CAT Odd Sentence Out: Build the Paragraph, Then See What Is Left
Odd-sentence-out questions look like para jumbles with an extra step, and candidates treat them that way — hunting for the sentence that seems out of place. That approach fails often, because the odd sentence is almost always on the same topic as the rest.
The reliable method is the reverse. Do not look for the intruder. Build the coherent paragraph from the sentences that link to each other, and the one left over is your answer by elimination.
Elimination, not detection
A well-set odd sentence shares the topic, the register and often the vocabulary of the paragraph. What it does not share is the chain of grammatical links — and building that chain is a mechanical task, while "spotting the odd one" is a judgement call under time pressure.
Quick Answer (30-Second Read)
Time per question
100 seconds
Then commit
Marking
No negative
Typed answer
- Build the paragraph first. The leftover sentence is the answer; do not try to identify it directly.
- Same topic is not a defence. The odd sentence usually shares the subject matter deliberately.
- Three tells: no grammatical link, different level of generality, or a shift from describing to evaluating.
- Never leave one blank — these are typed answers and carry no penalty for a wrong choice.
- If two sentences both look loose, the odd one is the one that nothing else refers back to.
Marking scheme per the official CAT bulletin at iimcat.ac.in.
CAT Mock Test — the non-RC questions under real timing
Odd-sentence questions are quick when the method holds and ruinous when it does not. Find out which is true for you.
The method, in four steps
| Step | What you do | Time |
|---|---|---|
| 1 | Read all sentences once, marking pronouns, demonstratives and connectors | ~30 sec |
| 2 | Build the longest chain of linked sentences you can | ~35 sec |
| 3 | Test whether the remaining sentence joins that chain anywhere | ~20 sec |
| 4 | Type the number, move on | ~15 sec |
Step two is identical to the anchor-pair work you would do on a para jumble, which is why the two question types should be practised together. The difference is only in what you do with the result — a jumble asks for the chain, an odd-sentence question asks for what would not fit into it.
Step three is not optional. A sentence that failed to join your chain may simply have been one you did not link correctly, so before committing, check whether it could slot in anywhere. If it can, your chain was wrong and the real intruder is elsewhere.
Three ways the odd sentence gives itself away
| Tell | What it looks like | Reliability |
|---|---|---|
| No grammatical link | Nothing refers back to it, and it refers back to nothing | Strongest — this is the actual definition |
| Wrong level of generality | A broad claim among specifics, or a specific example among general argument | Good, especially when the chain is short |
| Descriptive vs evaluative shift | An opinion in a paragraph that otherwise only reports, or vice versa | Useful as a tie-breaker, not as primary evidence |
The odd sentence is often the most quotable one
Setters frequently build the intruder as a clean, self-contained statement that reads well on its own — which is exactly why it does not belong. Sentences inside a paragraph lean on their neighbours; a sentence that stands entirely upright is suspicious. If one sentence would work as a caption and the others would not, look at it hard.
Two worked examples
Example 1. (1) Urban wetlands absorb runoff during heavy rain. (2) Their capacity depends on how much of the surrounding catchment remains unpaved. (3) Wetland conservation has become a common theme in municipal planning documents. (4) Where paving has advanced, this buffering effect declines sharply.
Build the chain: "Their capacity" in (2) refers to the wetlands in (1), so 1–2 links. "This buffering effect" in (4) refers to the absorption described in (1) and qualified in (2), so 2–4 follows.
Test the leftover: does (3) slot in anywhere? Nothing in the chain refers to planning documents, and (3) refers back to nothing. It is on the topic — wetlands — and shares no link with any sentence.
Answer: 3. Note that it was the sentence most likely to feel relevant, which is precisely the design.
Example 2. (1) Early radio broadcasters assumed listeners would attend to programmes as they had to concert performances. (2) Audience research quickly showed that most people listened while doing something else. (3) Programme formats were redesigned around interruption and repetition. (4) The gramophone had already changed how music was consumed at home.
Build the chain: (2) contradicts the assumption in (1). (3) is the consequence of what (2) found. A clean 1–2–3 sequence.
Test the leftover: (4) is about the gramophone, is on the broader topic of domestic listening, and connects to no sentence in the chain by pronoun, connector or echoed noun.
Answer: 4. It is also the most self-contained sentence of the four — the caption test flags it independently.
Where is VARC time actually going?
Non-RC questions are supposed to be the fast ones. The question-level time view shows whether they are.
- Time per question
- VARC accuracy
- Sectional percentile
The same-topic trap
The instinct that ruins this question type is looking for a sentence about something else. Setters know that instinct exists and design against it. In both examples above, the odd sentence was squarely on topic — wetlands, domestic listening — and a candidate scanning for a subject mismatch would have found nothing and then guessed.
Topic is not the criterion. Connection is. A sentence belongs to a paragraph because something in the paragraph points at it or it points at something in the paragraph, not because it discusses the same thing. Four sentences about wetlands can contain an intruder; four sentences on visibly different subjects rarely appear in this question type at all.
This is also why the method generalises across subject matter. You do not need to know anything about urban hydrology or the history of broadcasting to run it — the evidence is entirely on the surface of the sentences.
When two sentences both look loose
It happens, usually because the chain you built was shorter than it should have been. Three tie-breakers, in order of reliability.
| Tie-breaker | Apply it like this | Why it works |
|---|---|---|
| Direction of reference | Prefer the sentence that nothing else refers back to | A sentence with no backward reference of its own might simply be the opening sentence |
| Level of generality | If the chain is specific, a broad claim is the intruder | These paragraphs are built around one level of abstraction |
| The caption test | Whichever stands entirely on its own, needing nothing before or after | Paragraph sentences lean on their neighbours; captions do not |
And when the tie-breakers do not settle it, type an answer regardless. These questions carry no negative marking, so an uncertain choice is strictly better than an empty box.
Key Takeaways
- Build the paragraph, do not hunt the intruder. The leftover sentence is the answer.
- Same topic is not a defence — a well-set odd sentence deliberately shares the subject matter.
- Connection is the criterion, established by pronouns, demonstratives, connectors and echoed nouns.
- Always test the leftover against the chain before committing — a failed link may be your error, not the setter's design.
- The most self-contained sentence is suspicious. Paragraph sentences lean on their neighbours.
- No negative marking. Never leave one blank, whatever your level of confidence.
People also search for
How do I solve odd sentence out questions in CAT?
By elimination rather than detection. Read all the sentences once, marking pronouns, demonstratives and connectors, then build the longest chain of grammatically linked sentences you can. The sentence that will not join that chain is your answer. Before committing, test the leftover against the chain to check that it genuinely cannot slot in anywhere — if it can, the chain you built was wrong and the real intruder lies elsewhere.
Do odd sentence out questions have negative marking?
No. Like para jumbles, these appear as typed-answer questions rather than multiple choice, so an incorrect selection deducts nothing. That makes leaving one blank strictly worse than guessing under every circumstance. The only reason to move on without answering is that the time is better spent on a reading passage — and even then, enter your best guess before you leave rather than returning to it later.
Why is the odd sentence usually on the same topic?
Because setters design against the instinct candidates bring to the question. If the intruder discussed a visibly different subject, the question would be trivial. Instead it shares the topic, the register and often the vocabulary, and fails only on connection — nothing in the paragraph points at it and it points at nothing in the paragraph. Scanning for a subject mismatch therefore finds nothing and leads to a guess.
How long should an odd sentence out question take?
Around a hundred seconds — roughly thirty to read and mark, thirty-five to build the chain, twenty to test the leftover, and fifteen to enter the answer. Past two minutes the question has stopped being efficient. Because there is no negative marking, the correct action at that point is to type your best candidate and move on, rather than continuing to reread sentences that have already been read twice.
What if two sentences both seem out of place?
Use three tie-breakers in order. First, direction of reference: the sentence that nothing else refers back to is more likely the intruder than one that merely fails to refer backwards itself, since the latter might just be the opening sentence. Second, generality — a broad claim among specifics is usually the intruder. Third, the caption test: whichever stands entirely on its own without needing anything before or after it.
Is the odd sentence out question the same as a para jumble?
The work is nearly identical and the output differs. Both require building a chain of grammatically linked sentences using pronouns, demonstratives, connectors and echoed nouns. A jumble asks you to report the chain; an odd-sentence question asks which sentence would not fit into it. Practising the two together is efficient for exactly this reason — improvement in one transfers directly to the other, which is not true of most VARC question types.
Do I need to understand the paragraph to answer?
No. The evidence sits entirely on the surface of the sentences — a pronoun needing an antecedent, a demonstrative pointing back, a connector requiring something to contrast with. Paragraphs in this question type range across ecology, media history, economics and philosophy, and the method runs identically on all of them. Unfamiliarity with the subject is not a reason to skip, and can help by keeping outside knowledge from clouding the links.
What is the caption test?
A quick check on self-containment. Sentences inside a paragraph lean on their neighbours — they carry pronouns, connectors or definite articles that only make sense in context. A sentence that reads perfectly on its own, the way a caption or a pull quote would, is not doing the work that a paragraph sentence does. Setters frequently build the intruder as exactly this kind of clean, quotable statement, which makes it a useful secondary signal.
How many odd sentence out questions appear in CAT?
They form part of the non-reading-comprehension portion of VARC alongside para jumbles and para summary, and the exact number varies between cycles rather than being fixed. Prepare all three types rather than betting on a particular split. What is stable is the marking distinction: odd-sentence and jumble questions are typically typed answers with no penalty, while summary questions are MCQs that do deduct — which changes how each should be attempted.
How should I practise these questions?
Timed, in mixed sets with para jumbles, since the underlying skill is the same. When reviewing an error, write down the specific link you failed to build rather than noting the correct answer — "missed that 'their capacity' referred back to the wetlands" is usable later, while "chose 3 instead of 4" is not. Most candidates find their errors cluster on one signal type, usually echoed nouns, which makes the fix narrow rather than general.
Like what you read? Share with a friend
Senior Content Writer
Co-Founder at PrepGrind, working on education content, product development, and student-focused learning resources. I also write articles and guides for MBA/MBS entrance exams, covering preparation strategies, exam updates, and useful resources for students.
