General Intelligence & Reasoning

Analogy questions give you one pair (of words, numbers or letter clusters) and ask you to repeat its relationship on a second pair. CGL typically asks 3-5. The skill is naming the relationship precisely — as a one-line sentence or a shift tuple — and testing every option against it.

Classification questions list four or five similar items and ask which one does not belong. The hidden rule — a shared category, number property or letter structure — must hold for every option except one. CGL asks 2-4 per shift, drawn from number properties, letter clusters, word categories and pairs.

Series questions ask for the next or missing term of a number or letter sequence, or the term that spoils a given sequence. CGL asks 4-5 per Tier 1 shift. The universal tool is the difference ladder; the recurring families are AP, ×k±r, squares/cubes, interleaved pairs and stepped letter tracks.

Coding questions reveal how a word (or sentence) is transformed into a code and ask you to apply the same transformation to a new word. CGL asks 3-5 per shift. Rules are mechanical: letter shifts, opposites, reversal, position sums, renames. Compute the rule from the sample pair, then apply it without creativity.

Operator-substitution puzzles: signs are redefined or interchanged, equations must be balanced, and the correct equation picked from four. The arithmetic is primary-school level — marks are lost to BODMAS slips and half-applied substitutions, not to difficulty. A disciplined rewrite-then-evaluate habit makes this one of the safest scoring topics in the paper.

Figure-based puzzles — paired grids, single grids, circles — where one cell is hidden behind a numerical rule. Every rule connects the cells of a row (or segment), and the same rule is demonstrated by at least two complete instances. The winning habit is two-line verification: fit the rule on one complete row, confirm on the other, then apply. Digit-sum and digit-reversal variants reward students who switch families quickly when plain arithmetic stalls.

One of the highest-yield reasoning topics: family-tree puzzles, pointing-at-photograph statements and symbol-coded chains, all solvable by drawing a small tree with parent-to-child arrows. Marks are lost to direction reversals and assumed genders, never to difficulty. A fixed decoding habit — inside-out for pointing questions, pairwise for coded strings, top-down for statement lists — makes nearly every question here a 30-45 second correct answer.

Compass-walking puzzles: multi-leg journeys with left/right turns, questions on the shortest distance back, the direction faced, the endpoint's bearing from the start, and shadow-based facing. Everything reduces to two signed running totals plus one Pythagoras step, and most answers are 3-4-5 multiples. Shadow questions add one conversion line (morning shadow West, evening shadow East). A written trace of headings and components turns the whole topic into bookkeeping.

Rank-conversion and ordering puzzles: ranks from either end, totals from two ranks, people between two positions, queue shifts, and taller/heavier comparison chains. Two formulas (n + 1 − p and a + b − 1) plus the between-count run the entire numeric half; the comparison half is inequality chaining into one unique order. Speed comes from drawing a mini-row and from the built-in check a + b = n + 1.

Seating puzzles — linear rows and circular tables — are the highest-value reasoning block in CGL, appearing as a linked set of 3-5 questions. Mastery is mechanical: pin absolute seats (ends, middle, opposite) first, chain the fixed-distance and immediate-neighbour clues as arrows, apply negative clues last, and read the answer off the finished sketch. Circular variants hinge on one convention — facing centre, left = clockwise — and flip when people face outward.

Syllogism questions hand you 2-3 statements (All / No / Some / Some-not / Only-a-few) as true and ask which conclusions follow. The reliable method is minimal-diagram plus counter-example hunting: draw the least-committal picture, then try to break each conclusion — a conclusion follows only if no valid diagram breaks it. Modern twists (possibility conclusions, either-or complementary pairs, 'only a few') all reduce to the same discipline with one twist: possibility needs just one supporting diagram, and either-or needs the two conclusions to be exact negations.

Two question families: (1) region arithmetic — count union, intersection, only-regions, exactly-two, neither using inclusion-exclusion over labelled Venn regions; (2) relation selection — decide whether categories nest (definitional necessity), stay apart (impossibility) or intersect (possible but not forced). The arithmetic family is pure formula discipline; the relation family is two quick tests per pair. Label every region before computing and both become near-90% accuracy topics.

Arranging words as in a dictionary, alphabet-position arithmetic (EJOTY anchors, from-right = 27 − p, shifts), modified-alphabet sequences, and the dictionary rank of a word among its letter arrangements. The whole topic runs on one rule — the first differing letter decides, and a prefix comes before its extensions — plus small, careful counting on paper.

Judge what a short statement definitely supports: conclusions (forced or clearly implied by the statement alone), assumptions (what the speaker must take for granted — use the negation test) and courses of action (practical, proportionate responses). The traps are predictable: extreme words, outside facts, overgeneralisation, averages read as limits, and 'only X are eligible' misread as a guarantee.

Count triangles, squares, rectangles or straight lines in a figure. Recognise the standard shapes first (square with diagonals = 8, with midlines too = 16; apex lines (k+1)(k+2)/2; grid rectangles C(m+1,2)C(n+1,2); squares 1² + … + n²), and count anything irregular size by size so nothing is missed or counted twice.

A vertical mirror reverses left-right (character order reversed and every character flipped); water reverses top-bottom (order kept, characters upside down). Know the symmetric letters for each (AHIMOTUVWXY for mirrors, BCDEHIKOX for water), use 11:60 − time for clock mirrors, and track one asymmetric feature in figure questions.

A folded sheet is punched or cut, and you must predict the unfolded pattern. Unfold in reverse order, reflecting every hole across each fold line; count layers first (p × 2ⁿ for full folds) to eliminate options, then match positions. Diagonal folds swap the two coordinates; partial folds double only what the flap covers.

Find a small figure hidden inside a bigger one exactly as drawn — same size and orientation (rotation is not allowed). Anchor on the most distinctive stroke, usually a slant, and reject options that are flips or rotations of the true part.

Dice questions ask for opposite faces from several positions (adjacent-four, two-common-faces and same-position rules), or from an unfolded net (skip one; ends of a Z). Painted-cube questions use 8, 12(n−2), 6(n−2)² and (n−2)³.

A row of figures changes by fixed rules — an element moves round the border, an arrow rotates, a shaded quarter turns, or a count grows. Track each element separately, find its step, and pick the only option that satisfies every rule.