Abstract Reasoning and Pattern Recognition
What is actually tested
Abstract reasoning items present shapes and ask you to infer the rule. Formats include matrices (complete the missing cell), series (what comes next), analogies (A is to B as C is to ?), and odd-one-out. Rules are built from a finite set of attributes: number of sides or elements, size, position, rotation, shading or fill, line style, symmetry, and the presence of a small accompanying mark.
The recurring traps
The main danger is overfitting a rule to one row. Check a candidate rule against every row and column before committing. The second is the double rule: many matrices apply one rule across rows and a different one down columns, so a single-attribute solution is incomplete. The third is the attractive distractor that satisfies one attribute but violates another. The fourth is assuming progression rather than alternation - a sequence may flip between two states rather than advance steadily. The fifth is ignoring rotation direction and amount: a shape turning 90 degrees clockwise each step looks identical to one turning 270 degrees anticlockwise for the first two steps.
Worked example
A row shows a triangle, a square and a pentagon, each with one dot, then two dots, then three. The row rule is sides increasing by one; the column rule is dot count increasing by one. Any option must satisfy both, so an option with four sides and three dots is wrong for this cell even though it fits the column.
Rule of thumb
Isolate one attribute at a time and test it across the whole grid, writing the values down if the layout is complex. Only combine attributes once each has individually survived every row and column.