Classification (Odd One Out)

Every letter-group question looks like a puzzle about spelling, but letters carry no meaning. So the examiner can only hide one thing in them: the distance between letters in the alphabet. He builds three groups from the same hidden list of gaps and then spoils one of them. Your job is to find the list, not to read the letters.

First fix what the exam asks. The options are groups of three or four letters, such as ACE or PIR. The newer papers add a note: "the odd one is not based on consonants or vowels, or on their position". That note is a signal to ignore vowels and go straight to gaps. Sometimes the four options are just single letters, and then the same logic works with positions.

The method has two steps. Step 1: write the position of each letter. Step 2: write the gaps between neighbouring letters. For ACE the positions are 1, 3, 5 and the gaps are +2, +2.

[[fig:0]]

In the picture ACE, GIK and MOQ all have gaps (+2, +2), but SUX has (+2, +3). So SUX is odd. Notice that SUX also contains a vowel, U. That vowel is a decoy; the gap list is the real property.

Now the exam-level refinement. The setter often keeps the first gap equal in all four options so that a quick look finds nothing. Do not compare whole lists then; compare one gap at a time across the four options. Take gap 1 of every option, then gap 2, then gap 3. The column in which three values agree and one differs gives the answer, and the other columns may all be different. Compare gap by gap, column by column.

If every column fails, look for opposite letters (two letters that add up to 27, like A and Z). Note which positions hold such a pair; the odd group breaks that structure. Items that show a word and its rearrangements (LIFE, IFEL, FELI) are different again: watch how the letters move.

[[fig:1]]

Single letters. When the options are four single letters, write the positions and test in this order: vowel or consonant, odd or even position, multiple of 3 or 5, equal gap, opposite letters. In the picture D, G, J, L have positions 4, 7, 10, 12. The first three are 3 apart and L breaks the chain. Parity would also give a 3 and 1 split, but it points to G, which belongs to the chain. Only L breaks the chain, and with no vowel in the set, the gap test is the one to trust.

Now each option is a pair of clusters joined by a dash, for example QM – YF. The left cluster is turned into the right cluster by a coding pattern: every letter moves forward or backward by a fixed number of steps, and each position has its own step. Three options obey the pattern. One breaks it at one position.

The logic is simple. The pattern acts letter by letter, so the whole item is only two small questions: how far did the first letter move, and how far did the second letter move? That distance is the shift = position of the right letter − position of the left letter.

Step 1: write the shift of the first letter for all four options.
Step 2: write the shift of the second letter for all four options.
Step 3: the column in which three values agree and one differs gives the odd pair.

[[fig:0]]

In the picture, the first letters move +3 in every option. The second letters move −2 in three options but −1 in option (c), so DY – GX is odd. You never needed to decode the clusters or look at vowels.

Wrap round the alphabet. Shifts are counted round a circle of 26 letters, so +19 and −7 are the same shift (19 − 26 = −7). In HD – PW the second letter goes from D (4) to W (23), which is +19, and that is exactly the same as the −7 of QM – YF. Never call two options different only because one shows +19 and the other −7. Convert every shift to the same side (add or subtract 26) before you compare.

A second lens: the gap inside each cluster. When the shifts are fixed, the gap inside the right cluster is also fixed once you know the gap inside the left cluster. So you can also compare the inner gaps: in KO – RN the left gap is +4 and the right gap is −4, while the other options have +4 and −5. Use whichever view is quicker for the numbers in front of you.

Older papers gave single two-letter clusters such as EI, MQ, TX, CH and asked for the odd one. The logic is the same, with only one gap to compare.

Sometimes both clusters contain the same letters in a different order (PORT – TROP). Then ignore shifts and compare how the letters were rearranged: here the word is simply reversed. Compare the shift of the first letters, then of the second letters, and keep the column that cuts off exactly one option.

Each option is a pair of numbers, written 2 : 9, (3, 28) or 14 – 121. The second number is made from the first by a hidden pattern. Three pairs obey the pattern and one does not. The papers usually add a note: operations act on the whole number, not on its digits.

The logic is that the setter picks a pattern that can be done in the head, applies it three times, and spoils one result. So you do not need a clever idea; you need to guess the pattern from the size of the jump, test it on two pairs, and run it on the rest. The table tells you what to guess.

[[fig:0]]

In the picture the pairs are 2 : 9, 3 : 28, 4 : 65 and 5 : 124. The jump is about a cube, so try n³ + 1: 2³ + 1 = 9, 3³ + 1 = 28, 4³ + 1 = 65, but 5³ + 1 = 126, not 124. So 5 : 124 is odd.

Test the last digit first. In patterns such as n² + c, n³ ± c or × k, the units digit of the answer depends only on the units digit of the number. So you can reject a pair without any calculation. If the pattern is n² + 1, a number ending in 6 must give a result ending in 7.

Be careful of two things. A pair can fit a second pattern by coincidence, so confirm your pattern on three pairs before you call the fourth odd. And if the note says "whole number", do not break the number into digits. When both numbers have the same digits in a different order (317 – 713), reversal is the pattern. Guess the pattern from the size of the jump, then confirm it on three pairs.

Four single numbers are given and three of them share a property. Numbers have a great many properties, so the real skill is not knowledge but order: test the cheap properties first, and stop the moment one property leaves three numbers on one side and one on the other.

The logic is to let the numbers themselves tell you where to look. The table shows what to notice and what to test.

[[fig:0]]

In the picture all four numbers are squares (11², 13², 15², 17²), so "square" cuts off nobody. Now look at the bases: 11, 13 and 17 are prime, 15 is not. So 225 is odd. Whenever every option belongs to one broad family (all squares, all primes, all multiples of something), go one level deeper and look at the roots, the bases or the multipliers.

For digit-based sets, read the whole number. In 121, 232, 555, 766 three numbers read the same backwards and 766 does not. In 8765, 7654, 6543, 3456 the digits of three numbers fall one by one, and 3456 rises. In 234, 243, 324, 312 the first three use the digits 2, 3, 4 and 312 uses 1, 2, 3.

A double test helps with numbers like 8, 18, 32, 50, 72 and 98, which are 2 times a square: their doubles are perfect squares. And remember that last-digit patterns only help you reject: a square cannot end in 2, 3, 7 or 8, but ending in 1, 4, 5, 6 or 9 does not make a number a square. Let the numbers tell you the test, then keep the property with exactly one cross.

Each option is a group of three numbers, written (4, 11, 25), 11 : 21 : 121 or 6 – 18 – 54. Inside a group the three numbers are tied by one link. Three groups share the link and one breaks it. Recent papers also write the group as an equation, such as 4 # 3 = 12, where the symbol hides an operation; the idea is exactly the same, with a, b and c replaced by the numbers on either side of the equals sign.

The logic is that a link can live in only two places: between neighbours (the next number is made from the one before it) or among all three (the third number is made from the first two). The setter keeps the link simple, so you find it from one group, confirm it on a second, and then look for the group that fails.

Step 1: for every group write the two gaps (b − a and c − b) and the two ratios (b ÷ a and c ÷ b). These four numbers settle most items.
Step 2: equal gaps mean an AP; equal ratios mean a GP; gaps or ratios that are the same inside every group but fail in one group give the answer.
Step 3: if neither works, solve the chain next = k × previous + c from the first two numbers of one group.
Step 4: if the numbers go up and down, test the third number as a function of the first two (a + b, a × b ± c, b² ± a).

[[fig:0]]

In the picture, take (4, 11, 25). From 4 to 11, × 2 + 3 works (2 × 4 + 3 = 11), and 2 × 11 + 3 = 25 confirms it. The groups (5, 13, 29) and (7, 17, 37) obey the same link. The last group, (9, 21, 46), gives 21 correctly, but 2 × 21 + 3 = 45, not 46. So (9, 21, 46) is odd.

A very common hiding place is a multiplier that depends on the first number. In (3, 12, 48), (4, 20, 100), (6, 42, 294) the multipliers are 4, 5, 7, each one more than the first number, used twice. A group such as (5, 30, 150) uses × 6 and then × 5, so it breaks the link. If the ratios are not the same across groups, check whether they follow the first number.

Equations. For 4 # 3 = 12, 5 # 3 = 15, 5 # 4 = 20, 8 # 1 = 24 find the operation from the two cleanest equations: 4 × 3 = 12 and 5 × 3 = 15 point to a product. Then 8 × 1 = 8, not 24, so the last equation is odd.

Rows of items. Sometimes the options are sets of words, letters or numbers with no order. Then only presence matters: find the element that occurs in exactly three rows. The row without it is odd. An element in only two rows is a decoy. Write the gaps and ratios for every group, find the link on one group, and let the group that fails be the answer.

A word odd-one-out question looks like a list, but it is really a family photo with one stranger in it. Three words belong to one class (flowers, planets, birds, states, fruits, vehicles). The fourth belongs to a different class. Your job is to name the family and then spot the stranger.

Read all four words before you decide, not just the first two. Then say aloud what most of them are: "three of these are flowers". Now test the fourth word against that sentence. If it fails, it is the answer.

[[fig:0]]

In the picture, Tulip, Jasmine and Marigold sit inside the flower bubble. Banyan is a tree, so it stays outside. Notice that the examiner is not asking for the "best" family. He wants the family that holds exactly three of the words.

Sometimes all four words fit one broad family (all four are plants). Then go one step narrower and ask which smaller family holds three of them. Tulip, jasmine and marigold are grown for their flowers; the banyan is grown for its shade and size. Name the family out loud, then test the fourth word against it.

Use the plain, everyday meaning of the family and do not hunt for rare exceptions. In the list Train, Bus, Car, Bicycle, three run on an engine or motor, and the pedal-driven bicycle is the stranger. In Patna, Raipur, Hyderabad, Agra, three are state capitals and Agra is not.

What the latest papers use. The four words are everyday nouns and the exception is usually one of a handful of stock strangers: a state among cities (Delhi, Mumbai, Kolkata, Rajasthan); a city among state capitals (Patna, Raipur, Hyderabad, Agra); a cycle among engine-driven vehicles (Train, Bus, Car, Bicycle); a country that does not border India among those that do (Bhutan, Nepal, China, Singapore); an internal organ among parts you can see (Neck, Ear, Lips, Heart); an organ among household objects (Table, Fan, Bed, Kidney); a tool for eating among weapons or cutting tools (Gun, Knife, Sword, Spoon); a solid with no curved surface among those with one (Cube, Sphere, Cylinder, Cone).

When the family does not come at once, ask a second question about each word: What is it used for? Where is it found? Is it inside or outside the body? Does it run on an engine? Does it touch India? One of these questions will split the words 3 and 1.

In this type the family is easy to see, but the exception is hidden behind a famous misconception. The question tests whether you know the scientific class of each word, not where it lives or how it looks.

Take Whale, Bat, Dolphin and Shark. A whale and a dolphin live in water, a bat flies in the air, and a shark swims. If you group by habitat, you get a confusing mess. Group by class instead. Whale, dolphin and bat are mammals: they give birth to live young and feed them milk. A shark is a fish. So Shark is the stranger.

[[fig:0]]

The picture shows the trap. "Lives in water" and "is a mammal" are two different circles. Whale and dolphin sit in both, the shark sits only in the water circle, and the bat only in the mammal circle. The examiner decides the odd one by the property that excludes exactly one word: here "is a mammal" excludes only the shark.

For every word in a fact question, recall one fact: its class (mammal, fish, bird), its type (ocean or sea), whether it is a capital, whether it is hardware or software, whether it is renewable. Then group the words by that fact.

Typical pairs of fact and exception: a sea among oceans (Mediterranean), a mammal among birds (bat), a metal among non-metals (iron among the halogens), coal among renewable energies, a cyclone among tectonic hazards, a desert among rivers (Sahara), and software among hardware (Notepad). Decide by the class of the word, never by its habitat or looks.

If you do not know one fact, use the other three: if you are sure of three words, the fourth is the answer.

General-knowledge sets in the latest papers. Besides science families, 2025 and 2026 papers use fact sets from general studies. Rivers with a desert (Amazon, Nile, Ganges, Sahara). Metals with a non-metal (Gold, Silver, Copper, Diamond: diamond is carbon). Mountain peaks with one from another range (Anamudi, Agasthyamalai and Doddabetta lie in the Western Ghats (Doddabetta in the Nilgiris), while Nanda Devi is in the Himalaya). Articles of the Constitution (Articles 15, 16 and 18 belong to the Right to Equality, while Article 352 is about Emergency; Articles 15, 24 and 32 are Fundamental Rights, while Article 368 gives Parliament the power to amend the Constitution). For such sets, recall one anchor fact per word (which group, which part, which list) and let the three that agree outvote the fourth.

Now every option is a pair of words written with a colon, such as Sitar : String. The two words in a pair are linked by a relation. Three pairs carry the same relation, and one pair carries a different one. The odd pair is the one whose relation is different.

Do not read the pairs as four separate topics. Read each as a sentence: "X is a ___ of Y". Write the missing relation word next to every option. The relation word that appears only once marks the answer.

[[fig:0]]

In the picture every pair is "instrument : the family it belongs to". Sitar is a string instrument, flute is a wind instrument and tabla is a percussion instrument, so these three read correctly. A violin, however, is a string instrument, so "Violin : Wind" breaks the relation and is the odd pair.

The common relations are: synonym (Big : Large), antonym (Hot : Cold), tool and its job, instrument and what it measures, thing and its category, part and whole, worker and workplace. In one question, three pairs may be synonyms and one an antonym, or the other way round.

Two cautions. First, the relation must run in the same direction in all four options. "Doctor : Hospital" and "School : Teacher" are both about work places, but the second runs backwards, so it is the odd pair. Second, an odd-looking pair may be perfectly consistent. Do not choose by topic; choose by the relation word. Name the relation for each pair; the relation that stands alone is the answer.

Two kinds of word-pair items. The first kind asks for a different relation: three synonym pairs and one antonym pair, or "Dog : Bark, Cat : Meow, Cow : Moo, Snake : Slither" where three give the sound and one gives the movement. The second kind keeps the same relation in all four pairs and hides one false fact: country and currency (South Korea has the Won, not the "Rupian"), shape and number of sides (a hexagon has 6 sides, not 8), game and players in a team (a football team has 11, not 10). In the second kind you must check each fact; the pair whose fact is wrong is the answer.

Here every option mixes letters with numbers, and sometimes symbols as well. It looks messy, but the idea is always the same: a letter is a number in disguise, namely its position in the alphabet. Convert the letters and you are back to a number problem. The exam shows it in four forms.

1. Single tags such as C-3, R-19 or Z9$: the number is linked to the position of the letter. Three tags obey; one has a wrong number.
2. A sequence such as A3, C5, E7, G9, I10: letters and numbers move together in steps. One item breaks the steps.
3. Triplets such as P22 : N20 : L18: three items in a group, with the same pattern repeated in every group. Symbols (@, #, $) usually ride along in a fixed cycle.
4. Long codes and pairs such as A239X – B478Z: the left item becomes the right item by an operation on the whole number and shifts on the end letters. Also IDs and roll numbers, where only the shape matters (digit-letter-digit-letter).

[[fig:0]]

For single tags, convert each letter to its position and compare with the number. In the picture, C is 3rd and the tag is 3, H is 8th and the tag is 8, M is 13th and the tag is 13. R is the 18th letter, but its tag says 19, so R-19 is odd. The number need not equal the position; it can be twice the position, the position plus a constant, or 27 minus the position (the opposite letter). If every tag fits, the property is something else (even tags, vowels).

For triplets, keep two lists for each group: the steps of the letters and the steps of the numbers. A good group repeats both lists. The odd group changes one of them.

[[fig:1]]

In the picture all groups have letter steps (+2, +2) and number steps (+3, +3) except (b), where the letter step is +2 then +3. Check the symbols last. If the letter and number lists agree in all four groups, the symbols may be the difference: a cycle such as @, #, $ that suddenly starts differently.

For long codes, do the arithmetic first, because numbers are quicker than letters. In A239X – B478Z the number doubles (239 × 2 = 478). Then check the shift of the end letters, column by column, exactly as in letter-cluster pairs. Convert letters to positions, then compare the steps of letters and the steps of numbers separately.

Here each option is a word with its code, such as SMILE : TNKLI, and the question says "four of the following five are alike in a certain way" (or three of four). All the items but one are coded by the same pattern; one breaks it. You only have to compare, not decode a new word, so this type is quick once you know what to compare.

The logic is that the paper-setter picks a very simple pattern and spoils one item. In the recent papers the pattern is almost always a fixed shift: every letter moves +1 (BRING : CSJOH), sometimes +2, or the shift changes by position (+1, +2, +3). A second form gives a word : number pair where the number is the sum of the alphabet positions of the letters.

Step 1: test the commonest pattern, +1 for every letter, on all the items. The items that pass are the group.
Step 2: if no pattern is clear, take the shift at each position of the first clean item and test it on the others.
Step 3: the item that differs at any position is the odd one.

[[fig:0]]

In the picture SUN should become TVO if every letter moves +1. The given code is TVN: the first two letters moved +1, but N stayed at N, a shift of 0. So this item breaks the coding.

Keep three habits. Check every position; the odd item is often right at two positions and wrong at one. Remember the wrap: after Z comes A, so Z + 1 = A and A − 1 = Z, and such a letter is legitimate. Do not stop once the first letters agree.

For word : number items, add the alphabet positions of the letters (S 19 + T 20 + O 15 + N 14 + E 5 = 73). The item whose total differs from its number is odd. Check the units digit of your sum first: it is faster than adding every item completely. Test the commonest shift first, then compare position by position.