that breaks one of the three. Strengthen: the option that repairs one.
Which sample and number traps does CLAT test?
🎚️ The scale
How much can this evidence carry?
← AnecdoteStrong evidence →
My cousin
One case
e.g. “My cousin cracked it without coaching”
Opt-in poll
Self-selected
e.g. The app's own followers
One big study
Large but single
e.g. 2,000 students, one university
Independent studies
Many teams, many places
e.g. “Several studies in different countries”
Too small, or self-selected
Ten replies can't speak for thousands, and small samples turn flukes into trends. Worse is a sample that chose itself: people who answer a poll care more than average. A gym's members-only survey will find that most people exercise.
Percent vs absolute numbers
| Coaching centre | Students | NLU selections | Success rate |
|---|---|---|---|
| Centre A, Delhi | 1,200 | 96 | 8% |
| Centre B, Bhopal | 25 | 5 | 20% |
→ scroll to see more
A advertises 'most selections'; B, 'highest success rate'. Both true. But B's rate rests on 25 students, so each student moves it 4 points. 'Better teaching' needs rates on a decent base; 'more seats' needs totals.
⚠️ Percent of what?
'Hostel thefts up 200%' can mean 1 theft became 3. And a pass rate going from 40% to 50% is a 10-point rise but a 25% increase. Read the unit before the size.
Base rates and changed measures
🧭 Decision flow
Does this percentage prove anything?
❓ Does the passage give the base rate, the share in the whole group?
No →
You can't judge the number yet
Yes ↓
❓ Is the share among the cases bigger than the base rate?
No →
No over-representation: the claim collapses
Yes ↓
❓ Was the measure the same throughout (same tests, same definition)?
Yes →
The number supports the claim
No →
The rise may be in counting, not in reality
If 75% of vehicles are two-wheelers, two-wheelers being in 70% of accidents shows they're under-represented. Measures move numbers too: online FIRs raise reported thefts, more tests find more cases.
Can you spot what the number hides? Worked examples
✍️ Worked example
Warm-up
⏱ 30 sec
The 300% dengue headline
A channel flashes: 'Dengue in Sector 12 up 300% this month!' Last month the sector had 2 cases. How many now, and what should you ask first?
- 1Who and what: reported cases in one sector. 'Up 300%' means the rise is 3 × 2 = 6.
- 2New total = 2 + 6 = 8 cases.
- 3Compared with what? A base of 2 makes any change look huge. Ask for the sector's population and nearby sectors.
Answer: 8 cases; ask for the base
⚡ The exam-hall shortcut
Up n% means new = old × (1 + n/100). Up 300% is four times, not three.
✍️ Worked example
CLAT level
⏱ 45 sec
The hatchback speeding report
A report says 65% of cars fined for speeding last year were hatchbacks, so hatchback drivers are the most reckless. Which most weakens it?
(A) Fines doubled last year.
(B) About 70% of cars on the city's roads are hatchbacks.
(C) Hatchbacks are cheaper than sedans.
(D) Some sedan drivers were also fined.
(A) Fines doubled last year.
(B) About 70% of cars on the city's roads are hatchbacks.
(C) Hatchbacks are cheaper than sedans.
(D) Some sedan drivers were also fined.
- 1Who was counted? Fined cars. The claim compares hatchback drivers with others.
- 2Compared with what? 65% means nothing without the hatchbacks' share of traffic.
- 3B is the base rate: 70% of cars but only 65% of fines. Hatchbacks are fined less than their share. A, C and D don't touch the comparison.
Answer: (B): the base rate flips the claim
⚡ The exam-hall shortcut
See 'X% of the cases were Y'? Ask what % of everyone is Y.
✍️ Worked example
CLAT level
⏱ 90 sec
The cooking survey passage
A food-delivery app has published what it calls India's largest survey on eating habits. Over two weeks in May, 50,000 users answered a pop-up shown after they placed an order. 82% said they order food at least four times a week, and 61% said they 'rarely or never' cook dinner. The chief executive says urban Indians have given up on home cooking, and that the company will open 200 cloud kitchens in smaller towns. Fifty thousand responses, she adds, is more than many government surveys collect.
Which, if true, would most weaken the chief executive's conclusion? (Options below.)
Which, if true, would most weaken the chief executive's conclusion? (Options below.)
- 1Map: 50,000 app users → all urban Indians. Bridge: people who just ordered are typical.
- 2Who was counted? Users who had just placed an order: the most order-happy slice of the city.
- 3Compared with what? Nothing on non-users. Size can't fix skew: 50,000 biased replies are still biased.
- 4Point: the weakener shows the sample is untypical.
Answer: (B): only fresh orderers saw the pop-up
⚡ The exam-hall shortcut
Big is not representative. When an author brags about sample size, check who was in it.
🔬 Option autopsy
Option autopsy: the cooking survey
Which, if true, would most weaken the chief executive's conclusion?
A A few respondents may have tapped answers without reading the questions. Tag it
🌗 Half right
A real data worry, but 'a few' careless taps can't overturn 82% and 61%.
B The pop-up appeared only to users who had just ordered, and most urban households don't use a delivery app in a typical week. Tag it
✅ Correct
Breaks the bridge: the sample is the heaviest users, while most of the population isn't on the app at all.
C Nutritionists say frequent ordering is unhealthy. Tag it
🙈 True but irrelevant
About whether the trend is good, not whether it exists.
D Fifty thousand is more than many government surveys collect. Tag it
🔁 Just repeats the passage
The CEO already said it. Size was never the problem.
🌗 Half right✅ Correct🙈 True but irrelevant🔁 Just repeats the passage
Guess the trap behind each option, then tap to check. Wrong options fail in a handful of predictable ways.
How has CLAT asked statistics and survey questions?
📝 Asked in CLAT 2023
· Q128
Widen the evidence
While men and women are both considered to be more capable as they get older, only women bear the brunt of being seen as “less warm” as they age, new research has found. This series of studies is reportedly the first to look at both gender and age to determine how perceptions of women and men differ… In an analysis of professors' evaluations, female professors witnessed a decline as they moved from their 30s to 40s, hitting an all-time low around the age of 47. All this while, the evaluation of male professors remained consistent…
Which of the following, if true, would most strengthen the main argument in the passage?
ASeveral independent studies conducted in different countries have shown that women in the workplace are perceived positively and are favourably treated as they age.
BSeveral independent studies conducted in different countries have shown that women in the workplace are perceived negatively and are unfairly treated as they age.
CThe studies mentioned in the passage have been discredited after they were published, and no reliance should be placed on them.
DThe studies mentioned in the passage were conducted on very small sample sets and cannot be used to make general statements about the difference in perception between men and women.
Reveal answer + reasoning
Answer: B
Main argument: as they age, women, unlike men, are judged more harshly at work. The evidence is one series of studies. To strengthen an argument built on studies, widen it: independent teams, different countries, the same finding. Answer B.
Why the others fail
- A — Same broad evidence, opposite finding. It weakens.
- C — Discrediting the studies removes the evidence. It weakens.
- D — The textbook sample weakener: too small to generalise.
📝 Asked in CLAT 2021
· Q132
Is the data counting reality?
On the day of writing this, India had reported 116 deaths from COVID-19. In contrast, the US, with around one-fourth the population of India, reported 1,897 deaths… On other metrics too-new cases, active cases-the Indian curve has flattened… Instead of learning from India's experience, the first instinct is to doubt Indian data. We aren't counting the cases right, we aren't doing enough tests… In terms of testing, while a case might be made for a lot of Indians not getting tested, it is also true that random testing has increased in the last few months…
Which of the following weakens the author's argument that India has tackled the COVID-19 pandemic better than most advanced nations?
AIndia lacks the co-ordination and effort to execute conspiracies.
BMany Indians are possibly not getting tested.
CEvidence points to the fact India has flattened the corona curve.
DIndia has not imposed draconian lockdowns.
Reveal answer + reasoning
Answer: B
The evidence is reported numbers. The bridge: reported cases and deaths reflect real ones. The author sees the attack coming and tries to wave it off. B revives it: if many aren't tested, low numbers may reflect low testing, not low infection. Answer B.
Why the others fail
- A — The author's own point. It supports her.
- C — Restates her conclusion.
- D — A premise; silent on whether the data is reliable.
Also asked: CLAT 2021 Q129 asked what strengthens the claim that India's slump preceded COVID. Two options were April–June 2020 figures; one vouched for the survey. Only C had the right date: growth fell from 8.3% in 2016-17 to 4% by 2019-20. Match the period to the claim.
Which statistics myths cost you marks?
💥 Myth vs fact
Traps students fall for
🙅 Myth
A bigger sample is always a better sample.
Tap to bust it →
✅ Fact
Only if it's the right people. 50,000 users who just ordered food can't speak for people who never open the app.
🙅 Myth
A higher percentage means more people.
Tap to bust it →
✅ Fact
20% of 25 is 5; 8% of 1,200 is 96. Check the base before you compare.
🙅 Myth
Research by experts can't be weakened.
Tap to bust it →
✅ Fact
Small samples, one country, one method: any study can be. Equally, 'several independent studies in different countries' is CLAT's favourite strengthener.
🎯 Quick quiz
Five numbers to interrogate
0 / 5
A college fest committee wanted to know whether students want a celebrity DJ night. It posted a Google Form on the fest's own Instagram page for two days. Of the 340 students who filled it in, 88% voted yes. The committee told the principal that 'nearly nine in ten students want it' and asked for ₹6 lakh from the activity fund, which all 4,200 students pay into.
Q1. What is the biggest problem with the committee's claim?
340 ÷ 4,200 ≈ 8%, and they selected themselves from the keenest fest-goers. A is fine: 88% is close to nine in ten.Q2. Reported burglaries in a town rose 40% in the year police launched online FIRs. Which explanation needs no extra burglaries?
The measure changed: reporting got easier, so reported crime rose.Q3. A pass rate rises from 40% to 50%. Which is correct?
50 − 40 = 10 points; 10 ÷ 40 = 25%.A study of 3,000 adults in one Bengaluru neighbourhood found that those who cycled to work reported 30% fewer sick days than those who drove. A health blogger wrote that cycling 'cuts sick days across India by a third' and urged every state to build cycle lanes. The study's authors noted that the cyclists were, on average, eleven years younger than the drivers.
Q4. Which, if true, would most strengthen the blogger's claim?
It widens the sample beyond one neighbourhood and holds age constant. A repeats the weakness; D weakens.Q5. 'Four out of five dentists recommend Brand X.' Which question matters most before you believe it?
Size, selection and wording. If dentists could pick several brands, four in five might recommend every brand.🎯 Stats arguments in a nutshell
- Sample → population: the bridge is always 'the sample is typical'.
- Weaken: small, self-selected, untypical, or a changed measure.
- Strengthen: large, random, independent, many places.
- Percent needs a base; a share needs a base rate.
📎 Cheat sheet
- “studies show / new research” → evidence breadth → strengthen with independent studies elsewhere; weaken with small samples.
- “survey / poll / respondents” → who answered → hunt for self-selection.
- “X% of cases were Y” → base rate → what share of everyone is Y?
- “rose by n%” → base size → percent of what? Points or percent?
- “reported / recorded / detected” → the measure → did counting change?
Frequently asked questions
What is sample bias in an argument?
Sample bias means the group that was counted is not like the group the conclusion talks about. A poll of an app's own followers, or diners surveyed inside a restaurant, will skew the result. In CLAT, an option showing the sample is self-selected or untypical is a classic weakener.
Is a bigger sample always more reliable?
No. Size reduces chance flukes, but it cannot fix a skewed sample. Fifty thousand people who just ordered food online still can't tell you how all Indians eat. A smaller random sample of the right population beats a huge biased one every time.
How do you strengthen an argument based on a study?
Widen the evidence. The strongest options say independent teams, in different places or countries, with different methods, found the same thing. CLAT 2023 used exactly this: 'several independent studies conducted in different countries' strengthened a passage built on one series of studies.
What is a base rate and why does it matter?
A base rate is how common something is in the whole group. If 70% of cars are hatchbacks, hatchbacks making up 65% of speeding fines shows they're fined less than their share. Without the base rate, a percentage among the cases tells you very little.
Keep going
← Previous topic
Will the Plan Work? Strengthening and Weakening Proposals
Can it be done, will it work, will it backfire: the 3-test plan audit
Next topic →
Out-of-Scope Options: Why True-but-Irrelevant Never Weakens
True, on-topic and useless: the options to cross out first
📚
Part of the chapter
Strengthen & Weaken Questions for CLAT Logical Reasoning
See how this topic fits the whole of Strengthen / Weaken →
Connected topics in other chapters
Correlation vs Causation
Once the numbers are sound, ask whether they show a cause or just a link.
Hasty Generalisation and Weak Evidence
A tiny sample stretched over a nation, named as a flaw.
Explain the Paradox
Many 'paradoxes' dissolve once you check the base or the measure.
Cause and Effect
Data shows what happened; cause-and-effect stems ask why.
Every topic in Strengthen / Weaken
- 1.Strengthen Questions: How to Support an Argument
- 2.Weaken Questions: How to Attack an Argument
- 3.Correlation vs Causation: Other Causes and Reverse Cause
- 4.Explain the Paradox: Resolve-the-Discrepancy Questions
- 5.Will the Plan Work? Strengthening and Weakening Proposals
- 6.Surveys, Samples and Statistics in Arguments (you're here)
- 7.Out-of-Scope Options: Why True-but-Irrelevant Never Weakens