What Classification Tests
Classification โ popularly called "odd one out" โ is one of the highest-frequency topics in RPF Sub-Inspector reasoning. Out of roughly 35 reasoning questions, you can expect 3 to 5 from this single chapter. The questions look like a quick visual check, but they reward speed only if you know exactly which rule the setter is applying. The candidate who scans for one specific pattern at a time, instead of looking at the whole set vaguely, finishes the question in 10โ15 seconds.
Definition: A classification question presents four or five items and asks you to identify the one that does not share the property common to the others. The "odd" item is the answer; the others are linked by a rule you must first identify.
Why this topic is in the syllabus
Classification tests two narrow but important reasoning skills: pattern recognition under time pressure and the willingness to discard a partial match. In railway operations โ RPF SI's professional context โ staff must repeatedly identify the exception in a list (the suspicious parcel, the misplaced rake, the unlogged passenger). The exam mirrors that skill on paper.
Why it matters: A 10-second classification answer banks time you will need on the harder syllogisms and seating arrangements later in the paper. Three confident classification answers in 30 seconds is a candidate's quickest 3-mark cushion.
The four families RPF SI loves
Definition: A category family is the underlying type of relationship the setter has built into the four "matching" items. RPF SI questions almost always come from one of four families.
Family 1: Living and non-living groupings
Items that belong to broad biological or material categories: animals, birds, insects, fishes, vegetables, fruits, flowers, metals, gases, instruments, professions, vehicles. The odd one is usually a closely related impostor: in {Tiger, Lion, Leopard, Wolf}, all are big cats except Wolf (a canid). In {Lily, Rose, Marigold, Bamboo}, all are flowers except Bamboo (a grass).
The trap: an impostor that looks like a member of the group. Sparrow, parrot, ostrich, bat โ three are birds, but the bat is a mammal. The setter relies on the candidate's casual classification.
Family 2: Mathematical properties
Numbers form most of the high-scoring classification items. Common rules:
- Primes vs composites: {7, 11, 13, 15} โ fifteen is composite.
- Perfect squares: {16, 25, 49, 50} โ fifty is not.
- Perfect cubes: {8, 27, 64, 100} โ hundred is not.
- Multiples: {12, 24, 36, 50} โ fifty is not a multiple of 12.
- Even/odd: {3, 5, 7, 8} โ eight is even, the rest odd.
- Sum or product of digits: {12, 21, 30, 31} โ three of them have digit-sum 3.
- Two-digit reversals: {12, 21, 34, 43} โ usually paired.
Quick scan order: check digit sum, then primeness, then squareness, then divisibility by 2/3/5. Whichever check splits the set 3-vs-1 is your rule.
Family 3: Letter and alphabet patterns
Letters use rules that the alphabet itself enforces:
- Vowel position: {AEI, OUE, KMP, IOU} โ three have only vowels, KMP has only consonants.
- Equal gap: in {ACE, GIK, MOQ, SUV}, three follow the rule "skip one letter each time" (gap of 2) but SUV has a 2-then-1 gap.
- Consonant clusters: similar to the gap rule but with consonants only.
- Reverse alphabet pairs: items where letters are positioned symmetrically about the centre (A-Z, B-Y, C-X, etc.).
- Mirror in keyboard or alphabet half: occasional, but appears in advanced sets.
Speed habit: write down each item's letter positions (A=1, B=2 ... Z=26) and compute the differences. Three items with the same difference pattern give the rule; the fourth breaks it.
Family 4: Word meaning groups
Words tied by meaning, not letters. Subcategories:
- Metals: {Iron, Copper, Silver, Marble} โ marble is a rock.
- Instruments: {Sitar, Violin, Tabla, Flute} โ three are strings/wind, Tabla is percussion. (Or sometimes three are Indian, one is foreign.)
- Professions: {Doctor, Lawyer, Architect, Patient} โ patient is a recipient, not a profession.
- States/UTs of India: {Goa, Sikkim, Delhi, Mizoram} โ Delhi is a Union Territory; rest are states.
- Continents, oceans, languages: similar pattern questions.
The PAGE memory aid
A reliable scan order for any classification item:
- P โ Primes / Perfect squares. First number-only check.
- A โ Alphabet gap. For letter sets, compute differences.
- G โ Group / Category. For words, recall the broad category.
- E โ Even/Odd or End-letter. Last quick split for both numbers and words.
Run these checks in order. The moment a check produces a 3-vs-1 split, you have the rule. If none works, look for sum-of-digits or reverse-alphabet positions before guessing.
Why it matters: A defined scan order prevents the worst classification habit โ staring at the four items and "feeling" the odd one. Feeling fails on lookalike sets; systematic checking does not.
The double-trap problem
Definition: A double trap is a classification set where two of the items seem to break a rule, forcing you to pick the more fundamental break.
Example: {12, 16, 25, 36}. At first glance, 25 looks odd because it is the only odd number. But 12 is the only one that is not a perfect square. The setter expects the candidate to pick 25; the correct answer is 12, because being a perfect square is the deeper, defining rule shared by 16, 25 and 36.
How to handle double traps: rank the rule strengths. Mathematical category > digit-sum coincidence > even/odd > position curiosities. Pick the option that breaks the most defining rule.
Common misconception: That the first odd-looking option is the answer. RPF SI specifically sets these traps. Read the option matrix calmly and check both possible rules before locking in.
Real-world example: When a railway control room scans a list of cargo entries, several may look suspicious โ undocumented weight, mismatched destination, unusual labelling. The senior controller is trained to focus on the most fundamental anomaly. Classification is the paper version of that habit.
Worked example:
Question: Find the odd one out: 144, 169, 196, 200.
Solution:
Step 1: Check even/odd โ all are even, so this rule does not split.
Step 2: Check perfect squares โ 144 = 12ยฒ, 169 = 13ยฒ, 196 = 14ยฒ. 200 is not a perfect square.
Step 3: Confirm there is no stronger rule the squares break.
Conclusion: The odd one is 200.
Worked example (letters):
Question: Find the odd one out: BD, FH, JL, NQ.
Solution:
Step 1: Compute alphabet positions. B=2, D=4 โ gap 2; F=6, H=8 โ gap 2; J=10, L=12 โ gap 2; N=14, Q=17 โ gap 3.
Step 2: The first three pairs have a gap of 2; NQ has a gap of 3.
Conclusion: NQ is the odd one out.
Worked example (words):
Question: Find the odd one out: Mumbai, Chennai, Kolkata, Bengaluru.
Solution:
Step 1: All four are major Indian metros.
Step 2: Mumbai, Chennai and Kolkata are coastal port cities; Bengaluru is inland.
Conclusion: Bengaluru is the odd one out.
Final habits that save marks
- Never spend more than 20 seconds on a classification item. If the rule does not emerge, mark and move.
- Verify your candidate by reading the rule the other three share, not by why your pick is odd. If you can express the shared rule cleanly, you are right.
- Beware brand-new option formats. Recent RPF SI papers have inserted Hindi alphabet sets and Roman numeral sets. The rules above still apply; only the surface changes.
| Family | Typical setter rule | Quick check |
|---|---|---|
| Living / non-living | Same kingdom or class | Look for the impostor (bat in birds) |
| Mathematical | Primes, squares, cubes, multiples | Run PAGE on digits and divisibility |
| Letters | Alphabet gap, vowel-consonant | Compute A=1, B=2 differences |
| Word meaning | Metals, instruments, professions | Name the shared category in one word |
- โ- Classification rewards quick rule identification, not deep analysis.
- โ- Use the PAGE order: Primes/squares โ Alphabet gap โ Group โ Even/odd.
- โ- In double traps, the deeper structural rule wins.
- โ- Verify by stating the rule the other three share.
- โ- 10โ15 seconds is the target time per classification question.
- โ- Letter questions are solved by mapping letters to numbers and checking gaps.
- โ- Word questions often hide impostors โ bat among birds, Delhi among states.
- โ- Never lock in a guess unless one rule cleanly splits the set 3-vs-1.
"PAGE" โ Primes/squares, Alphabet gap, Group/category, Even-odd. Run the scan in this order; the first check that produces a 3-vs-1 split is your rule. For double traps, remember "DEEPER WINS" โ pick the option breaking the more fundamental rule.
- โ- Classification asks for the one item that does not share the common rule.
- โ- Four families: living/non-living, mathematical, letter-pattern, word-meaning.
- โ- Scan with the PAGE order; the first split rule is usually correct.
- โ- Double traps pick the deeper rule, not the first odd-looking option.
Number Classification Checks
Number classification questions in RPF Sub-Inspector look harmless โ four numbers, pick the odd one out โ yet they trap thousands of aspirants every shift because the rule that seems obvious is often the trap. A systematic checklist beats inspiration, every time.
Definition: A number classification (or "odd one out") problem gives four or five numbers and asks which one does not share a hidden property with the rest. The property may be parity, divisibility, primeness, being a perfect power, or a digit-based rule.
The Fixed Six-Step Checklist
When you see four numbers in an RPF SI question, do not stare and "feel" for the answer. Run the same six tests in the same order, every single time. This is muscle memory, not creativity.
Test 1 โ Primes vs composites. Are most of them prime (2, 3, 5, 7, 11, 13, 17, 19, 23, 29 โฆ)? If three are prime, the composite is your odd one out. If three are composite, the prime is odd. Memorise primes up to 50 cold โ RPF SI almost always picks from that range.
Test 2 โ Perfect squares. Are most of them perfect squares (1, 4, 9, 16, 25, 36, 49, 64, 81, 100, 121, 144, 169, 196, 225)? If yes, the non-square is your answer.
Test 3 โ Perfect cubes. Are most of them perfect cubes (1, 8, 27, 64, 125, 216, 343)? If yes, the non-cube is the answer. Watch out โ 64 is both a square (8ยฒ) and a cube (4ยณ), so 64 can sit in either group depending on what the other numbers are.
Test 4 โ Parity (even vs odd). Are all of them even except one, or all odd except one? This is the simplest test, so it should actually be your first sanity check. If three are even and one is odd, you almost always have your answer.
Test 5 โ Multiples of a fixed number. Are most of them multiples of 5, 7, 11, or 13? Try a small set of common factors (3, 5, 7, 11). Three numbers from a multiplication table with one outsider is a very common pattern.
Test 6 โ Digit-based rules. Sum of digits, product of digits, digit-reversal property (palindromes), or alternating digits. Try this last, only if Tests 1โ5 fail.
Definition: A prime number is a natural number greater than 1 that has exactly two distinct positive divisors โ 1 and itself. A composite number has more than two divisors. Note carefully: 1 is neither prime nor composite โ it is the unit. This is a textbook trap that examiners exploit relentlessly.
Trap Rules You Must Know Cold
- 1 is neither prime nor composite. A set like {1, 2, 3, 5} looks like "four small primes" but 1 is the imposter.
- 2 is the only even prime. A set like {2, 4, 6, 8} has 2 as the odd one not because of size but because three of the four are composite while 2 is prime; or in a set like {2, 3, 5, 7}, 2 is the only even โ both interpretations are valid, so a careful examiner will design only one to fit perfectly.
- 64 is the only number โค 100 that is both a perfect square and a perfect cube (since 64 = 8ยฒ = 4ยณ = 2โถ). Other dual-power numbers (1, 729 = 27ยฒ = 9ยณ) lie outside common RPF ranges.
- Confirm by counting. A valid odd-one-out question must have exactly three numbers sharing a property and exactly one breaking it. If two-and-two share a property, you have misidentified the rule โ try the next test.
Why It Matters
In RPF SI Phase I, three to four marks of Reasoning come from classification โ pure speed marks. A trained candidate solves each in 15 seconds; an untrained one wastes a minute and still picks wrong. Over the full paper, that is two extra minutes for arithmetic or general awareness, and arithmetic at the cutoff is what decides selection.
Real-world example: When the railway recruitment cell tabulates candidate roll numbers and flags duplicates, the underlying logic is the same โ find the entry that breaks the pattern shared by the rest. The reasoning syllabus is training a skill the service actually uses on the job: spotting the outlier in a clean dataset.
Common misconception: Many students believe the "harder" the rule, the more likely it is the answer โ so they jump to digit-sum or some clever cubic relation. In reality, the simplest rule that fits three numbers is almost always the intended one. If even/odd already explains the split, do not look for cubes. Occam's razor wins in RPF reasoning.
Question: Find the odd one out โ (8, 27, 64, 100).
Solution:
Step 1: Apply Test 4 (parity) โ 8, 64, 100 are even; 27 is odd. Three even, one odd. So 27 could be the answer.
Step 2: Apply Test 3 (cubes) โ 8 = 2ยณ, 27 = 3ยณ, 64 = 4ยณ are perfect cubes; 100 = 10ยฒ is not a cube. Three cubes, one non-cube. So 100 could also be the answer.
Step 3: When two tests give two different answers, pick the rule that uses a stronger, more specific property. Cubes are a tighter property than parity (parity is true for half of all numbers; cubes are rare). So the intended classifier is cubes.
Conclusion: 100 is the odd one out โ it is a perfect square but not a perfect cube, while the other three are perfect cubes.
Question: Find the odd one out โ (15, 25, 35, 45).
Solution:
Step 1: Test 4 (parity) โ all four are odd. No split. Move on.
Step 2: Test 5 (multiples) โ all four are multiples of 5. No split. Move on.
Step 3: Test 1 (composite check) โ 15 = 3ร5, 35 = 5ร7, 45 = 9ร5 each have two distinct prime factors; 25 = 5ร5 has only one distinct prime factor.
Conclusion: 25 is the odd one โ it is a perfect square (5ยฒ) and has only one distinct prime factor, unlike the other three.
How to Drill This
In your last 30 days before the exam, do two minutes of classification a day โ twenty questions, timed. Track the test number that solved each one. Within two weeks you will discover that Tests 1 and 4 alone solve 70 percent of all RPF SI classification questions. That data tells you where to look first when the clock is running.
| Test | Property checked | Frequency in RPF SI | Time to run |
|---|---|---|---|
| Test 1 | Prime vs composite | High | 10 sec |
| Test 2 | Perfect squares | High | 5 sec |
| Test 3 | Perfect cubes | Medium | 5 sec |
| Test 4 | Even vs odd | Very high | 2 sec |
| Test 5 | Multiples of a fixed number | Medium | 10 sec |
| Test 6 | Digit-sum / reversal | Low | 15 sec |
- โ- Always run the six-test checklist in order โ parity first, digit-rules last.
- โ- 1 is neither prime nor composite.
- โ- 2 is the only even prime.
- โ- 64 is both a perfect square and a perfect cube โ handle with care.
- โ- A valid question must have exactly three numbers sharing the rule.
- โ- If two rules both seem to fit, pick the rarer (more specific) one.
- โ- Memorise primes up to 50 and squares/cubes up to 225.
PSEC-MD: Primes, Squares, cEubes (yes โ pronounce it "EE-cubes" to keep order), pariTy, Multiples, Digits. Or use the simpler English chant: "Prime, Square, Cube, Even, Multiple, Digit". The first letters spell PSCEMD โ say it before each set and you will never skip a test.
- โ- Classification is a checklist exercise, not a flash of insight.
- โ- The simplest rule that fits three of four numbers is the right one.
- โ- Trap numbers โ 1, 2, 64 โ appear in roughly one in five RPF SI questions.
- โ- Two minutes of daily timed practice will lock the method in for exam day.
Worked Example: Spotting the Odd Item
Q: (a) Rose (b) Lotus (c) Marigold (d) Mango. Three are flowers; Mango is a fruit -> answer (d). Q: (a) 8 (b) 27 (c) 64 (d) 81. 8=2^3, 27=3^3, 64=4^3 are cubes; 81=3^4 (or 9^2) is not a cube -> answer (d). Q: (a) BD (b) FH (c) JM (d) PR. Gaps: B-D=1 skip, F-H=1 skip, P-R=1 skip, but J-M has 2 letters skipped (J,K,L,M) -> JM is odd, answer (c). TIP: when 3 options have an equal alphabet gap, the one with a different gap is the misfit. Always count gaps using position numbers to avoid careless errors.
Classification (Odd One Out) โ Flashcards
Find the shared property, then the misfit. 12 cards on odd-one-out for RPF SI.
Q1. Core method for classification questions?
A1. Find the property common to most options; the one lacking it is the answer. There is always exactly one intended common thread.
Q2. 3, 5, 7, 9, 11 โ odd one out?
A2. 9 โ all others are prime; 9 = 3ร3 is composite.
Q3. When four options seem similar, what secondary property do RPF SI setters use?
A3. A hidden second layer: digit-sum, perfect square/cube, prime, even/odd, or position in alphabet. Test these when the surface category matches all.
Q4. How to attack "which number is different" when all share a surface trait?
A4. Check exactly one property holds for three of four: prime, perfect square, perfect cube, or a fixed nยฒยฑ1 / nยณยฑ1 pattern. The one breaking it is the answer.
Q5. Letter classification BD, FH, JL, MN โ odd one?
A5. MN โ others have a one-letter gap (skip one); M and N are consecutive.
Q6. Trap when the category is "animals"?
A6. Sub-classify: mammal vs bird vs reptile, or domestic vs wild, or number of legs. E.g., Whale among fish (it's a mammal), or Ostrich among flying birds.
Q7. 64, 216, 512, 729, 1000 โ odd one out?
A7. These are 4ยณ,6ยณ,8ยณ,9ยณ,10ยณ. 729 (9ยณ, odd base) breaks the even-base run.
Q8. How to handle word classification like Copper, Iron, Silver, Bronze?
A8. Bronze โ the others are pure metals/elements; bronze is an alloy.
Q9. Two options look equally "odd" โ how to decide?
A9. Prefer the property that leaves a cleaner group of three sharing an exact trait. The setter always has one tidy majority class.
Q10. 15, 25, 35, 45, 55 โ is there an odd one?
A10. All are multiples of 5 ending in 5. Look deeper: 25 is a perfect square; others aren't. So 25.
Q11. Direction of eliminating in "choose the pair that is different"?
A11. Find the relation inside each pair; the pair whose internal relation differs is the answer (e.g., Dog:Puppy vs Horse:Stable โ one is animal:young, other animal:home).
Q12. Fastest safeguard against overthinking classification?
A12. Take the most obvious common category first; only dig for a hidden layer if two or more options fit that obvious one equally.